Imported from ../bash-2.01.tar.gz.

3 files changed, 1238 insertions, 7912 deletions

diff --git a/CWRU/PLATFORMS b/CWRU/PLATFORMS
index e479a296..4d2b4844 100644
--- a/CWRU/PLATFORMS
+++ b/CWRU/PLATFORMS
@@ -6,10 +6,13 @@ By chet:
 SunOS 4.1.4
 SunOS 5.5
 BSDI BSD/OS 2.1
-Motorola SVR3.2
-FreeBSD 2.1.5
+FreeBSD 2.1.7
+NetBSD 1.2
 AIX 4.2
+AIX 4.1.4
 HP/UX 9.05, 10.01, 10.10, 10.20
+Linux 2.0.29 (libc 5.3.12)
+Linux 2.0.4 (libc 5.3.12)
 
 By other testers:
 
diff --git a/CWRU/POSIX.NOTES b/CWRU/POSIX.NOTES
index 304289e1..af3acbcc 100644
--- a/CWRU/POSIX.NOTES
+++ b/CWRU/POSIX.NOTES
@@ -77,6 +77,9 @@ The following list is what's changed when `POSIX mode' is in effect:
  20. Assignment statements preceding POSIX.2 `special' builtins persist
      in the shell environment after the builtin completes.
 
+ 21. The `export' and `readonly' builtin commands display their output
+     in the format required by POSIX.2.
+
 
 There is other POSIX.2 behavior that Bash does not implement.
 Specifically:
diff --git a/CWRU/changelog b/CWRU/changelog
index d09a5b5a..331c9e60 100644
--- a/CWRU/changelog
+++ b/CWRU/changelog
@@ -1,8399 +1,1719 @@
-				    6/2
-				    ---
-lib/readline/readline.c
-	- fixed an off-by-one error in the kill ring reallocation code
-	  in rl_kill_text
-
-Makefile
-	- replaced instances of /bin/sh with $(SHELL)
-
-Makefile, cpp-Makefile, documentation/Makefile
-	- added a working `uninstall' target
-
-[1.14.0 FCS release made available for FTP]
-
-				    6/3
-				    ---
-README
-	- added note about building with gcc, same as in Makefile
-
-documentation/Makefile
-	- some versions of make don't understand `$*' in a regular recipe;
-	  replace with features.dvi in the recipe for features.ps
-
-				    6/4
-				    ---
-subst.c
-	- fix up the calls to string_extract_double_quoted and 
-	  string_extract_single_quoted in char_is_quoted so the initial
-	  value of the index is *after* the opening quote character
-	- make sure we only return 1 from char_is_quoted if the index
-	  into the string after a call to string_extract_double_quoted or
-	  string_extract_single_quoted is *greater than* `eindex'
-
-lib/readline/complete.c
-	- change the order and sense of the quoting tests in
-	  rl_complete_internal so that the expensive char_is_quoted
-	  is only called if `scan' actually is a word break character
+				12/23/1996
+				----------
+[bash-2.0 released]
 
-shell.c
-	- fixed a typo in the test for a restricted shell
+				   12/27
+				   -----
+configure.in
+	- don't automatically use GNU malloc on FreeBSD, the builtin one
+	  is supposed to be better (but doesn't have the debugging hooks)
 
 builtins/exec.def
-	- need to include flags.h
+	- check that full_pathname(command) doesn't return NULL because of
+	  inaccessible parent directories
 
-				    6/6
-				    ---
-make_cmd.c
-	- make sure that we don't try to walk down a null command tree
-	  in connect_async_list (tickled by `(command &) &')
-	- if a command has the CMD_WANT_SUBSHELL bit set in its flags,
-	  don't even try to walk the command tree and move the `&';
-	  just connect the command with a null command using `&'
+support/config.sub
+	- recognize OpenBSD
 
 execute_cmd.c
-	- make sure we don't try to reference a command struct without
-	  first checking it (case '&')
-
-				    6/7
-				    ---
-machines.h
-	- fix a typo in the NeXT/i386 description; change the NeXT description
-	  to #define HAVE_RESOURCE if not already defined rather than putting
-	  -DHAVE_RESOURCE into the SYSDEP_CFLAGS
+	- broke the code that opens a file for output redirection when in
+	  noclobber mode out into a separate function: noclobber_open().
+	  This tries to avoid race conditions and file replacement between
+	  stat(2) and open(2)
 
-				    6/9
-				    ---
-Makefile
-	- make sure all of the rules that reinvoke `make' have the
-	  `-f bash-Makefile' before the assignment statements; some
-	  versions of make like it that way
-
-variables.c
-	- make sure that `interactive' is set to 0 before evaluating the
-	  string containing an exported function obtained from the
-	  environment, so that it does not try to execute PROMPT_COMMAND
-	  (this may not be the best fix)
-
-				   6/13
-				   ----
-documentation/Makefile
-	- make sure all of the directories exist before trying to install
-	  doc files into them
-
-lib/readline/history.c
-	- add a missing O_TRUNC to the open call for writing in
-	  history_truncate_file
-
-trap.c
-	- run_interrupt_trap should only try to run the trap command if
-	  the value is not IMPOSSIBLE_TRAP_HANDLER
-
-Makefile
-	- add `realclean'
-
-lib/readline/complete.c
-	- do the same kind of double-quoting a replacement string if the
-	  user supplies the opening double quote as we would if we were
-	  adding both quotes ourselves
-
-variables.c
-	- fixed the variable initialization so that history_control/HISTCONTROL
-	  can be inherited from a parent shell
-
-INSTALL, README, cpp-Makefile, documentation/bash.1, documentation/readline.3
-	- ai.mit.edu -> prep.ai.mit.edu
-
-shell.c
-	- fixed a problem with setting no_line_editing to the result of the
-	  check for running inside emacs, thereby losing any value
-	  initialized by the `-nolineediting' flag
-
-cpp-Makefile
-	- `make distclean' will now remove the `installed-bash' link
+subst.c
+	- make sure pat_subst does not run off the end of its return
+	  string when copying the unmatched portion of the input string
 
-print_cmd.c
-	- fixed a problem with undefined variables when HAVE_VARARGS_H is
-	  not defined
+				   12/30
+				   -----
+doc/Makefile.in
+	- don't install readline.3 by default
 
-				   6/14
-				   ----
-lib/readline/history.c
-	- fixed an error in the csh history expansion code so that the
-	  `-y' word designator now expands to `0-y' rather than `1-y'
+lib/tilde/tilde.c
+	- removed an unnecessary check for string[i] being non-null in
+	  tilde_find_suffix
 
-lib/readline/isearch.c
-	- changed an absolute check for a character falling within the
-	  ASCII 32-126 range (printable chars) with checks for CTRL_P
-	  and META_CHAR and a check against the value RUBOUT
-	- changed a `break' to a `continue' so that the first non-matching
-	  character in the search string does not cause the search to
-	  end abruptly
-	- initialize prev_line_found to 0 at the top of rl_search_history
-	  to avoid duplicate history lines being saved across searches
-
-lib/readline/rltty.c
-	- consolidated repeated code for setting special characters into
-	  `SET_SPECIAL' defines
+				   12/31
+				   -----
+support/config.{sub,guess}
+	- merged in changes from config.sub in autoconf-2.12 distribution
 
 lib/readline/readline.c
-	- include <sys/ioctl.h> if VSTATUS is defined
-	- add bindable command rl_tty_status (unbound by default)
-
-lib/readline/funmap.c
-	- assign bindable command name `tty-status' to rl_tty_status
-
-INSTALL
-	- add note about compiling with gcc, same text as README
+	- in rl_newline, only call _rl_vi_done_inserting and _rl_vi_reset_last
+	  if readline is currently in vi mode
 
 lib/readline/display.c
-	- many changes and tweaks to make redisplay work better when the
-	  prompt has invisible characters.  These changes are non-optimal
-	  in that the prompt is redrawn more than it needs to be, but
-	  things are a hell of a lot better than they were
-
-				   6/15
-				   ----
-documentation/Makefile
-	- make a variable NROFF that people can set to `groff -Tascii' if
-	  they don't have real nroff
-	- changed the suffix rules to use $<
+	- corrected a misuse of inv_lbreaks where vis_lbreaks was wanted in
+	  the code that decides whether the cursor is at the beginning of
+	  an otherwise-empty line
 
+				    1/2
+				    ---
 support/bashbug.sh
-	- if rmail doesn't exist or fails, save the bug report in
-	  ~/dead.bashbug
-
-execute_cmd.c
-	- in setup_async_signals, only ignore SIGINT and SIGQUIT if job
-	  control is not active.  If it is active and the job is restarted,
-	  SIGINT remains ignored, and the now-foregrounded job is not
-	  interruptible
+	- fixed a typo, thanks to eggert@twinsun.com
 
-subst.c
-	- fixed up a problem with char_is_quoted that caused backslash-
-	  escaped characters to cause incorrect results
-
-tests/run-dollars, tests/dollar-at.sh, tests/dollar-star.sh
-	- since the error messages produced by `cat' vary, changed `cat'
-	  to `recho' and updated the correct answers file
-
-machines.h
-	- fixes to CRAY_STACKSEG_END definitions for different versions of
-	  Unicos on the YMP (from Bill Jones)
-	- Motorola SVR4 machines have getcwd() and should not undef
-	  HAVE_GETCWD
-	- on hpux 9.x, don't try to link with -lPW if compiling with gcc
-	  (for alloca)
-
-parse.y
-	- an ugly fix for a compiler problem with structure assignment on
-	  the cray
-
-				   6/16
-				   ----
-builtins/wait.def
-	- replaced a call to sscanf with a validity check using all_digits()
-	  and a call to atoi, since a pid_t is not necessarily an int, and
-	  the sscanf clobbered the stack frame on systems where it is shorter
-	  than an int (e.g., SCO)
-
-lib/readline/display.c
-	- since META_CHAR and the other macros only work with unsigned
-	  chars, make rl_character_len convert its argument to an unsigned
-	  char before testing it
-
-documentation/Makefile
-	- use $(INSTALL_DATA) instead of $(CP) to install the man pages and
-	  info files
-
-cpp-Makefile
-	- use INSTALL_PROGRAM and INSTALL_DATA to install binaries and
-	  documentation; pass both values to installs in subdirectories
+aclocal.m4
+	- new test, BASH_STRUCT_DIRENT_D_FILENO, testing for d_fileno member
+	  of struct dirent
 
-				   6/18
-				   ----
-builtins/ulimit.def
-	- compensate for systems which define RLIMIT_OFILE instead of
-	  RLIMIT_NOFILE, or don't provide such a compatibility define
-	  themselves
+configure.in
+	- call BASH_STRUCT_DIRENT_D_FILENO
 
-shell.c
-	- make maybe_execute_file check for directories and print an
-	  appropriate error message, since it's doing an fstat anyway
+config.h.in
+	- new #define for STRUCT_DIRENT_HAS_D_FILENO
 
-support/mksysdefs
-	- added support for a `-s srcdir' option so it can find
-	  cpp-Makefile if ansi-Makefile is to be created
+lib/posixheaders/posixdir.h
+	- only define d_fileno as d_ino if STRUCT_DIRENT_HAS_D_INO is defined
+	  and STRUCT_DIRENT_HAS_D_FILENO is not defined.  This fixed the
+	  problem of compiling getcwd.c on SunOS4.1.4 with cc
 
-Makefile
-	- call mksysdefs with -s $(srcdir)
+				    1/3
+				    ---
+lib/readline/complete.c
+	- fix a memory-used-after-freed bug reported by Andreas Schwab
 
-jobs.c
-	- add the magic #undef lines to avoid redefinition warnings on
-	  SunOS 4 only
+configure.in
+	- call BASH_CHECK_SOCKLIB only if getpeername is not found in libc
+	  (ac_cv_func_getpeername = no)
 
-				   6/20
+				   1/13
 				   ----
-cpp-Makefile
-	- install `bashbug' with `make install'
 
-trap.c
-	- make sure that `interactive' is set to 0 when running trap
-	  commands
+builtins/getopt.h
+	- change #define guard to _SH_GETOPT_H to avoid similar guards in
+	  /usr/include/getopt.h (Dec OSF/1 4.x, for example)
 
-builtins/umask.c
-	- fixed typo in usage error message
+variables.h
+	- fix assign_array_from_string so that it rejects attempts to assign
+	  to readonly variables
 
 subst.c
-	- fix process_substitute to set subshell_environment
-
-jobs.c, nojobs.c
-	- only mess with the terminal settings for an interactive shell
-	  that is not in a subshell environment
-
-				   6/21
-				   ----
-lib/readline/history.h
-	- add extern declaration of history_get
-
-builtins/fc.def
-	- make history replacement when using `r' or `fc -s' obey the
-	  setting of HISTCONTROL
-
-general.c
-	- in canonicalize_pathname, preserve a double // at the start
-	  of an absolute pathname, since that means something special
-	  for the network directory system
+	- fix verify_substring_values to handle offsets that are past the
+	  end or before the beginning (in the case of a negative offset)
+	  of the expanded variable value
 
-README, INSTALL
-	- updated information about submitting bug reports
-
-lib/readline/vi_mode.c, lib/readline/isearch.c
-	- make sure unistd.h is included before rldefs.h, if
-	  HAVE_UNISTD_H is defined
-
-				   6/24
-				   ----
-lib/readline/complete.c
-	- add `#' to the list of characters which cause a completed filename
-	  to be quoted
-
-execute_cmd.c
-	- be more careful about closing pipe file descriptors in do_piping;
-	  don't want to have `dup2(i, i); close(i);' problem
-
-lib/readline/{keymaps,readline}.h
-	- include local copies of include files if READLINE_LIBRARY is
-	  defined, otherwise include the `official, installed' versions
-	  using #include <readline/xxx.h>
-
-lib/readline/*.c
-	- define READLINE_LIBRARY before including any files
-	- include only `local' copies of include files using #include "xxx.h"
-	  rather than #include <readline/xxx.h>
-
-				   6/26
+				   1/14
 				   ----
-execute_cmd.c
-	- check for clobbering the bash input stream before closing a file
-	  descriptor due to an r_close_this redirection
-
-lib/readline/history.c
-	- made history_expand inhibit history expansion if the history
-	  expansion char is set to 0
+bashline.c
+	- fix a problem where any completion after a `cmd` command
+	  substitution would inappropriately attempt command completion,
+	  even if the previous command substitution was correctly closed
 
-lib/readline/chardefs.h
-	- moved savestring() definition to rldefs.h
-	- changed lowercase_p, uppercase_p, to_lower, to_upper defines to
-	  use <ctype.h> macros rather than assume ASCII
+builtins/evalstring.c
+	- unwind_protect remember_on_history even if the shell is not
+	  interactive, since history can now be used in scripts, and
+	  `source' will turn off interactive_shell before calling
+	  parse_and_execute
 
-lib/readline/bind.c, general.c, general.h
-	- use strcasecmp, strncasecmp instead of str[n]icmp if
-	  HAVE_STRCASECMP is defined
+jobs.c
+	- new function get_job_by_pid(pid, block) to translate a pid to
+	  a job number.  The block argument says whether or not to block
+	  SIGCHLD
 
-cpp-Makefile
-	- pass -DHAVE_STRCASECMP to builds in the libraries, primarily
-	  readline
+jobs.h
+	- new extern declaration for get_job_by_pid
 
-machines.h
-	- add HAVE_STRCASECMP to the entries for BSD/386, NetBSD, FreeBSD,
-	  and 4.4 BSD
+builtins/jobs.def
+	- call get_job_by_pid if an argument appears to be a pid rather
+	  than a jobspec
 
-builtins/hash.def
-	- add a fourth parameter to remember_filename, the initial value
-	  of times_found (0 if we're just looking it up for `hash', 1
-	  for the command execution code)
+configure.in
+	- configure --without-gnu-malloc automatically on MachTen 4.x
 
-execute_cmd.c
-	- call remember_filename with an initial value of 1 for times_found
+builtins/cd.def
+	- change to mindist() so that a best guess of `.' returns 3,
+	  which means not reasonable
 
-builtins/wait.def
-	- handle a null argument with an error message
+lib/posixheaders/memalloc.h
+	- changed hpux_9 to __hpux, since the new config stuff doesn't
+	  define hpux_9
 
-builtins/common.c
-	- parse_and_execute now takes a third parameter: the value for
-	  `interactive' while it is executing commands
+subst.c
+	- fix parameter_brace_patsub to handle null patterns (doesn't
+	  do anything)
 
-bashline.c, jobs.c, parse.y, shell.c, subst.c, trap.c, variables.c
-	- set the new third argument to parse_and_execute appropriately
+oslib.c
+	- slight change to bzero replacement
 
-builtins/eval.def, builtins/fc.def, builtins/source.def
-	- set the new third argument to parse_and_execute appropriately
+support/bashbug.sh
+	- changed TEMP to be /tmp/bbug.$$ as a sop to systems with 14-char
+	  file names
 
-builtins/help.def
-	- changed a call to strnicmp to strncmp when trying to find what
-	  to give help on; it seems more correct
+doc/bashref.texi
+	- add note to POSIX Mode section that the output of `set' when
+	  invoked without arguments is sorted when in POSIX mode
 
-				   6/27
+				   1/15
 				   ----
-machines.h
-	- cleaned up the SunOS section so it no longer relies on
-	  HAVE_SHARED_LIBS being defined; it uses SunOS4 and SunOS5
-	  instead
+support/recho.c
+	- a couple of changes prompted by a `gcc -Wall'
 
-support/mksysdefs
-	- define SYSDEF to be SunOS4 or SunOS5 depending on the output
-	  of uname rather than looking for ld.so
+subst.c
+	- changed ASSIGN_RETURN macro to use the do...while(0) idiom to
+	  avoid problems with its multiple statements
 
-				   6/29
-				   ----
-machines.h
-	- minor change to the ardent titan machine description
-	- move the ardent and stardent descriptions before the
-	  mips riscos description
+builtins/setattr.def
+	- a bad identifier given to readonly or export without an assignment
+	  statement (e.g. `readonly a[5]') is an error, but not an assignment
+	  error (i.e., return EXECUTION_FAILURE rather than EX_BADASSIGN)
 
-print_cmd.c
-	- ardent machines also need the extern declaration for printf
+tests/{{array,new-exp}.,tilde-}tests
+	- added `set +o posix' at the beginning to disable POSIX mode, which
+	  causes some of the tests to fail.  Some systems (e.g. LINUX-FT),
+	  set POSIXLY_CORRECT by default
 
-make_cmd.c
-	- connect_async_list should do its work only if the lists to be
-	  backgrounded are connected with `;'.  This makes `;' bind tighter
-	  than `&', so only the last job in the list is backgrounded.  All
-	  other lists should have the entire thing put in the background
+tests/test-tests
+	- added a warning if $UID is 0 to the effect that the test suite
+	  should not be run as root
+	- worked around the `noread' and `nowrite' tests failing when run
+	  as root
 
-parse.y
-	- added a function `print_prompt' to take care of displaying the
-	  prompt string if readline is not being used.  This fixes problems
-	  with the prompt being displayed before the status of completed
-	  jobs is printed
+test.c
+	- began removing the remains of the STANDALONE code, since test is
+	  included in GNU shellutils
 
-				   6/30
+				   1/16
 				   ----
-builtins/fg_bg.def
-	- `fg' and `bg' now print error messages if invoked when job control
-	  is disabled
-
-lib/readline/rltty.c
-	- if not compiled into the shell, make get_tty_settings get and set
-	  the window size.  This noop stops the process if it is started in
-	  the background
+lib/readline/{readline,display}.c
+	- a couple of slight changes to build on Win95 using djgpp (reported
+	  by x-aes@telelogic.se)
 
-lib/readline/readline.c
-	- provide a function version of savestring, if not being compiled
-	  into the shell, since the macro has been removed from the
-	  `public' header files
-
-lib/readline/readline.h
-	- provide all extern function declarations without checking whether
-	  VI_MODE or PAREN_MATCHING are defined.  It does not hurt to define
-	  them if they are not used and not in the library, and other
-	  applications using readline can't tell whether or not VI_MODE was
-	  defined when the library was compiled anyway
-
-				    7/1
-				    ---
-machines.h
-	- add #undef HAVE_DIRENT_H to the ardent titan description
-
-				    7/2
-				    ---
-lib/readline/chardefs.h
-	- removed META_P define, renamed CTRL_P to CTRL_CHAR
-
-lib/readline/bind.c, lib/readline/isearch.c
-	- changed instances of CTRL_P to CTRL_CHAR
-
-lib/readline/search.c
-	- include <unistd.h> before rldefs.h, if HAVE_UNISTD_H is defined
-
-lib/readline/readline.c
-	- declare PC, UP, and BC as extern rather than `local' to the
-	  readline library
+Makefile.in
+	- changed TERMCAP_LIBDIR to TERM_LIBDIR, so building the termcap
+	  library (if necessary) should work now
 
-				    7/5
-				    ---
 bashline.c
-	- implement command word completion inside of command substitution
-	  with a new function: `command_subst_completion_function'
-
-subst.c
-	- new function to help with command subst completion: unclosed_pair
-
-lib/readline/complete.c
-	- new variable rl_filename_quoting_desired, which can be set to 0
-	  to inhibit the quoting of filenames after completion
-
-lib/readline/readline.h
-	- declare rl_filename_completion_desired and
-	  rl_filename_quoting_desired
-
-builtins/bind.def
-	- don't save the old value of rl_outstream before initializing
-	  readline -- it saves garbage values and screws up readline
-
-parse.y
-	- don't have private state telling whether or not readline has
-	  been initialized -- use bash_readline_initialized like other
-	  functions in bashline.c
-
-lib/readline/readline.c
-	- make the default 8-bit behavior be based on whether LC_CTYPE is
-	  defined and its value (accept iso-8859-1 or iso_8859_1)
-
-				    7/6
-				    ---
-variables.c
-	- fix up the declaration of getenv() for convex machines
-
-				    7/7
-				    ---
-lib/readline/readline.c
-	- fixed up typos in the declaration of `savestring'
-
-lib/readline/history.c
-	- fixed an off-by-one error in the ADD_CHAR macro which caused one
-	  extra character to be overwritten, causing the gnu malloc to abort
-	  when that one character was at the end of an allocated block
-	- changed the ADD_STRING macro to avoid some unnecessary xreallocs
-
-lib/readline/display.c
-	- fixed a problem with move_cursor_relative -- function now returns
-	  immediately if it has nothing to do
-	- fixed another problem with displaying prompts with invisible chars
-
-lib/readline/chardefs.h
-	- fixed the CTRL macro to be right (agree with the BSD kernel, for
-	  example)
+	- new function: quote_word_break_characters(), to backslash-quote 
+	  characters in a filename that the readline completion code would
+	  treat as word breaks
+	- change bash_quote_filename to call quote_word_break_characters if
+	  the completion quoting style says to use backslashes
+	- add `:' to rl_filename_quote_characters, since it's part of
+	  filename_word_break_characters
+
+lib/posixheaders/posixjmp.h
+	- new file, with half of bashjmp.h
+	- posixjmp.h and lib/readline/posixjmp.h are symlinks to this file
 
-cpp-Makefile
-	- fixed typo in the `install' recipe
-
-				    7/8
-				    ---
-support/srcdir
-	- fixed to handle srcdir when it begins with ./ or ../ to handle
-	  $(srcdir) being a relative path better
-
-cpp-Makefile
-	- changed some include paths to $(BUILTIN_ABSSRC) when building in
-	  `builtins' to handle $(srcdir) being a relative path
-	- change the `chmod' on bashbug to turn on read and execute for all
-	- added a couple of definitions to make it easier for a later
-	  `configure' program
+bashjmp.h
+	- include "posixjmp.h" for possible redefinitions of setjmp/longjmp
+	  and procenv_t
 
-support/mksysdefs
-	- added a -i option to specify an alternate set of directories to
-	  search for include files
+				   1/17
+				   ----
+shell.c
+	- bad options now cause the standard shell usage message (a subset
+	  of what `bash --help' prints) to be displayed on stderr
 
-lib/readline/bind.c
-	- in rl_read_init_file, when skipping whitespace at the start of
-	  the line, decrement `i' so that we don't jump past the start
-	  of the next line
+trap.c
+	- don't free the trap command for a DEBUG trap in
+	  restore_default_signal if SIG_INPROGRESS is set -- there's already
+	  a pointer saved to the old value in _run_trap_internal.  This
+	  makes bashdb run better, too
 
-machines.h
-	- SCOv4 has a `robust' opendir that checks that you're actually
-	  opening a directory
+examples/bashdb/bashdb.{pre,fns}
+	- a couple of minor fixes; it actually has a chance of working now
 
-				   7/11
+				   1/21
 				   ----
-lib/readline/complete.c
-	- make sure a word break character is unquoted before using it to
-	  separate out the current word for completing
-
-machines.h
-	- new machine description: NetBSD on motorola m68k machines like
-	  the hp300
-	- undef HAVE_GETWD in the generic svr4 machine description, like
-	  other svr4 descriptions
-
-lib/readline/rltty.c
-	- make sure to fflush (rl_outstream) after toggling the setting
-	  of the keypad and meta key
+config.h.in
+	- add a define for <dlfcn.h>, HAVE_DLFCN_H
 
-portbash/libc.sh
-	- add a test for OPENDIR_NOT_ROBUST
+configure.in
+	- look for <dlfcn.h>, define HAVE_DLFCN_H if found
 
-support/getcppsyms.c
-	- output __svr4__ if we find __uxps__ (this makes the Fujitsu port of
-	  SVR4 to the sparc build OK)
+builtins/enable.def
+	- include <dlfcn.h> only if HAVE_DLFCN_H is defined
 
-				   7/12
-				   ----
 lib/readline/display.c
-	- more display-related fixes when the prompt has invisible chars;
-	  this time for screen updates when moving between screen lines
+	- renamed clear_to_eol to _rl_clear_to_eol and made it global, so
+	  other library files (readline.c) can use it
+	- new function _rl_clear_screen, to clear the screen with the right
+	  termcap escape sequence
 
-lib/readline/readline.c, lib/readline/display.c
-	- changes to make readline work with terminals that have auto-wrap
-	  from Per Bothner (new function _rl_update_final, term_xn changes,
-	  some efficiency speedups, new function space_to_eol)
+lib/readline/readline.c
+	- call _rl_clear_to_eol and _rl_clear_screen instead of using tputs
+	- extern declarations for _rl_clear_to_eol and _rl_clear_screen
 
-				   7/13
+				   1/22
 				   ----
-lib/readline/display.c
-	- after moving up screen lines using term_up in _rl_move_vert, if
-	  the new screen line is 0, _rl_last_c_pos needs to be adjusted
-	  to take invisible characters into account.  This was the source
-	  of many bugs
+mailcheck.c
+	- fixed a problem in make_default_mailpath() where a slash was not
+	  added between the default mail directory and the username
 
-				   
-				   7/14
+				   1/23
 				   ----
-documentation/Makefile
-	- change instances of `groff' to `${GROFF}', GROFF is set to
-	  `groff' by default
-
-general.c, variables.c
-	- moved `qsort_string_compare' from variables.c to general.c
+stringlib.c
+	- added a fourth parameter to ansicstr: the length of the returned
+	  string
 
-general.h, variables.h
-	- moved declaration of `qsort_string_compare' from variables.h
-	  to general.h
+externs.h
+	- changed declaration of ansicstr
 
-alias.c, lib/readline/funmap.c
-	- moved qsort auxiliary functions after their use and added
-	  forward declarations to avoid warnings from ANSI C compilers
+parse.y
+	- changed call to ansicstr -- saves a call to strlen
 
-memalloc.h
-	- hpux_9 needs alloca declared as `extern void *' if __STDC__
-	  is defined
+builtins/echo.def
+	- changed call to ansicstr
+	- if do_v9 is non-zero, use putchar to output the string instead
+	  of printf, since there may be embedded NULL characters
 
-support/mksysdefs
-	- removed HAVE_SHARED_LIBS entirely
-	- make a call to /bin/uname -X for SCO machines to avoid running
-	  a different uname from the $PATH
+doc/{bash.1,bashref.texi}, builtins/trap.def
+	- modified the `trap' documentation to make it clearer that trap
+	  takes multiple signal specs as arguments
 
-machines.h
-	- new descriptions: Intel i860 running SVR4, Tahoe running 4.3 BSD
-	- changed descriptions: Mips/RiscOS, DG AViiON, unknown machine
+jobs.c, nojobs.c, jobs.h
+	- renamed initialize_jobs to initialize_job_control, added an
+	  argument (force)
 
 jobs.c
-	- changes to how the shell handles foreground jobs dying of SIGINT:
-		an interactive shell using job control will no longer
-		act as if it received a SIGINT if the foreground job
-		dies from a SIGINT
-
-		a non-interactive shell or shell without job control tries
-		to differentiate between SIGINTs it has seen (in
-		wait_sigint_handler) and a foreground job dying of a SIGINT
-		not sent from the keyboard, and runs the normal SIGINT code
-		only in the former case
-
-				   7/15
-				   ----
-support/mksysdefs
-	- check for ${UNAME}${RELEASE} expanding to `SunOS4*' or `SunOS5*'
-	  to set SYSDEF to SunOS4 or SunOS5, respectively.  Apparently
-	  this does not work for Solbourne
+	- set shell_tty to fileno(stderr) in initialize_job_control if
+	  the shell is not interactive.  This fixes the problem of bad
+	  tty pgrps when monitor mode is turned on in a non-interactive
+	  shell
 
-				   7/18
-				   ----
+sig.c
+	- made initialize_terminating_signals do only that; moved the rest
+	  of the code that used to be there into a new function:
+	  initialize_shell_signals, which calls initialize_terminating_signals
+	  if the shell is interactive
+	- initialize_terminating_signals is now extern
+	- made reset_terminating_signals return immediately if
+	  termsigs_intitialized is zero, meaning that
+	  initialize_terminating_signals has not been called
 
-lib/readline/rltty.c
-	- if output is being flushed on termios systems, loop until the
-	  FLUSHO bit is no longer set in the termios struct
+sig.h
+	- new extern declaration for initialize_terminating_signals
 
-support/mksysdefs
-	- added a -A flag to force creation of ansi-Makefile
+trap.c
+	- call initialize_terminating_signals from set_signal if sig is
+	  EXIT_TRAP and the shell is not interactive.  Since the terminating
+	  signals do not need to be initialized until a trap on exit is
+	  set, not doing that at startup should result in a speed increase
+	  for scripts
 
-machines.h
-	- new entry for Tandem machines running SVR3
+execute_cmd.c
+	- save and restore command->flags in time_command, so you can use
+	  `time command' in a loop
 
-				   7/19
+				   1/24
 				   ----
-lib/readline/rldefs.h
-	- include <termcap.h> if HAVE_TERMCAP_H is defined
-	- use <termio.h> stuff if HAVE_TERMIO_H is defined and _POSIX_VERSION
-	  is not defined
-
-lib/readline/rldefs.h, lib/readline/history.c
-	- include "config.h" if HAVE_CONFIG_H is defined
-
-lib/readline/{rldefs.h,signals.c,readline.c}
-	- WINSIZE_IN_IOCTL_H -> GWINSZ_IN_SYS_IOCTL for compatibility with
-	  other GNU programs
+lib/readline/display.c
+	- fix redisplay code to wrap correctly if the prompt is longer than
+	  the screen width (reported by bos@Eng.Sun.COM)
 
-lib/readline/doc/Makefile
-	- fixed up to create the readline and history manuals in dvi and
-	  ps format
+lib/readline/undo.c
+	- don't include <setjmp.h>, it's not needed
+
+lib/readline/{util,readline}.c
+	- include "posixjmp.h" instead of <setjmp.h>
+	- readline_top_level is now a `procenv_t' instead of a `jmp_buf'
+	  (now readline uses the correct posix semantics for preserving
+	  the signal mask and other things across longjmp)
+
+parse.y
+	- fixes to push_string/pop_string to make them more general -- they
+	  now can be used generally, instead of having to be associated
+	  with an alias being expanded
+	- fixes to the parser so that it parses (( ls abc; ls def); ls ghi)
+	  as a nested subshell command for backwards compatibility.  Broke
+	  the double-paren expression parsing off into a new function:
+	  parse_arith_cmd, called from read_token when a `((' is seen.  If
+	  it looks like an arithmetic command, return `let' and set things
+	  up so that the expression is returned as a double-quoted string
+	  by the next call to read_token.  If it's a nested subshell, push
+	  the text we parsed onto the list of strings for later consumption
+	  and return `('
 
-lib/readline/Makefile
-	- changes inspired by the standalone readline-2.0 distribution
+lib/glob/glob.c
+	- fix glob_vector so it doesn't short-circuit checking a filename if
+	  it starts with a `.' and the pattern starts with `\.' (if
+	  noglob_dot_filenames is set).  This makes `".junk2"*' match
+	  `.junk2.txt' correctly
 
-				   7/20
+				   1/27
 				   ----
-lib/readline/history.c
-	- new function, history_is_stifled (), returns history_stifled
-	- set history_state flags member in the history state functions
-
-lib/readline/history.h
-	- reorganized the function declarations, added missing declarations
-	- history_stifled is no longer exported by the library
-	- added a `flags' member to the HISTORY_STATE structure
-
-bashline.c
-	- use history_is_stifled () instead of history_stifled
-
-lib/readline/readline.c, lib/readline/vi_mode.c
-	- filled in correct argument declarations for functions called via
-	  keymaps (count, key)
+support/bashbug.sh
+	- ask for confirmation before sending the bug report
 
-lib/readline/complete.c
-	- efficiency improvement for compare_strings
+builtins/fc.def
+	- when editing and re-executing a multiline command, make sure
+	  current_command_line_count is initialized and then incremented
+	  for each line read from the file so that the lines added to
+	  the history list by fc_replhist and fc_addhist obey the `lithist'
+	  and `cmdhist' shopt options.  Reported by tibbitts@pb.seflin.org
 
-				   7/21
+				   1/28
 				   ----
-examples/dirfuncs
-	- new directory functions from ksh book, contributed by
-	  Ken Konecki (kenk@wfg.com)
-
-machines.h
-	- hpux_8 and hpux_9 should both #undef HAVE_ALLOCA unless gcc is
-	  being used
+lib/readline/readline.h
+	- added a couple of extern declarations for variables described in
+	  the documentation but heretofore undeclared
 
-				   7/22
-				   ----
-bashline.c
-	- fixed up command_word_completion_function so that filenames with
-	  leading tildes are completed correctly
+builtins/ulimit.def
+	- try to catch some classes of integer overflows before calling
+	  set_limit
 
-				   7/26
+				   1/29
 				   ----
-builtins/read.def
-	- if -r not given, make sure CTLESC is removed from input string
-	  when reading \<newline>
-
-lib/readline/readline.c
-	- new function bind_arrow_keys, which binds vt100/ansi arrow key
-	  escape sequences after reading the termcap definition and the
-	  inputrc file
-	- new function rl_yank_last_arg, which does what insert-last-arg
-	  does in bash
-
-lib/readline/emacs_keymap.c
-	- remove default bindings to rl_arrow_keys for M-[ and M-O
-	- rl_yank_last_arg is now bound to `M-.' and `M-_' in
-	  emacs_meta_keymap
-
-subst.c
-	- when performing process substitution on systems with /dev/fd,
-	  make sure the child clears the slot in dev_fd_list it gets
-	  from its parent so the file descriptor does not get closed
-	  inappropriately if reallocated by, e.g., pipe(2)
-
-bashline.c
-	- removed insert_last_arg and the calls to bind in to `M-.' and `M-_'.
-	  `insert-last-argument' is now bound to rl_yank_last_arg for
-	  backwards compatibility
-
-lib/readline/funmap.c
-	- `yank-last-arg' is now a named command for rl_yank_last_arg
-
-
-documentation/bash.1, documentation/readline.3
-	- add description of yank-last-arg as one of the readline user
-	  commands
+parse.y
+	- push and pop a `(' delimiter while parsing a $(...) construct, so
+	  the history code doesn't try to inappropriately add a `;' when
+	  a newline appears in the `...'
 
-lib/readline/doc/rluser.texinfo
-	- added description of yank-last-arg
+aclocal.m4
+	- new macro, BASH_STRUCT_WINSIZE, which looks for `struct winsize'
+	  in <sys/ioctl.h> (or one of the files it includes)
 
-builtins/getopts.def
-	- fixed a typo in the int-to-string code computing the value to set
-	  OPTIND to: had '\0' instead of '0'
-	- made getopts handle the case where there are more than 9 dollar
-	  variables (where rest_of_args is non-null) correctly
+configure.in
+	- call BASH_STRUCT_WINSIZE
+	- slightly reorganized the calls to bash-specific macros
 
-				   7/28
-				   ----
-lib/readline/display.c
-	- fixes to the display code for single-line-display in the presence
-	  of prompts containing invisible characters
+config.h.in
+	- add an `#undef STRUCT_WINSIZE_IN_SYS_IOCTL'
 
-lib/readline/readline.c
-	- if we are using horizontal scrolling and we have term_xn, decrement
-	  the screenwidth by 1, since we won't be doing any line wrapping
+jobs.c, nojobs.c
+	- only look in sys/ptem.h for struct winsize if
+	  STRUCT_WINSIZE_IN_SYS_IOCTL is not defined to cpp
 
-				   7/31
+				   1/30
 				   ----
-jobs.c
-	- new variable `freeze_jobs_list' to set when changes to the jobs
-	  list or status of jobs in the list (other than calling something
-	  like `jobs -n') are undesirable.  This is set when execuing traps
-	  on SIGCHLD
-
-				    8/1
-				    ---
-subst.c
-	- check that `~' is unquoted before performing tilde expansion in
-	  an assignment statement
+.{distribution,patchlevel}
+	- renamed to _{distribution,patchlevel}
 
-				    8/3
-				    ---
-bracecomp.c
-	- keep brace completion from dumping core if there is only one
-	  match
-
-lib/readline/chardefs.h
-	- add a define for digit_p, which returns the value of isdigit()
-
-lib/readline/readline.c
-	- added function equivalents for uppercase_p, lowercase_p, to_upper,
-	  to_lower, pure_alphabetic, digit_p, and digit_value
-	- replaced calls to numeric () with calls to digit_p, removed
-	  definition of numeric ()
-
-lib/readline/history.c
-	- digit -> digit_p
-
-lib/readline/vi_mode.c
-	- replaced uses of the `isletter' define to use pure_alphabetic
-	  from chartypes.h
-	- replaced uses of `numeric' with calls to digit_p
-	- added do...while(0) to `exchange' define
-
-				    
-				    8/4
-				    ---
-execute_cmd.c
-	- make sure execute_function saves and restores the current loop
-	  count with unwind_protect_int
-
-documentation/features.texi
-	- change the `Shell Command Line Options' section to `Invoking
-	  Bash' to be closer to the GNU coding standards
-
-				    8/5
-				    ---
-builtins/read.def
-	- fixed up a memory leak and made behavior correct when no
-	  variables given and backslash escaped at least one input char
-	- if we added CTLESC anywhere while reading the input string,
-	  make sure we call dequote_string on each word of the input
-	  before calling bind_variable with that string
-
-				    8/6
-				    ---
-lib/readline/readline.h
-	- removed definition of rl_show_star -- the variable no longer
-	  exists
-
-lib/readline/readline.h, lib/readline/history.h
-	- reorganized to follow texinfo manual structure
-
-lib/readline/bind.c
-	- rl_function_dumper is now static
-	- rl_list_funmap names does not take any arguments
-
-builtins/bind.def
-	- rl_list_funmap names does not take any arguments
-
-lib/readline/readline.c, lib/readline/signals.c
-	- rl_init_argument ->_rl_init_argument
+configure.in
+	- create a variable BASHVERS, from the contents of _distribution,
+	  and a variable BASHPATCH, from the contents of _patchlevel,
+	  (using m4 magic so we don't have to distribute .distribution
+	  and .patchlevel) and substitute them into Makefile.in
 
-lib/readline/rltty.c
-	- remove declaration and references to output_was_flushed
+Makefile.in
+	- use Version and PatchLevel variables instead of the contents of
+	  .distribution and .patchlevel, respectively.  These are set by
+	  configure
+	- removed `.machine' from targets and dependencies
 
-support/getcppsyms.c
-	- if __uxps__ is defined, define __uxps__ and __svr4__
+support/mkversion.sh
+	- new shell script to handle updating version.h, replaces mkversion.c
+	  (which is now overkill)
 
-				    8/9
-				    ---
 support/mkdist
-	- added a -r option to set the `root name' of the distribution
-	  (e.g., `bash' or `readline')
-
-				   8/10
-				   ----
-input.c
-	- if check_bash_input sees that default_buffered_input is about
-	  to be the target of a redirection, but the buffered input stream
-	  is not the current input stream (e.g. as the result of an
-	  `eval' while in a script, change default_buffered_input to a
-	  new fd and use a new variable `bash_input_fd_changed' to note
-	  the change
-
-parse.y
-	- if pop_stream sees that default_buffered_input has changed while
-	  a buffered stream was on the save stack (because
-	  bash_input_fd_changed is non-zero), then make the popped stream
-	  and the corresponding buffer use the new fd
+	- don't bother with writing .distribution and .patchlevel files,
+	  since they're no longer distributed
 
-test.c
-	- fixed an off-by-one error in test_stat when testing /dev/fd/n.
+support/mknewvers.sh
+	- simple bash script to make new version files.  It can increment
+	  the major or minor version numbers or patchlevel, or take a
+	  completely new version number (e.g., 2.01) as an argument
 
-cpp-Makefile
-	- add -I$(incdir) to CPPFLAGS
-	- add `incdir = $(prefix)/include' assignment
+doc/Makefile.in
+	- added support for the `install-info' command to update the info
+	  directory file after installing bash.info
 
-				   8/11
+				   1/31
 				   ----
-lib/readline/signals.c
-	- changed some calls to signal () to call rl_set_sighandler()
-	- call _rl_redisplay_after_sigwinch from the sigwinch handler
-	  to do better redisplay of wrapped command lines and multi-line
-	  prompts
-
-builtins/suspend.def
-	- changed some calls to signal () to call set_signal_handler()
-
-lib/readline/*.c
-	- include config.h before anything else if HAVE_CONFIG_H is defined
-
-lib/readline/readline.c
-	- renamed rl_set_mark to _rl_set_mark_at_pos
-	- call _rl_vi_initialize_line from rl_initialize if the current
-	  editing mode is vi mode
-	- only output a term_cr in crlf if term_cr is non-null
-
-lib/readline/readline.h
-	- added declarations for rl_vi_set_mark and rl_vi_goto_mark
-
-lib/readline/readline.c, lib/readline/readline.h
-	- added a new bindable command to set the mark, rl_set_mark
-
-lib/readline/funmap.c
-	- new bindable function with name `set-mark', bound to rl_set_mark
-	- new bindable vi-mode functions with names `vi-set-mark' and
-	  `vi-goto-mark'
-
-documentation/{bash.1,readline.3}, lib/readline/doc/rluser.texinfo
-	- added documentation for `set-mark' (unbound)
-
-lib/readline/vi_mode.c
-	- new vi-command mode commands to save marks (the vi `m' command)
-	  and go to saved marks (the ``' command)
-	- new function _rl_vi_initialize_line called from rl_initialize
-	  to set up the line state in vi mode.  Right now this just resets
-	  all the marks
+builtins/help.def
+	- fix core dump with `help --'
 
-lib/readline/vi_keymap.c
-	- new vi command-mode bindings to set the mark (`m') and to go to
-	  a saved mark (``')
+susbt.c
+	- make call_expand_word_internal obey the convention that if
+	  expand_word_internal returns &expand_word_{fatal,error}, then
+	  w->word has already been freed
+	- return &expand_word_fatal from expand_word_internal if the shell
+	  is not interactive and `set -u' has been executed
 
-documentation/readline.3
-	- added vi-set-mark and vi-goto-mark to list of key bindings for
-	  vi command mode
+test.c, general.c
+	- moved group_member from test.c to general.c
 
-lib/readline/{readline.c,display.c}
-	- renamed term_xn to _rl_term_autowrap
+externs.h, general.h
+	- moved extern declaration of group_member from externs.h to general.h
 
-lib/readline/display.c
-	- new function _rl_redisplay_after_sigwinch to encapsulate the
-	  necessary redisplay code after a SIGWINCH is received
+general.c
+	- broke the code that initializes the group array out into a new
+	  function, initialize_group_array()
+	- initialize_group_array() now initializes an array even if the OS
+	  does not have getgroups().  If it does not, an array with one
+	  element (the real gid) is created
+	- call sysconf(_SC_NGROUPS_MAX) if sysconf is available and
+	  _SC_NGROUPS_MAX is defined
+	- new function, char **get_group_list(int *), to return an array
+	  of strings made from the groups list
 
 variables.c
-	- only call find_user_command on the shell name in shell_initialize
-	  if shell_name is not an absolute program name
-	- call canonicalize_pathname on the full shell pathname after
-	  calling make_absolute on it if the shell name starts with a `.'
-
-siglist.h
-	- FreeBSD does not need a definition of sys_siglist[]
+	- new dynamic array variable: GROUPS, expands to the group set as
+	  obtained with getgroups() (or whatever initialize_group_array()
+	  makes)
 
-				   8/12
-				   ----
-lib/readline/rltty.c
-	- don't try to use FLUSHO unless FLUSHO is defined
+doc/{bash.{1,html},bashref.texi}
+	- added description of GROUPS variable
 
-cpp-Makefile
-	- make sure to pass $(CC) to the make in the lib/malloc subdirectory
+test/test-tests
+	- before modifying the setgid bit on /tmp/setgid, try to change its
+	  group to ${GROUPS[0]} 
 
-lib/readline/readline.c
-	- don't declare PC, UP, and BC as extern; it causes too many problems
+				    2/3
+				    ---
+aclocal.m4
+	- new autconf macro, BASH_MISC_SPEED_T, to see if speed_t is defined
+	  in <sys/types.h>
 
-				   8/13
-				   ----
-machines.h
-	- add -DNO_SBRK_DECL to SYSDEP_CFLAGS for SVR4.2 on i386
+configure.in
+	- call BASH_MISC_SPEED_T
 
-support/mksysdefs
-	- new variable UNAME_S for the output of `uname -s'
-	- if uname -s does not output the same thing as uname, and uname -s
-	  outputs UNIX_SV, then set UNAME to UNIX_SV.  Some versions of i386
-	  SVR4.2 make `uname' equivalent to `uname -n'
+config.h.in
+	- add `#undef SPEED_T_IN_SYS_TYPES'
 
-lib/readline/memalloc.h
-	- removed from the library; not needed since readline does not use
-	  alloca
+lib/readline/tcap.h
+	- include `rltty.h' if HAVE_TERMCAP_H and __linux__ are defined, but
+	  SPEED_T_IN_SYS_STYPES is not, before including <termcap.h>
 
-lib/readline/{history.c,isearch.c,search.c,rldefs.h}, lib/tilde/tilde.c
-	- no longer include memalloc.h
+support/mksignames.c, siglist.c
+	- add support for 4.4 BSD SIGLOST
 
-lib/readline/Makefile
-	- remove dependencies on memalloc.h
+support/config.guess
+	- add support for recognizing QNX based on `uname' output
 
-lib/readline/signals.c
-	- overhauled signal handling based on ideas from Bruno Haible
-	  (haible@ma2s2.mathematik.uni-karlsruhe.de) so that readline
-	  saves and restores the signal mask and flags on Posix systems
-	- don't include fcntl.h, sys/file.h, errno.h, or stdlib.h
-	- ignore SIGALRM during rl_signal_handler like SIGINT until the
-	  old handler is installed, on non-Posix or non-BSD systems
-	- new define SIGHANDLER_RETURN, dependent on the value of
-	  VOID_SIGHANDLER
-	- use RETSIGTYPE instead of `sighandler' as the signal handler
-	  return type; it's the name autoconf uses.  If not defined,
-	  set up from VOID_SIGHANDLER
-
-lib/readline/rltty.c
-	- call control_meta_key and control_keypad from rl_deprep_terminal
-	  before putting the terminal back into ICANON mode to avoid
-	  messing up the tty driver
-
-				   8/15
-				   ----
-cpp-Makefile
-	- changes so that fewer -Idir arguments are passed to the make
-	  in ./builtins
+Makefile.in
+	- make sure recho and zecho are compiled with the same set of CC
+	  options as the rest of the sources
 
-				   8/16
-				   ----
 lib/readline/bind.c
-	- use KEYMAP_SIZE instead of 128 in rl_invoking_keyseqs_in_map
-
-lib/readline/complete.c
-	- qsort the array of matches omitting matches[0], which must stay
-	  in place no matter what, even if strcmp doesn't compare unsigned
-	  chars correctly
-	- some fixes to the loop that prints the list of completions
-
-lib/readline/keymaps.c
-	- run the loop up to 127 when creating a new keymap in rl_make_keymap
-
-input.h, general.h, lib/readline/keymaps.h, lib/readline/tilde.h
-	- use _FUNCTION_DEF instead of __FUNCTION_DEF
-
-lib/readline/parens.c
-	- include "config.h" if HAVE_CONFIG_H is defined
-	- include <sys/select.h> if HAVE_SYS_SELECT_H is defined
-	- use HAVE_SELECT instead of FD_SET as the `configuring define';
-	  define HAVE_SELECT if it is not already defined but FD_SET is
+	- change calls to rl_generic_bind to cast the third argument to
+	  (char *) where necessary
 
-lib/readline/readline.c, lib/readline/rltty.c
-	- the __GO32__ include file is <pc.h>
-
-lib/tilde/tilde.c
-	- rewrote tilde_expand_word, eliminated static u_name array in
-	  favor of using xmalloc
-
-lib/tilde/tilde.h
-	- use single leading underscore for file inclusion guard
-
-lib/readline/vi_mode.c
-	- use KEYMAP_SIZE rather than a literal 127 as the loop limit when
-	  creating the overstrike keymap
-
-support/install.sh
-	- new file, from the make-3.71 distribution
-
-				   8/17
-				   ----
-lib/readline/complete.c, lib/readline/readline.h
-	- new variable, rl_basic_quote_characters, used to see
-	  if a word break character was a quoting character, so we can
-	  do appropriate quoting after the completion
-
-lib/readline/parens.c
-	- use rl_basic_quote_characters when trying to find a matching
-	  open paren
-
-lib/readline/doc/rltech.texinfo
-	- description of rl_basic_quote_characters
-
-				   8/18
-				   ----
-newversion.c
-	- renamed to support/mkversion.c, which builds to mkversion
-
-cpp-Makefile
-	- changed to use support/mkversion
-	- removed instances of $< except in suffix rules -- some makes don't
-	  handle those right
-
-lib/readline/rldefs.h
-	- make sure to #undef HAVE_DIRENT_H on NeXT machines
-
-				   
-				   8/19
-				   ----
-lib/readline/readline.h
-	- removed some duplicate function declarations
-
-trap.h
-	- define DEBUG_TRAP as NSIG for the last slot in trap_list
-	- define EXIT_TRAP as 0
-	- declare trap_list as array of unknown size, since this is just a
-	  header file
-
-trap.c
-	- extend trap_list and sigmodes to NSIG+1
-	- initialize DEBUG_TRAP to do nothing; changes to functions to handle
-	  DEBUG_TRAP
-	- let decode_signal return OK if NSIG is given, even though this will
-	  vary between systems, for DEBUG_TRAP
-	- make reset_signal_handlers and restore_original_signals loop from
-	  1 to NSIG, not 0 to NSIG
-	- change GET_ORIGINAL_SIGNAL to try to get the original signal handler
-	  only if `sig' is less than NSIG
-	- restore_default_signal just frees the trap string for both EXIT_TRAP
-	  and DEBUG_TRAP
-	- run_debug_trap: new function to run traps on DEBUG
-	- free_trap_command: new function to free a trap string if it is
-	  really a command and not one of the special trap values
-	- if a SIGINT trap resets the SIGINT trap string, free the old string
-	- new function: set_debug_trap(char *), which will reset the DEBUG
-	  trap string, suitable for calling from an unwind-protect handler
-
-signames.c
-	- arrange things so that signal_names[NSIG] == "DEBUG"
-
-builtins/trap.def
-	- make sure traps on DEBUG are printed
+command.h
+	- two new flags: CMD_AMPERSAND (currently unused), and CMD_STDIN_REDIR,
+	  which means that this command should have its standard input
+	  redirected from /dev/null if there are not any explicit redirections
+	  to stdin
 
 execute_cmd.c
-	- call run_debug_trap at the end of the `cm_simple' case in
-	  execute_command_internal
-	- change execute_function so that traps on DEBUG do not propagate
-	  into the function
-
-documentation/bash.1, documentation/features.texi
-	- documented the new DEBUG trap
-
-				   8/20
-				   ----
-shell.c, config.h
-	- made the name of the restricted shell configurable in config.h;
-	  the option is RESTRICTED_SHELL_NAME
-
-shell.c
-	- if the shell is invoked as `sh', enter Posix.2 mode after running
-	  /etc/profile and ~/.profile
-
-builtins/read.def
-	- added a -p option for a prompt string, which is displayed before
-	  trying to read anything
-
-documentation/bash.1
-	- description of `read -p'
-	- description of new -r and -s options to `jobs' 
-
-documentation/features.texi
-	- description of new -r and -s options to `jobs' 
-
-jobs.c
-	- new functions: list_running_jobs, list_stopped_jobs
-	- list_jobs -> list_all_jobs
-	- new function `print_job' to do the work for all of the list*jobs
-	  functions
-	- list_one_job type now void, calls print_job to do the work
-
-jobs.h
-	- new declarations for list_running_jobs, list_stopped_jobs
-	- list_one_job now void
-	- list_jobs -> list_all_jobs
-
-builtins/jobs.def
-	- new options: -r to print only running jobs, -s to print only
-	  stopped jobs
+	- a redirection of type r_inputa_direction is no longer added at the
+	  beginning of the redirection chain for an async command;
+	  CMD_STDIN_REDIR is set in the flags instead
+	- new function: stdin_redirects: returns the number of redirections to
+	  stdin in a chain of redirections
+	- new functions: async_redirect_stdin() to open /dev/null and make it
+	  file descriptor 0
+	- changes to make CMD_STDIN_REDIR propagate to all of the necessary
+	  functions (like execute_simple_command)
+	- execute_disk command now takes the flags from simple_command rather
+	  than just the CMD_NO_FORK flag as its last argument
+	- various places after make_child is executed (in the child) check for
+	  CMD_STDIN_REDIRECT (and no stdin redirections or piping) and call
+	  async_redirect_stdin
+	- stdin_redir is now global
 
-				   8/22
-				   ----
-lib/readline/readline.c
-	- don't try to expand a null prompt string
+eval.c
+	- set stdin_redir to 0 just before calling execute_command so it
+	  gets reset to a known value and doesn't persist across commands
 
-subst.c
-	- remove special handling of hostname_completion_file and
-	  history_control; HOSTFILE and HISTCONTROL are the acceptable
-	  names
-	- new function sv_histignore; call if HISTIGNORE variable is
-	  assigned to
+				    2/4
+				    ---
+builtins/ulimit.def
+	- add a stub function for ulimit(2) on systems without HAVE_RESOURCE
+	  or HAVE_ULIMIT that just sets errno to EINVAL and returns -1 --
+	  QNX is one such system
 
 bashhist.c
-	- new functions and declarations for handling HISTIGNORE variable
-	  and its specifications for command lines to ignore
-	- change maybe_add_history to call history_should_ignore if the
-	  HISTCONTROL tests are passed
-
-documentation/bash.1, documentation/features.texi
-	- removed mention of history_control and hostname_completion_file
-	- added description of HISTIGNORE
-
-				   8/23
-				   ----
-array.c, array.h
-	- promoted to the shell mainline code; the `array' subdir is now
-	  gone
-
-variables.h
-	- include `array.h' for the ARRAY typedef
+	- pre_process_line needs to protect all occurrences of hist_verify
+	  with #ifdef READLINE
 
-variables.c
-	- changes to assignment() to handle array assignments a[x]=b
-	- split off the creation of new variables into a new function:
-	  make_new_variable
-	- split off the construction of variable values for assignments and
-	  bindings into a new function: make_variable_value
-	- new function: bind_array_variable, to handle the addition of
-	  new array variables and indices
-
-subst.c
-	- changes to do_assignment_internal to handle a[x]=b:
-		parse the subscript out of the name and call evalexp() on it
-		call bind_array_variable to do the value assignment
-	- change the printing of assignment statements when -x is in effect
-	  to before the variable binding is attempted, so any error messages
-	  look better
-	- new functions for referencing indexed arrays and to find the length
-	  of arrays or array indices
-
-				   8/24
-				   ----
-expr.c
-	- fixed up error reporting to use get_name_for_error if not executing
-	  a builtin
-
-subst.c
-	- everywhere evalexp() is called, use maybe_expand_string to expand
-	  the expression (calls expand_string if necessary)
-
-subst.h
-	- extern declarations for array convenience functions exported to the
-	  rest of the shell
+builtins/shopt.def
+	- hist_verify needs to be protected with #ifdef READLINE, not
+	  #ifdef HISTORY
 
-variables.c
-	- more new convenience array functions: make_new_array_variable and
-	  convert_var_to_array
+				    2/5
+				    ---
+support/config.{guess,sub}, configure.in
+	- chages to better support the Harris Night Hawk
 
-builtins/declare.def
-	- additions to declare_internal to add a -a flag which makes array
-	  variables, and the requisite semantics (declare +a does not work,
-	  declare -a name=value does not work, etc.)
-	- code to support the ksh-like declare a[] syntax to make array
-	  variables
+[many files]
+	- changes for things pointed out by gcc -Wall
 
-builtins/setattr.def
-	- made set_or_show_attributes understand array variables and `-a'
+lib/glob/Makefile.in
+	- make sure -DSHELL is included in CCFLAGS so that globbing is
+	  interruptible
 
-builtins/set.def
-	- changes to make `unset name' work, where name is an array variable
-	- changes to make `unset name[xxx]' work
+lib/malloc/malloc.c
+	- extern declaration for botch: if botch is #defined, it should be
+	  the name of a void function
 
-				   8/25
-				   ----
+configure.in,Makefile.in,builtins/Makefile.in,
+lib/{glob,malloc,readline,tilde}/Makefile.in
+	- add a LOCAL_DEFS variable, substituted from configure into the
+	  various Makefiles.  It's set to -DSHELL, so that define gets
+	  passed to all sub-makes without doing anything special
 
 lib/readline/readline.c
-	- make sure meta characters are added to a keyboard macro as
-	  ESC-char if we are converting meta chars to ascii
-	- change rl_unix_word_rubout to handle repeat counts
-	- move the check of LC_CTYPE from rl_initialize to
-	  readline_initialize_everything, where it should have been
-	  all along
+	- change to rl_digit_loop to make it compatible with GNU emacs:
+	  if a key bound to `universal-argument' is read after reading
+	  some digits, it terminates the argument but is otherwise
+	  ignored.  This is how people can insert repeated digits
 
-error.c
-	- make sure <sys/types.h> is included before <fcntl.h>
+doc/{bash.{1,html},readline.3}, lib/readline/doc/rluser.texinfo
+	- changed description of `universal-argument' to describe how
+	  to terminate a numeric argument
 
-machines.h
-	- new entry for Bull DPX2
+				    2/6
+				    ---
+jobs.c
+	- changed kill_pid to diddle the job and process data structures
+	  if we're sending SIGCONT to a stopped job with `kill' so that
+	  the shell knows the job is running again.  ksh93 does this
 
-lib/tilde/tilde.c
-	- include <sys/types.h> before <pwd.h>
+				    2/7
+				    ---
+unwind_prot.c
+	- changed bcopy to FASTCOPY
 
-variables.c
-	- PS1 and PS2 are no longer non-unsettable
+builtins/enable.def
+	- changed pointer arithmetic in delete_builtin to just subtract
+	  the head of the builtins list (shell_builtins) from a pointer
+	  to the builtin to be deleted (b) to find its index in the list
+	  on ANSI C systems
 
-				   8/26
+				   2/10
 				   ----
-variables.c
-	- PPID, UID, and EUID, since they are read-only, need not appear on
-	  the `non-unsettable' list
-	- PATH and IFS now have the `nounset' attribute set rather than use
-	  a separate `non-unsettable' list
-	- the non-unsettable list and non_unsettable() are gone
-
-variables.h
-	- new define `non_unsettable_p' to test nounset attribute
-
-builtins/set.def
-	- `unset' now uses non_unsettable_p to tell whether a variable can
-	  be unset even if it's not readonly
-
-builtins/read.def
-	- read -a arrayname will read a list of values from stdin and assign
-	  them to the array `arrayname', splitting on " \t\n" (honoring
-	  backslash quoting if -r is not supplied)
-
-builtins/declare.def
-	- print values of arrays when `display -a' is invoked
+lib/readline/bind.c
+	- when using old-style keynames to bind to a new-style macro,
+	  pass an array of type (unsigned char *) to rl_macro_bind
 
-builtins/exit.def
-	- remove references to `bye'
+builtins/getopt.c
+	- change sh_getopt to return EOF if nextchar is empty or NULL
+	  and sh_optind is greater than argc.  This can happen if a
+	  script mixes calls to `getopts' with calls to `shift'
 
-				   8/28
+				   2/11
 				   ----
-lib/readline/rltty.c
-	- removed calls to control_keypad; they cause problems on some
-	  machines and should not really be performed by an application
-
-cpp-Makefile
-	- support for optionally linking array.o into the shell
-
-lib/glob/glob.c
-	- when returning an error after glob_filename on the directory name
-	  returns an error result, make sure to free `result'.  This fixes
-	  a memory leak for names passed to glob_filename which contain
-	  globbing characters before the first `/' but do not match any
-	  existing files
+print_cmd.c
+	- fixed make_command_string_internal so that commands with the
+	  CMD_TIME_POSIX flag set print `time -p' instead of `time'
+	- changed print_redirection_list so the here documents are
+	  printed after all the other redirections, and followed by a
+	  newline (rather than a possible semicolon)
+	- added a new variable, was_heredoc, to avoid adding a semicolon
+	  at the beginning of an empty line after printing the heredoc
+	  ending delimiter
 
-				   8/29
-				   ----
 execute_cmd.c
-	- the `for' and `select' commands should expand their word lists
-	  using expand_words_no_vars
-
-command.h, execute_cmd.c, general.c, make_cmd.c, print_cmd.c, subst.c,parse.y
-	- changed the `dollar_present', `quoted', and `assignment' members
-	  of WORD_DESC to a single flags word
-
-				   8/30
-				   ----
-variables.c
-	- make get_string_value return array[0] if called with a variable
-	  that is an array
+	- don't put the gnu_argv_flags into the environment any more; it's
+	  proven to be a bad idea
 
-				   8/31
-				   ----
-lib/readline/display.c
-	- don't take the value of _rl_term_autowrap into account when
-	  computing which screen line the cursor should be on; it screws
-	  up wrapping on terminals without the `xn' capability
-
-lib/readline/examples/rltest.c
-	- new file, test code moved here from readline.c
-
-documentation/features.texi, lib/readline/doc/hsuser.texinfo
-	- fixed up printing of some of the shell options and the history
-	  commands and modifiers
-
-subst.c
-	- make sure $name, where `name' is an array variable, returns
-	  ${name[0]}
-
-variables.c
-	- convert a variable to an array even if index 0 is being assigned to,
-	  instead of leaving it a `normal' variable
-
-				    9/1
-				    ---
-
-builtins/setattr.def
-	- make sure that array variables printed by `declare' single-quote
-	  the value after the `=' so the statement can be reused as input
-
-variables.c
-	- make sure array variables can be exported (name=(assignments)), and
-	  that the export code works.  The code is commented out because of
-	  the possible confusion between an array and a regular string that
-	  looks like an array assignment string.
-
-				    9/2
-				    ---
-variables.c
-	- make sure there is an executable file with the same name as
-	  `shell' name in the current directory before assigning it to
-	  $BASH at startup.  If there is not, just make $BASH the login
-	  shell name
-
-				    9/3
-				    ---
-parse.y
-	- removed a reduction from the `list0' production that could cause
-	  statements which require semicolons to be allowed without error
-	  (like before a `}' in a group command)
-
-lib/readline/display.c
-	- new functions: _rl_save_prompt and _rl_restore_prompt to save and
-	  restore prompt invisible character info
-	- if the redisplay code encounters a line shorter than the old one
-          and containing invisible characters, make sure that the cursor is
-	  at the end of the new text before calling clear_to_eol
-	- new variable last_invisible containing the index in the prompt
-	  string of the last invisible character.  We only have to redraw
-	  the prompt string of _rl_last_c_pos is < last_invisible
-	- new function _rl_make_prompt_for_search to take care of setting
-	  up the prompt string for a non-incremental search when the prompt
-	  contains invisible characters
-	- fix so that a prompt string with invisible characters is not
-	  redrawn each time through update_line (don't tputs term_cr unless
-	  the cursor is before the last invisible character in the prompt
-	  and will be moving past the last invisible char of the prompt
-
-lib/readline/isearch.c
-	- call _rl_save_prompt and _rl_restore_prompt when changing the prompt
-	  to do i-search
-
-				    9/4
-				    ---
-lib/readline/readline.c
-	- doing_an_undo -> _rl_doing_an_undo for use by the vi-mode code
-	- rewrote rl_do_undo to reformat and eliminate a clumsy goto
-	- new split the add-to-kill-ring code off into a separate function,
-	  _rl_copy_to_kill_ring
-	- added a new bindable function rl_copy_region_as_kill
-	- added a new bindable function rl_kill_region
-	- moved _rl_char_search_internal to here from vi_mode.c
-	- new bindable function rl_char_search
-
-lib/readline/vi_mode.c
-	- made `.' work for the `cw' and `[Ss]' commands, which implicitly
-	  put the editor into insertion mode after they run
-	- split rl_vi_char_search code that actually searches for a character
-	  into a new function, _rl_char_search_internal
-
-lib/readline/emacs_keymap.c
-	- bind M-= to possible-completions for ksh compatibility
-	- bind M-* to insert-completions
-	- bind C-x C-x to exchange-point-and-mark
-	- bind C-] to character-search
-	- bind C-@ to set-mark
-
-lib/readline/funmap.c
-	- new bindable function names: exchange-point-and-mark, kill-region,
-	  copy-region-as-kill, character-search
-
-lib/readline/readline.h
-	- declare rl_exchange_point_and_mark, rl_copy_region_to_kill,
-	  rl_kill_region, and rl_char_search externally
-
-documentation/{bash.1,readline.3}, lib/readline/doc/rluser.texinfo
-	- document new exchange-point-and-mark bindable command
-	- documented M-* default emacs-mode binding to insert-completions
-	- documented new copy-region-as-kill and kill-region commands
-	- documented new character search emacs-mode command
-	- documented new binding for set-mark
-
-lib/readline/rldefs.h
-	- moved values for `dir' when searching for characters in the line
-	  from vi_mode.c to here
-
-				    9/6
-				    ---
-general.c
-	- new function ansicstr, which decodes ANSI-C backslash-escaped
-	  characters (with the addition of \e and \E to mean escape) and
-	  returns a new string
-
-subst.c
-	- new expansion $'...' which translates ANSI-C backslash escapes
-	  in `...' and expands to the result
+configure.in
+	- set LOCAL_CFLAGS to `-DSunOS5' on Solaris 5.5[.x]
 
 builtins/echo.def
-	- ifdefs for ANSI-C for \a and \v rather than just expanding to
-	  literal ASCII values
+	- only call fflush() after printing each word on SunOS 5.5, since
+	  that's the system with the bug that prompted its inclusion in
+	  the first place
 
-documentation/bash.1
-	- documented new $'...' expansion
-
-				    9/7
-				    ---
-builtins/enable.def
-	- fixed up the help text
-	- added a -s flag to restrict operation to Posix.2 `special' builtins
-	- removed -all option; -a is the way to do it now
-	- converted to use the builtin getopt
-
-builtins/common.c
-	- changed builtin_address_internal to return a pointer to a
-	  struct builtin, which makes it much more useful.  Changed
-	  find_shell_builtin and builtin_address accordingly
-	- new function find_special_builtin, which returns special builtins
-	  ((flags & SPECIAL_BUILTIN) != 0)
+support/mksignames.c
+	- added support for more system-specific signals from AIX 4.2,
+	  changed initialization order so that system-specific signals
+	  are done first, before the common signals
 
+				   2/12
+				   ----
 execute_cmd.c
-	- in Posix.2 mode, find special builtins before shell functions
-	  when performing command lookup
-	- in find_user_command_internal, return NULL if there is no $PATH
-	- in Posix.2 mode, failure of a special builtin causes a non-
-	  interactive shell to exit
-
-variables.c
-	- allow $PATH to be unset
+	- broke the code that creates a file containing the text of a
+	  here document out into a separate function: here_doc_to_fd
+	- create the temp file used for a here document with mode 600
+	- open the temp file used for a here document with O_EXCL
 
-				    9/8
-				    ---
-input.h
-	- added a new `input type': st_stdin, for use when using readline,
-	  since readline is not properly a string
+shell.h
+	- changed the uid members of struct user_info to be of type uid_t
+	  and the gid members to be of type gid_t
 
 parse.y
-	- surgery on the grammar:
-		o added new `compound_list' production: a list that can end
-		  without a newline, `;' or `&' (used in subshell commands
-		  and case clause commands)
-		o removed shell_command_1, folded rules into command
-		o eliminated pattern_list_1; changed case_clause_sequence to
-		  directly incorporate the SEMI_SEMI token (a pattern_list_1
-		  was just a pattern_list with a trailing SEMI_SEMI)
-		o new `for_command' and `case_command' productions
-		o redirections -> redirection_list
-		o newlines -> newline_list
-		o redid the code that attaches redirections to the function
-		  command rather than the function definition and eliminated
-		  all of the shift/reduce conflicts
-	- changed with_input_from_stdin in the readline case to set
-	  bash_input.type to st_stdin
-
-print_cmd.c
-	- a couple of changes to the way functions are printed
-
-shell.c
-	- only execute the PROMPT_COMMAND if input is not coming from a
-	  string (bash_input.type != st_string).  This fixes the problem
-	  of PROMPT_COMMAND being executed by `eval' commands in an
-	  interactive shell
-
-				   9/12
-				   ----
-shell.c
-	- all interactive shells SIGHUP running jobs when exiting due to
-	  a SIGHUP, not just login shells
-
-				   9/13
-				   ----
-config.h, config.h.mini
-	- changed the default primary prompt (PPROMPT) to "\s\$ "
-	- removed the INTERACTIVE_COMMENTS define
-
-flags.c
-	- interactive comments are now enabled by default, regardless of the
-	  INTERACTIVE_COMMENTS define
+	- changed the type of the `type' argument to init_yy_io() to be
+	  `enum stream_type', since that's what's always passed
 
 input.h
-	- a new enum: stream_type, used in the BASH_INPUT struct
-
-parse.y
-	- bash_input.type is now initialized to st_none
-
-bashhist.c
-	- command_oriented_history is now the default
-
-documentation/bash.1
-	- removed description of `nolinks' variable, updated set -P
-	  description
-	- removed description of `notify' variable, updated set -b
-	  description
-	- removed description of `noclobber' variable, updated set -C
-	  description
-
-variables.c
-	- IFS may now be unset
-
-builtins/read.def
-	- now does the right thing if IFS is unset (acts as if it is
-	  set to " \t\n")
+	- changed the function prototype for init_yy_io so the third arg is
+	  type `enum stream_type'
 
-builtins/cd.def
-	- added a new parameter to change_to_directory () which says
-	  whether or not to follow symlinks, instead of using the
-	  global no_symbolic_links
-	- changed cd_builtin to accept a -P option to disable symlink
-	  following temporarily
-
-builtins/ulimit.def
-	- instead of using a hardcoded `long' for the return type of the
-	  rlimit functions, use RLIMTYPE, which defaults to long
-	- new defines string_to_rlimtype -> string_to_long and
-	  print_rlimtype for systems which do not need `RLIMTYPE' defined
-	  to something other than long
-
-general.c
-	- function replacements for string_to_rlimtype and print_rlimtype
-	  for machines which have RLIMTYPE defined in the machine description
-
-general.h
-	- extern declarations for string_to_rlimtype and print_rlimtype
-
-shell.c
-	- don't source file named by $ENV at script startup if act_like_sh
-	  is turned on
-
-machines.h
-	- new entry for amiga/netbsd
-	- change all the netbsd entries to add -DRLIMTYPE=quad_t to
-	  SYSDEP_CFLAGS
-
-siglist.h
-	- define sys_siglist as _sys_siglist on the amiga only if USGr4 is
-	  defined
-
-print_cmd.c
-	- change so that printf is not prototyped on LynxOS with gcc
-
-lib/readline/Makefile
-	- changed the `installdirs' target to not fail if the directories
-	  do not need creating
-
-lib/readline/history.c
-	- fixed a bug in history_arg_extract which manifested itself when
-	  !* was used after a command without arguments (e.g., pwd ; echo !*)
-
-				   9/15
-				   ----
-subst.c
-	- expand_word_internal should preserve the flags (other than
-	  W_QUOTED) from the word passed as a parameter on the word it returns
-	- expand_words_internal renamed to expand_word_list_internal
-	- expand_word_list_internal does not perform word splitting or
-	  globbing on words with the W_ASSIGNMENT bit set
-	- removed special treatment of `notify', `command_oriented_history',
-	  `history_control', and `nolinks'
-	- rewrote some of the sv_* functions to remove or avoid calls to
-	  sscanf, one of the most expensive C library functions
-
-variables.c
-	- don't look for $command_oriented_history at startup, since it now
-	  defaults to `on'
-
-general.h, general.c
-	- new function posix_initialize to do whatever is necessary to enable
-	  `Posix mode'
-
-shell.c, subst.c
-	- call posix_initialize
-
-general.c
-	- rewrote replacements for strchr, strrchr to make them faster
-
-builtins/enable.def
-	- new -f option and necessary support to load builtins from a shared
-	  object file on systems supporting both dlopen() and dlsym().  New
-	  builtins loaded this way can replace existing shell builtins or
-	  add completely new functionality
-
-builtins.h
-	- mkbuiltins.c creates `static_shell_builtins', which is a fixed
-	  array, `shell_builtins' points to this array initially
-	- extern declaration for current_builtin
-
-builtins/mkbuiltins.c
-	- change to creat static_shell_builtins[] and declare shell_builtins
-	  as a pointer to it
-	- change to declare `current_builtin' in created builtins.c
-
-builtins/source.def
-	- print an error message and return failure if no filename argument
-	  is supplied
-
-builtins/common.c
-	- current_builtin is set by find_shell_builtin, find_special_builtin,
-	  and builtin_address_internal
-	- new function builtin_usage, which prints the contents of
-	  current_builtin->short_doc
-
-builtins/{bind,declare,enable,fc,getopts,hash,history,jobs,kill,read,set,
-	  setattr,trap,ulimit,umask}.def
-	- changed the builtins in these files to use builtin_usage() to
-	  print usage messages
-
-cpp-Makefile
-	- support for the HAVE_DLOPEN and HAVE_DLSYM defines
-
-machines.h
-	- added -ldl and -Bdynamic to the SunOS4 entry
+externs.h
+	- added a prototype for getcwd, if HAVE_GETCWD is not defined
 
-documentation/{bash.1,features.texi}
-	- documented new enable -f option to dynamically load builtins
+builtins/umask.def
+	- changed all variables that are used as arguments to or save the
+	  return value from umask(2) to be of type mode_t
+	- changed print_symbolic_umask to take an argument of type mode_t
 
-				   9/16
+				   2/13
 				   ----
-test.c
-	- added `==' as a synonym for `='
+jobs.c
+	- if old_sigint_handler is set to SIG_DFL, as it will be in a script
+	  without a trap on SIGINT installed, call termination_unwind_protect()
+	  directly from waitchld()
 
-				   9/18
+				   2/14
 				   ----
-bashline.c
-	- include readline/rlconf.h so that VI_MODE is defined if
-	  appropriate
-
-shell.c
-	- removed `-nobraceexpansion' option
-
-flags.c, flags.h
-	- added new -B flag; enabled if brace expansion is turned on
-
-subst.c
-	- use brace_expand instead of !no_brace_expand
-
-builtins/set.def
-	- change set -o braceexpand to be equivalent to set -B
+configure.in
+	- added a section before the call to BASH_CHECK_LIB_TERMCAP to
+	  set a variable prefer_curses on some systems (AIX, for one)
 
-parse.y
-	- in with_input_from_stdin, do nothing if bash_input.type already
-	  is st_stdin or if there is already a stream with type st_stdin
-	  on the saved stream stack
-	- new function stream_on_stack to find out if there is a saved
-	  stream of a specified type
+aclocal.m4
+	- in BASH_CHECK_LIB_TERMCAP, don't return -ltermcap if $prefer_curses
+	  is non-empty
 
-documentation/bash.1, documentation/features.texi
-	- doeumented new set -B option
-	- removed -nobraceexpansion shell startup option
+lib/readline/Makefile.in
+	- redid the dependencies
 
-				   9/19
+				   2/17
 				   ----
-builtins/reserved.def
-	- made `help select' work
+hashlib.c, getcwd.c
+	- include `bashansi.h' instead of stdlib.h and string.h
 
-cpp-Makefile
-	- removed references to the nonexistant `load.def'
-	- removed conditionals based on GETOPTS_BUILTIN
+error.c, siglist.c, xmalloc.c, builtins/{common,evalfile,mkbuiltins,psize}.c,
+builtins/{exec,exit,fg_bg,hash,history}.def
+	- include `bashtypes.h' instead of directly including <sys/types.h> 
 
-builtins/getopts.def
-	- no longer $DEPENDS_ON GETOPTS_BUILTIN
-
-config.h, config.h.mini, builtins/help.def
-	- made the `help' builtin dependent on the HELP_BUILTIN define
-	- removed GETOPTS_BUILTIN
+builtins/fc.def
+	- include ../bashtypes.h and ../posixstat.h instead of bashtypes.h
+	  and posixstat.h
 
 builtins/mkbuiltins.c
-	- made the `$DEPENDS_ON' clause work for reserved words and
-	  shell control structures in reserved.def by writing out
-	  dependencies to builtext.h even if there is no `function'
-	- made the long_doc for each builtin and reserved word get
-	  written out to builtins.c with #ifdef HELP_BUILTIN surrounding
-	  the doc strings -- this makes the minimal shell much smaller
-
-shell.c, documentation/bash.1
-	- removed `-quiet' long option
-
-parse.y
-	- make sure that word splitting is not performed on the prompt
-	  string after expanding in prompt_string_decode by calling
-	  expand_string_unsplit instead of expand_string
-
-lib/readline/bind.c
-	- new readline variable `mark-directories': if set, completed
-	  directory names have a slash appended
-
-lib/readline/complete.c
-	- support for `mark-directories'
-
-documentation/{bash.1,readline.3}, lib/readline/doc/rluser.texinfo
-	- documented the new `mark-directories' variable
-
-builtins/bind.def
-	- new option `-r' to remove a binding for a specified key sequence
-	- make sure that a failure return from bind resets the keymap
-	  correctly if a -m option was supplied
-
-documentation/{bash.1,features.texi}
-	- documented new bind -r option
-
-				   9/20
-				   ----
-builtins/jobs.def
-	- new `disown' builtin
-
-documentation/{bash.1,features.texi}
-	- documented new `disown' builtin
+	- include ../posixstat.h instead of <sys/stat.h>
 
-cpp-Makefile
-	- removed support for `MAKE_SHELL' cpp variable
-	- renamed endian.aux to mkendian, look for endian.c in support dir
-	- link in array.o, alias.o, braces.o, bracecomp.o
-	  unconditionally, rely on cpp defines in the files to exclude code
-	- pass -DHAVE_SYS_SIGLIST if sys_siglist is defined
-	- pass -DHAVE_GETCWD if HAVE_GETCWD is defined in machines.h
-	- pass -DHAVE_VFPRINTF_EMUALTION through from machines.h to
-	  compilation
-
-array.c
-	- don't compile in body of file unless ARRAY_VARS is defined
-
-braces.c
-	- don't compile in body of file unless BRACE_EXPANSION is defined
-
-alias.c
-	- don't compile in body of file unless ALIAS is defined
-
-bracecomp.c
-	- don't compile in body of file unless BRACE_EXPANSION and READLINE
-	  are defined
-
-bashline.c
-	- don't compile in body of file unless READLINE is defined
-
-bashhist.c
-	- don't compile in body of file unless HISTORY is defined
-
-siglist.c
-	- don't compile in body of file unless HAVE_SYS_SIGLIST is not
-	  defined
-
-getcwd.c
-	- don't compile in body of file unless HAVE_GETCWD is not defined
-
-vprint.c
-	- don't compile in body of file unless USE_VFPRINTF_EMULATION
-	  is defined
-
-support/mksysdefs
-	- for SCO machines, define SYSDEF as SCO or SCOv4
-
-lib/readline/readline.[ch]
-	- new function rl_push_macro_input (s), which makes s the current
-	  macro input string
+general.h
+	- include `bashtypes.h' if RLIMTYPE is defined, so we can fetch
+	  a definition of quad_t (or whatever) from <sys/types.h> before
+	  using it in a function prototype
 
-endian.c
-	- renamed to support/endian.c
+Makefile.in, builtins/Makefile.in
+	- updated dependencies
 
-				   9/21
+				   2/18
 				   ----
-lib/readline/complete.c
-	- more file types for the VISIBLE_STATS code:
-		|	FIFOs
-		%	character special devices
-		#	block special devices
-
-lib/readline/isearch.c
-	- made RETURN an alternate search string terminator
-
-builtins/read.def
-	- removed use of stdio -- unbuffered stdio on a dup of fd 0 is
-	  a big loss
-
 builtins/set.def
-	- new function: minus_o_option_value, returns 1, 0, or -1 given
-	  a -o option name (-1 means a bad name)
-	- reorganized list_minus_o_options to be more efficient
+	- new function, set_posix_mode, called when `set [-+]o posix'
+	  or `shopt -[su] -o posix' is executed.  It sets or unsets
+	  $POSIXLY_CORRECT and calls sv_strict_posix
 
-execute_cmd.c
-	- when expanding a here document, use maybe_expand_string rather
-	  than a simple expand_string to try a speed things up a bit
-
-input.c
-	- don't compile in the body of the file unless BUFFERED_INPUT
-	  is defined in config.h
-
-				   9/22
-				   ----
-jobs.h, siglist.h
-	- replaced instances of Solaris with SunOS5
+subst.c
+	- in sv_strict_posix, call posix_readline_initialize only if
+	  the shell is interactive (interactive_shell != 0)
 
 shell.c
-	- changed long option parsing code so that --arg is equivalent
-	  to -arg, when `arg' is one of the recognized multichar options
-
-builtins/history.def
-	- rewrote to regularize the option parsing and use internal_getopt
-	- now allows only one of -awrn to be specified
-	- added a new -p option that adds each of its arguments to the
-	  end of the history list, and deletes the `history -p' history
-	  entry
-
-builtins/trap.def
-	- added new -p option to display specified trap values, or all
-	  trap values if no other arguments supplied
-
-documentation/{bash.1,features.texi}
-	- documented new history -p option
-	- documented new trap -p option
-
-				   9/25
-				   ----
-lib/readline/display.c
-	- fixed up the calcluation of the correct cursor line number (a `+'
-	  and `-' were transposed in the calculation of `nleft')
-
-				   9/26
-				   ----
-general.c
-	- moved isint() here from test.c, renamed to legal_number
-
-general.h
-	- extern declaration of legal_number
-
-test.c
-	- only compile isint() if SHELL is not defined, define it as
-	  legal_number otherwise
-
-trap.c
-	- use legal_number in decode_signal instead of sscanf()
-
-builtins/common.c
-	- moved list_sigs to here from trap.def, renamed to
-	  display_signal_list
-	- changed get_numeric_arg to use legal_number instead of doing
-	  the parsing and calculation itself
-
-builtins/common.h
-	- extern declaration for display_signal_list
+	- if we are acting like `sh', call posix_readline_initialize
+	  if the shell is interactive
+	- moved the code that does posix.2 mode initialization after
+	  interactive_shell is set, and call posix_readline_initialize
+	  if interactive_shell is non-zero
 
-builtins/{trap.def
-	- changed to use display_signal_list for trap -l
-
-builtins/kill.def
-	- changed to use display_signal_list for kill -l
-	- added new kill -n signum option
-
-documentation/bash.1
-	- added description of new kill -l signame functionality
-	- added description of new kill -n signum feature
-	- added description of new enable -d option
-
-builtins.h
-	- added a char *handle member to `struct builtin' for later use
-	- new flags value: BUILTIN_DELETED
-
-builtins/mkbuiltins.c
-	- added code to inintialize the `handle' member to null in the
-	  static builtin array definition
-
-builtins/common.c
-	- don't `find' a builtin if the BUILTIN_DELETED flag is set in the
-	  flags word of the struct builtins array
-
-builtins/enable.def
-	- new option -d to remove a builtin loaded with -f, depends on
-	  HAVE_DLCLOSE
-
-cpp-Makefile
-	- pass -DHAVE_DLCLOSE through from machines.h to compilation
-
-machines.h
-	- change SunOS4 machine description to define HAVE_DLCLOSE
-
-				   9/27
-				   ----
-shell.c
-	- split shell exit code off into a separate function: exit_shell(status)
+bashwait.h
+	- renamed to unionwait.h, since that is what it defines
 
-builtins/exec.def
-	- rewrote for clarity and speed and to use the builtin getopt()
-	- added new -a, -c, and -l options
-	- now calls exit_shell if shell_execve fails and the shell is not
-	  interactive
+posixwait.h
+	- moved the POSIX 1003.1 job control defines here from jobs.h
 
-documentation/bash.1
-	- documented the new options to `exec'
+jobs.h
+	- include `posixwait.h'
 
-				   9/28
+				   2/19
 				   ----
-builtins/exec.def
-	- if the execve fails and the shell is not going to exit, reinitialize
-	  traps and signals
-	- only call end_job_control if subshell_environment != 0
-	- exec should exit unconditionally if the execve fails and
-	  subshell_environment != 0
+braces.c
+	- if SHELL is defined, pass the contents of new-style command
+	  substitution through without expanding brace constructs between
+	  the parens -- let the subshell do it
 
 subst.c
-	- if valid_brace_expansion_word fails, make sure `temp' is set to
-	  NULL before trying to free it after the `goto bad_substitution'
+	- when brace-expanding words, preserve the flags (word->flags) if
+	  brace expansion does not change the word.  This fixes the problem
+	  of things like
 
-cpp-Makefile
-	- add $(CPPFLAGS) to the compilation flags when making `mksignames'
+		 local -a avar=( ${PATH//: } );
 
-documentation/features.texi
-	- fixed a typo in the tilde expansion section
-
-				   9/29
-				   ----
-machines.h
-	- DEC OSF/1 has the dlopen/dlsym/dlclose set of library functions
-
-shell.c
-	- don't execute /etc/profile if -noprofile given
-
-builtins/pushd.def
-	- new file, pushd/popd/dirs split off from cd.def
-	- replaced calls to sscanf with calls to legal_number
-
-builtins/Makefile, cpp-Makefile
-	- changes for pushd.def
-
-config.h, config.h.mini
-	- ALLOW_RIGID_POSIX_COMPLIANCE is no longer used
-
-subst.c, variables.c
-	- GETOPTS_BUILTIN is no longer used
-
-variables.c
-	- if the first character of argv[0] is not a `/', search the path
-	  and canonicalize the result to find out how to set $BASH
-
-				   10/2
-				   ----
-builtins/enable.def
-	- changed enable_shell_builtin to use builtin_address_internal to find
-	  the builtin rather than searching the list itself
-	- list_some_builtins skips a builtin if flags & BUILTIN_DISABLED != 0
-	- rewrote dyn_load_builtin to take a list of names to load from a
-	  single filename
-	- don't dlclose the shared object in dyn_unload_builtin unless its
-	  reference count drops to 0
-
-builtins/test.def
-	- don't bother making new copies of everything in the argument list
-	  when constructing the argc and argv for test_command; just make
-	  sure not to free anything but ARGV
-
-				   10/3
-				   ----
 bashline.c
-	- remove C-e binding in vi movement mode which switches into emacs
-	  mode
+	- have shell_expand_line pass a copy of rl_line_buffer to expand_string
+	  in case there are substitution errors and the string gets freed
 
-general.c
-	- make xfree only try to call free on non-null strings
-
-				   10/4
+				   2/20
 				   ----
-builtins/read.def
-	- new `-e' option that uses readline to read the line
-	- if one of the arguments is not a legal variable name, print an
-	  error message and return failure
-
-builtins/reserved.def
-	- changed the `Variables' to `variables' so `help variables' works
+expr.c
+	- make sure that expland and explor set `lasttok' to LAND and LOR,
+	  respectively, if they parse `&&' or `||'.  This makes the
+	  precedence code work right
 
 subst.c
-	- set startup_state to 2 in child of command substitution to try to
-	  avoid some unneeded forks
+	- changes so that non-interactive shells exit immediately when a
+	  parameter that is unset is referenced after `set -u' has been
+	  executed causes the shell to exit immediately
 
-trap.c
-	- removed call to reset_terminating_signals in restore_original
-	  signals; callers are now required to take care of that themselves,
-	  if necessary
-
-execute_cmd.c, subst.c
-	- added necessary calls to reset_terminating_signals before calls
-	  to restore_original_signals
-
-execute_cmd.c
-	- when executing a null command in a subshell, don't bother passing
-	  a string to make_child; just pass NULL
-	- in execute_builtin_or_function, don't add so many unwind-protects
-	  if `subshell' == 1
-	- in command_substitute, call cleanup_the_pipeline to discard the
-	  old pipeline, so pipeline_pgrp does not get set to 0 in
-	  start_pipeline, which is called by make_child via making_children
-
-jobs.c, nojobs.c
-	- new function, ignore_tty_job_signals, to set SIGTTIN, SIGTTOU, and
-	  SIGTSTP to SIG_IGN
-	- new function, default_tty_job_signals, to set those signals
-	  to SIG_DFL
-	- new function, cleanup_the_pipeline to free up the_pipeline and
-	  set it to NULL
-
-				   10/5
+				   2/21
 				   ----
-builtins/history.def
-	- the history -p option is now -s (to sort of parallel the ksh
-	  print -s option)
-	- the -s option now combines all of its arguments into a single
-	  string and appends the string to the history list
-	- new history -p option to history expand each argument and print
-	  the result without modifying the history list
-
-documentation/bash.1, documentation/features.texi
-	- documented the new -e option to read
-	- documented the new history -s and -p options
-	- documented the new cd and pwd -L options
-
-builtins/cd.def
-	- changed cd and pwd to use internal_getopt
-	- added the -L option to cd and pwd to follow symlinks (like if
-	  set +P were issued)
-
-builtins/pushd.def
-	- added text for dirs +N and dirs -N to the dirs builtin long doc
-	- added -v option to dirs to print dirstack one dir per line with
-	  stack index prepended
+flags.c
+	- if `set -r' is executed, call maybe_make_restricted so that $PATH
+	  and $SHELL are made read-only
 
-				   10/6
-				   ----
 execute_cmd.c
-	- split the command searching code into a new function:
-	  search_for_command
-	- removed a bunch of dead code from shell_execve
-	- removed call to reset_terminating_signals when executing a function
-	  or builtin in a subshell (either via (xxx) or xxx &)
-	- don't add unwind protects at all in execute_function if subshell == 1
+	- if `set -e' has been executed, and we're inverting a pipeline's
+	  return status with `!', set CMD_IGNORE_RETURN so a failing
+	  command does not inadvertently cause the shell to exit before
+	  the exit status can be inverted.  This is probably only a problem
+	  with the `eval' builtin.
 
-lib/readline/chardefs.h
-	- new macro ALPHABETIC(c), returns 1 if c is a letter or digit
-
-lib/readline/readline.c
-	- don't call abort() in rl_change_case(); it's impolite in a
-	  library function
-	- new macro, SWAP, used to swap values of two integers
-	- changed alphabetic to use ALPHABETIC, made it slightly faster
-	- modified rl_change_case() so that word capitalization is the
-	  same as GNU Emacs
-
-lib/readline/search.c
-	- in noninc_dosearch, don't reset the history positition to what
-	  it was if we're currently in vi editing mode (as per Posix.2
-	  `/' and `?' vi-mode editing commands)
-
-				   10/7
-				   ----
-builtins/common.c
-	- changed single_quote and double_quote to use char pointers
-	  rather than string indexing
-	- new function backslash_quote(string), which quotes special
-	  characters in STRING using backslashes
-
-				   10/8
+				   2/24
 				   ----
-alias.h
-	- added an extern declaration for alias_expand_word
-
-parse.y
-	- broke the alias expansion code off into a function
-	  alias_expand_token; its return value says whether to re-read
-	  a token or go on
-	- changed the alias expansion code to handle aliases that expand
-	  to nothing better
-	- broke the code that does special-case token recognition off into
-	  a function: special_case_tokens
-	- used the new functions to make sure that the special-case tokens
-	  can be the expansion of an alias
-	- made sure that if in `posix mode' that reserved words cannot be
-	  aliased and that all reserved words can be the values of aliases
-	  and be recognized after expansion
-
-				   10/10
-				   -----
-lib/readline/complete.c
-	- replaced #ifdef SHELL code with two new exported readline interfaces:
-	  rl_filename_quoting_function and rl_filename_dequoting_function.
-	  Both return a pointer to char.
-	- new extern variable rl_filename_quote_characters, containing a list
-	  of characters that cause a word to be quoted by the completer if
-	  they appear in a file name
-
-bashline.c
-	- new functions for rl_filename_quoting_function and
-	  rl_filename_dequoting_function.
-	- initialize rl_filename_quoting_function and rl_filename_dequoting_function
-	  in initialize_readline
-	- initialize rl_filename_quote_characters
-
-lib/readline/readline.h
-	- extern declarations for new public interfaces
-	  rl_filename_quoting_function and rl_filename_dequoting_function
-	- added declarations for NO_MATCH, SINGLE_MATCH, and MULT_MATCH for
-	  use by the filename quoting functions
-	- new extern declaration for rl_filename_quote_characters
-
-lib/readline/history.c
-	- made a version of single_quote be compiled in if SHELL is not
-	  defined
-	- the `q' and `x' modifiers are now compiled in by default, not
-	  just if SHELL is defined
-
-				   10/11
-				   -----
-subst.c
-	- string_quote_removal was being a little overzealous in stripping
-	  things within embedded quoted strings when `quoted' was == 1.
-	  Only remove one level of quotes each time through the function
-	  This fixes the problem of quotes being stripped incorrectly in
-		var="The text \"hello\" should show up inside double quotes."
-
-lib/readline/history.c
-	- made the behavior of single quotes inhibiting history expansion
-	  configurable with a variable: history_quotes_inhibit_expansion,
-	  not just shell-specific
-	- added a new variable: history_search_delimiter_chars, which is a
-	  list of characters that can also delimit a history search string
-
-lib/readline/history.h
-	- extern declaration of history_quotes_inhibit_expansion
-	- extern declaration of history_search_delimiter_chars
-
-bashhist.c
-	- set history_quotes_inhibit_expansion to 1 in bash_initialize_history
-	- initialize history_search_delimiter_chars to ";&()|<>"
-
-lib/readline/doc/{rltech,hstech}.texinfo
-	- documented new readline and history library interfaces
-
-parse.y
-	- split the part of read_token that reads a single word off into
-	  a new function: read_token_word
-
-lib/readline/chardefs.h
-	- include <string.h> by default; only check HAVE_STRING_H if
-	  HAVE_CONFIG_H is defined
-
-				   10/12
-				   -----
-parse.y
-	- moved the `RESET' code out of read_token into reset_parser
-	- rewrote some of decode_prompt_string to make it more efficient
-	- rewrote more of read_token_word to make it more efficient
-	- make shell_getc cast its result to `unsigned char' before
-	  returning it.  This fixes the problem of \255 appearing in a
-	  line
-
-machines.h
-	- new entry for m68k machines running Linux
-
-				   10/13
-				   -----
-builtins/exec.def
-	- use search_for_command rather than find_user_command to look up
-	  the path to exec, so the hash table and temp environment are used
-
-variables.c
-	- don't rebuild the export environment after binding a shell
-	  function unless that function is exported
-	- make sure that copy_variable copies arrays correctly, using
-	  dup_array()
-	- in assign_in_env, only call tilde_expand if a `~' appears somewhere
-	  in the value
-
-execute_cmd.c
-	- if we found $PATH in the temp environment in search_for_command,
-	  call find_user_command_in_path instead of find_user_command, so
-	  we don't try to search the temporary env again.  Call
-	  find_user_command as normal if PATH is not in the temp environment
-
-subst.c
-	- char_is_quoted should not be compiled in if READLINE is not
-	  defined
-
-lib/readline/complete.c
-	- add a new external interface: Function *rl_char_is_quoted_p,
-	  which is called to find out whether a word break character is
-	  quoted and should be skipped over when breaking words for
-	  the completer
-
-lib/readline/readline.h
-	- extern declaration for rl_char_is_quoted_p
-
-bashline.c
-	- initialize rl_char_is_quoted_p to char_is_quoted
-
-lib/readline/doc/rltech.texinfo
-	- documented rl_char_is_quoted_p
-
-lib/readline/readline.c
-	- extend the undo records so that a `start' or `end' value of -1
-	  means rl_point and a value of -2 means rl_end.  This is a start
-	  to better support for undoing vi-mode commands like `C'
-
-lib/readline/vi_mode.c
-	- don't save what's entered in insert mode after a `C' command
-	  for later insertion when doing a `redo'
-
-				   10/16
-				   -----
-test.c
-	- rewrote unop() to use a switch statement instead of a call to
-	  strchr
-	- remove #ifdef SHELL blocks by defining getuid, geteuid, getgid
-	  and getegid as references to current_user.{uid,euid,gid,egid}
-	  respectively
-	- change group_member to only fetch the group list once and to use
-	  NGROUPS_MAX or NGROUPS to find the maximum number of groups
-
-documentation/bash.1
-	- fixed description of ${#@} expansion
-
-				   10/17
-				   -----
-support/bashbug.sh
-	- add a `From:' line to the mail message handed to rmail
-
-				   10/18
-				   -----
-test.c
-	- rewrote binary_operator for speed and clarity
-	- removed age_of, added arithcomp(), filecomp() to support new
-	  binary_operator
-	- removed support for `-l string'
-
-documentation/bash.1
-	- removed mention of `-l string' from `test' description
-
-				   10/19
-				   -----
-cpp-Makefile
-	- pass PROGRAM as the double-quoted shell name to compilation of
-	  shell.c and error.c
-
-				   10/20
-				   -----
-support/bashbug.sh
-	- don't try to use ${word:-expansion}; ultrix sh doesn't understand it
-
-hash.c
-	- new function: flush_hash_table (table, free_data) to delete the
-	  contents of a given hash table.  *free_data is called to free
-	  each item's data, if free() is inappropriate
-
 builtins/hash.def
-	- new functions free_hashed_filenames and free_filename_data to
-	  flush the table of hashed filenames
-
-subst.c
-	- change sv_path to call flush_hashed_filenames directly
-
-variables.c
-	- only sort arrays of variables or functions for the environment or
-	  `set' output if `posixly_correct' is set.  sh does it; ksh does
-	  not, and there's no real requirement to do so
-	- rewrote delete_all_variables so it looks like flush_hash_table
-
-trap.c
-	- two new flag values for the `sigmodes' array: SIG_INPROGRESS,
-	  which is set for sigmodes[sig] while a trap handler for sig
-	  is executing, and SIG_CHANGED, which is set if a new trap
-	  value is set when SIG_INPROGRESS is set.  This should obviate
-	  the need to set the trap value to IMPOSSIBLE_TRAP_HANDLER while
-	  the trap handler is executing
-
-alias.c
-	- rewrote delete_all_aliases so it looks like flush_hash_table
-
-				   10/21
-				   -----
-alias.c
-	- changed delete_all_aliases to call flush_hash_table directly
-	  and use free_alias_data as the `free function' argument
-
-variables.c
-	- changed delete_all_variables to call flush_hash_table directly
-
-tests/run-test, tests/test-tests
-	- new scripts to run tests of the `test' builtin as part of the
-	  regression test
-
-				   10/24
-				   -----
-bashline.c
-	- initialize_hostname_list needs to look for HOSTFILE first
-	- hostname list is no longer sorted
-	- replaced binary search in hostnames_matching with a simple
-	  linear search
-	- made the code that reads hostnames skip over the first word on
-	  a line only if its first character is a digit, assuming it's
-	  an Internet address
-
-copy_cmd.c
-	- removed copy_select_command; overload copy_for_command, since the
-	  select and for command structs are exactly the same
-
-make_cmd.c
-	- combined make_for_command and make_select_command into a new
-	  function, make_for_or_select
-	- rewrote make_here_document to remove the unneeded `switch' statement
-
-builtins/common.c, builtins/hash.def
-	- moved remove_hashed_filename from common.c to hash.def
-
-builtins/common.c
-	- remove the \r from error message printed by get_working_directory
-	- change parse_and_execute to call dispose_fd_bitmap directly then
-	  discard the `pe_dispose' unwind-protect frame rather than running
-	  the frame
-
-builtins/set.def
-	- changed how set -o options are set and retrieved, using set and get
-	  functions to avoid all that special-case inline code
-
-				   10/26
-				   -----
-test.c
-	- added unary operator `-o', which returns true of the shell option
-	  name given as an argument is set
-
-lib/readline/readline.c
-	- added a definition of set_lines_and_columns to be called if the
-	  library is not compiled -DSHELL
-
-shell.c
-	- added a --verbose startup long option
-	- renamed `--nolineediting' to `--noediting'
-
-				   10/27
-				   -----
-lib/readline/util.c
-	- new file, for readline utility functions
-
-lib/readline/readline.c
-	- moved a bunch of functions to util.c
-
-subst.c
-	- make sure set_sigint_handler is called only by the subshells doing
-	  command and process substitution
-
-builtins/read.def
-	- make sure that rlbuf is initialized to null
-
-trap.c
-	- new flag for sigmodes[] members: SIG_IGNORED, set when signal is
-	  ignored, even if it's special or untrappable
-	- new function: signal_is_ignored (sig), which returns 1 if SIG
-	  has been ignored with trap ''
-
-builtins/trap.def
-	- changed to use the builtin getopt
-
-shell.c
-	- added a check to sigint_sighandler for whether or not SIGINT has
-	  been ignored with trap '' in an interactive shell.  This fixes
-	  the problem with `read' being interruptible in an interactive
-	  shell even if SIGINT is being ignored
-
-				   11/8
-				   ----
-lib/readline/Makefile
-	- added definition of INSTALLED_HEADERS, just in case
-
-				   11/11
-				   -----
-variables.c
-	- change assign_in_env so that it doesn't use savestring ("") to
-	  set `value' to a dummy value; don't call strcpy if there's
-	  nothing to copy
+	- add a missing argument of 0 to add_hashed_command
 
-				   11/15
-				   -----
-general.h
-	- new defines, legal_variable_starter and legal_variable_char
-	- new define SIGRETURN(n) which encapsulates the VOID_SIGHANDLER
-	  differences when returning from a signal handler
-
-general.c, variables.c, subst.c, expr.c
-	- use legal_variable_starter and legal_variable_char
-
-shell.c, nojobs.c, trap.c, jobs.c, builtins/suspend.def
-	- change to use SIGRETURN macro
-
-subst.c
-	- massive changes to clean up the code and remove unused code and
-	  variables
-	- expanded the ${#param} code so that all of the shell special
-	  variables may have their length taken
-
-tests/run-tilde
-	- new test for tilde expansion
-
-bashline.c
-	- fix a bug in command_subst_completion_matches: make sure that
-	  `matches' is static
-
-parse.y
-	- don't print a prompt when not using readline if the current
-	  input type is st_string
-
-machines.h
-	- add -DINT_GROUPS_ARRAY to SYSDEP_CFLAGS on ultrix
-	- fixes to the cray machine description from Bill Jones
-
-braces.c
-	- fixed a bug in brace_gobbler that prevented a backslash from
-	  escaping an open brace
-
-				   11/16
-				   -----
-tests/braces-tests, tests/run-braces
-	- new regression tests for brace expansion
-
-builtins/pushd.def
-	- new -p option for `dirs' that prints dirstack on per line
-	  without numbers
-
-				   11/17
-				   -----
-command.h
-	- move redirection error values here from execute_cmd.c
-	- new defines INPUT_REDIRECT and OUTPUT_REDIRECT
-
-shell.c
-	- make the default MAINTAINER `bash-maintainers@prep.ai.mit.edu'
-
-execute_cmd.c
-	- many changes to clean up the code and remove unused variables and
-	  functions
-	- new functions: redirection_error, find_in_path_element,
-	  find_absolute_program
-	- fixes to redirection error reporting, so things like exec 4<&y*
-	  and exec 4<&$FOO are displayed correctly
-	- removed the `lexical_scoping' code
-
-flags.c, flags.h
-	- removed the lexical_scoping code and variable
-
-documentation/{bash.1,features.texi}
-	- removed the description of the `-l' option to `set'
+builtins/kill.def
+	- job identifiers can be used in non-interactive shells as long as
+	  job control has been turned on with `set -m'
 
 jobs.c
-	- don't try to open /dev/tty to get the controlling tty, use
-	  fd 2 like other job control shells
-
-lib/readline/vi_mode.c
-	- when using `d%', make sure the matching character found by
-	  the `%' is deleted by the `d'.  Ditto for `c%'.
-	- stub function for vi undo: rl_vi_undo.  Right now it just
-	  calls rl_undo_command
-
-lib/readline/vi_keymap.c
-	- change to call rl_vi_undo instead of rl_undo_command
-
-lib/readline/readline.h
-	- extern declaration for rl_vi_undo
-
-				   11/21
-				   -----
-execute_cmd.c
-	- fix to print_select_list to avoid a possible divide-by-zero error
-	  and subsequent core dump
-	- fix to execute_select_command to just return 0 if there is no
-	  select list
-
-parse.y
-	- remove `in' from the list of tokens that cannot take trailing
-	  semicolons
-
-builtins/read.def
-	- make sure that leading IFS whitespace is removed before calling
-	  get_word_from_string the first time.  This matters when IFS is
-	  not " \t\n" but non-null
-	- make sure the array code uses IFS to split the input string
-	  before assigning it to the array
-
-				   11/22
-				   -----
-parse.y
-	- make sure that if \nnn expands to CTLESC or CTLNUL, the char
-	  is protected by a CTLESC
-	- new variable `promptvars', which, if non-zero, causes all the
-	  variable expansions to be performed in decode_prompt_string.
-	  If zero, only quote removal is performed.
-
-builtins/source.def
-	- new variable: `source_uses_path', set to 1 by default.  If
-	  non-zero, the `.' builtin uses $PATH to find the script to
-	  source
-
-builtins/getopt.h
-	- cut out everything not needed by bash
-
-builtins/getopt.c
-	- cut out everything not needed by bash
-
-builtins/getopts.def
-	- removed the call to getopt_set_posix_option_order, which is no
-	  longer necessary
+	- we want to be notified of stopped jobs if job_control is non-zero,
+	  even if the shell is not interactive
 
 execute_cmd.c
-	- split the code that writes out here documents to files out into
-	  a separate function
+	- make sure shell_execve returns EX_NOTFOUND if execve fails and
+	  errno is set to ENOENT
+	- makes sure execute_builtin saves the temporary environment to
+	  builtin_env for the `eval' builtin, since it can destroy the
+	  temporary environment when it calls parse_and_execute
 
-				   11/23
-				   -----
-builtins/getopt.c, builtins/getopt.h, builtins/getopts.def, subst.c
-	- prefix all of the getopt variables and functions with `sh_'
-	  (that is, optind becomes sh_optind and getopt becomes sh_getopt)
-	  to avoid confusion with a system's getopt(3) implementation
-
-subst.c
-	- new functions: parameter_brace_substring and verify_substring_values
-	  and changes to expand_word_internal to support the ksh-93
-	  ${var:exp1:exp2} substring syntax
-
-documentation/bash.1
-	- documented the new ${var:exp1[:exp2]} syntax
-
-				   11/25
-				   -----
-builtins/setattr.def
-	- don't allow readonly -n at all
-
-array.c
-	- split array_to_string into two parts; a new function
-	  array_to_string_internal does the real work
-	- new function array_subrange () to return a subset of the elements
-	  in an array
-
-subst.c
-	- augmented the substring code to handle the positional parameters
-	  and array variables
-	- made quote_list and dequote_list return their WORD_LIST *
-	  arguments so they can be used like
-
-		z = string_list ((quoted ? quote_list (l) : l), xxx);
-
-	- augmented the ${xxx} expansion code to do indirect variable
-	  references if the first character of the variable name is `!'
-
-tests/new-exp.tests
-	- added regression tests for substring expansion
-	- added regression tests for indirect variable references
-
-				   11/28
-				   -----
-builtins/set.def
-	- added set -o hashfunc and set -o onecmd, synonyms for set -h
-	  and set -t, respectively
-
-builtins/shift.def
-	- changed shift so that the positional parameters are not changed
-	  if the argument is > $# (this is ksh and Posix.2, unlike sh)
-
-documentation/bash.1
-	- documented true behavior of `shift'
-
-lib/readline/kill.c
-	- split the kill ring management code and the kill commands out
-	  from readline.c into this file
-
-lib/readline/undo.c
-	- split the code that does undoing out of readline.c into this file
-
-shell.c
-	- force the shell to exit with status 127 if a longjmp back to
-	  run_one_command occurs with bash -c
-	- force the last command exit status to 1 if a
-	  longjmp (top_level, DISCARD) is performed
-
-builtins/read.def
-	- don't throw away partial lines after reading EOF
-
-subst.c
-	- command substitution should not inherit the -e flag
-
-builtins/source.def
-	- make sure to set the exit status correctly when in posix mode
-	  and the filename argument to `.' does not exist
-
-				   11/29
-				   -----
-lib/readline/input.c
-	- split the input buffering and character input code out of
-	  readline.c to here
-
-lib/readline/macro.c
-	- moved the keyboard macro management code to here from readline.c
-
-lib/readline/readline.c
-	- removed the STATIC_MALLOC code
-
-lib/readline/rltty.c
-	- return -1 in POSIX get_tty_settings if tcgetattr returns -1 and
-	  errno != EINTR, even if output is being flushed
+bashhist.c
+	- new variable: hist_last_line_added, set to 1 if the last command
+	  line was added to the history successfully as a separate entry.
+	  Used by `history' and `fc'
 
-				   12/1
+				   2/25
 				   ----
-machines.h
-	- fixes to the hpux_8 and hpux_9 machine descriptions
-
 trap.c
-	- make run_exit_trap return the right exit status
-	  (last_command_exit_value)
-	- run_exit_trap no longer preserves the value of last_command_exit_value
-	  around the execution of the trap commands
-	- run_exit_trap now turns off SIG_TRAPPED and sets SIG_INPROGRESS
-	  and will not try to run anything if SIG_INPROGRESS is set
-
-trap.h
-	- change definition for run_exit_trap
-
-shell.c
-	- call run_exit_trap only if trap[0] is set and not ignored
-
-builtins/exit.def
-	- make sure we only source the .bash_logout file once, even if it
-	  contains a call to `exit'
+	- save line number before executing trap command, because
+	  parse_and_execute resets it to 1 and the trap command might
+	  want it
 
 execute_cmd.c
-	- if we run an exit trap in a (...) user subshell, allow it to
-	  override the exit status of the subshell
-
-lib/readline/readline.c
-	- made rl_delete_text bounds check its `to' argument, and limit
-	  it at rl_end
-
-lib/readline/vi_mode.c
-	- make rl_vi_subst call `rl_delete_text' directly for the `s'
-	  command
-
-support/mksysdefs
-	- define a new variable for the sysdefs.h file for ISC machines:
-	  ISC_release, which can be ISC_2, ISC_3, or ISC_4
-
-machines.h
-	- don't `#undef' HAVE_GETCWD on ISC 4.x machines
-	- `#undef' HAVE_RESOURCE on ISC 4.x machines
-
-support/mkversion.c
-	- include "posixstat.h" rather than <sys/stat.h> for the benefit
-	  of ISC machines
+	- change to executing_line_number to return trap_line_number if
+	  the shell is currently running a trap
 
-				   12/5
-				   ----
-lib/readline/complete.c
-	- changed username_completion_function so that a null username
-	  generates a list of all users as possible completions
-
-lib/readline/readline.h
-	- added definitions for STREQ, STREQN
-
-lib/readline/{search.c,isearch.c,kill.c}
-	- removed private definitions of STREQ, STREQN
-
-execute_cmd.c
-	- in find_user_command_internal, just return a copy of the pathname
-	  passed as an argument if there is no PATH
-
-				   12/6
-				   ----
-siglist.h
-	- NetBSD 1.0 does not need a define for strsignal()
-
-				   12/8
-				   ----
-subst.c
-	- removed assignment_name, word_list_quote_removal, word_quote_removal,
-	  and sub_append_number -- unused functions
-	- removed some unexecuted code from expand_word_internal
-
-				   12/9
+				   2/26
 				   ----
 execute_cmd.c
-	- if PATH is set to the empty string, find executables in the
-	  current directory
-
-shell.c, parse.y, trap.c
-	- before setting the SIGINT sighandler unconditionally to one of
-	  sigint_sighandler or termination_unwind_protect, check that it
-	  is not ignored.  Now trap '' 2 really sets the SIGINT signal
-	  handler to SIG_IGN. [In 1.14.3]
-
-trap.c
-	- rewrote set_sigint_handler to use SIG_IGNORED rather than checking
-	  against IGNORE_SIG
-	- changed ignore_signal, run_exit_trap, maybe_call_trap_handler, and
-	  run_trap_internal to check SIG_IGNORED
-
-shell.c
-	- removed the check for signal_is_ignored(SIGINT) in sigint_sighandler
-
-				   12/11
-				   -----
-sig.c, sig.h
-	- new files, moved signal-related definitions and code here from
-	  shell.c, general.h, general.c, jobs.c, jobs.h, externs.h
-
-unwind_prot.c, shell.h, nojobs.c
-	- include sig.h
-
-				   12/12
-				   -----
-jobs.c
-	- only break out of loops if SIGTSTP was used to stop a job in the
-	  loop, the shell is currently interactive, and job control is on.
-	  SIGSTOP does not break loops. [In 1.14.3]
-
-				   12/13
-				   -----
-expr.c
-	- bases < 2 or > 36 are now accepted without silently being reset
-	  to 10
-
-braces.c
-	- make sure array_concat copies the array it returns if one of
-	  the arguments is null [In 1.14.3]
-
-				   12/14
-				   -----
-subst.c
-	- split the ${...} expansion code out into a separate function,
-	  static char *parameter_brace_expand()
-	- changes to array_value so that any variable can be referred to
-	  as an array with an integer subscript.  The value will be
-	  returned if a non-array variable is referred to as ${var[0]};
-	  if the subscript is > 0 a null string is returned
-
-				   12/15
-				   -----
-machines.h
-	- fixes to freebsd description for FreeBSD 2 [In 1.14.3]
-
-support/bashbug.sh
-	- changed to use /usr/lib/sendmail if present or /usr/sbin/sendmail
-	  if present, defaulting to rmail [In 1.14.3]
-
-bashhist.c
-	- HISTFILESIZE now controls how large the history file is after
-	  it is written.  After saving the shell history, sv_histfilesize
-	  will truncate it if necessary.  history -w can override this.
-
-documentation/bash.1
-	- documented change to treatment of HISTSIZE when saving history
-
-lib/malloc/malloc.c
-	- removed the calls to sigsetmask() in malloc().  This should
-	  result in a speed improvement
-
-				   12/19
-				   -----
-builtins/enable.def
-	- don't allow -f or -d in a restricted shell
-
-builtins/alias.def
-	- rewrote alias and unalias to use the internal getopt
-	- added -p option to print the alias list to alias
-	- fixed up the documentation for `alias'
-
-documentation/{bash.1,features.texi}
-	- updated the documentation for `alias'
-	- updated the documentation for $_
-
-array.c,array.h
-	- new function `empty_array (ARRAY *a)' removes all of the
-	  elements in a without destroying the array variable in
-	  preparation for overwriting it.  Used by read -a.
-
-builtins/read.def
-	- call empty_array() before assigning list of values with read -a
-	- changed dispose_array to use empty_array to destroy the
-	  array elements
-
-variables.c
-	- set $_ to argv[0] at variable initialization time
-
-				   12/20
-				   -----
-subst.c
-	- broke the pattern removal code into a few separate functions:
-	  getpatspec to get the pattern specifier, getpattern to do the
-	  necessary word expansions and return the pattern to be matched
-	- new function: parameter_list_remove_pattern(), which implements
-	  the ${param[#%][[%#]]pattern} where param is `@' or `*'
-
-documentation/bash.1
-	- documented new pattern removal functionality for the positional
-	  paramters
-
-general.c
-	- new function strsub (s, pat, rep, gflag) replaces PAT with REP
-	  in S.  All occurrences are replaced if GFLAG != 0; the first is
-	  replaced otherwise
-	- changed strindex() to avoid multiple calls to strnicmp by
-	  checking first character of the string
-
-builtins/fc.def
-	- changed to use builtin_getopt with a check for fc numbers as
-	  arguments
-	- some code rearranging for efficiency and clarity
-	- fc_dosubs now just calls strsub(); fc_replace is gone
-
-				   12/21
-				   -----
-subst.c
-	- new function, match_pattern, which matches a shell globbing
-	  pattern anywhere in a string and returns the boundaries of the
-	  match
-
-lib/readline/readline.h
-	- declarations for rl_insert_command and rl_backward_char_search
-
-lib/readline/readline.c
-	- new function rl_insert_comment, no longer vi-mode-specific
-	- new function rl_backward_char_search
-
-lib/readline/bind.c
-	- comment-begin now sets the comment char for emacs and vi modes
-	- variable holding the value is now _rl_comment_begin
-
-lib/readline/vi_mode.c
-	- move the `comment-begin' stuff to readline.c and bind.c
-
-lib/readline/funmap.c
-	- vi-comment is now insert-comment
-	- new bindable command character-search-backward
-
-lib/readline/vi_mode.c
-	- command mode `#' now invokes rl_insert_comment
-
-lib/readline/emacs_keymap.c
-	- M-# now bound to insert-comment
-	- M-space now bound to set-mark
-	- M-^] now bound to character-search-backward
-
-bashline.c
-	- posix_readline_initialize now calls rl_variable_bind to set the
-	  value of comment-begin rather than directly modifying
-	  _rl_comment_begin
-
-documentation/{bash.1,readline.3}, lib/readline/doc/rluser.texinfo
-	- added description of new bindable `insert-comment' command
-	- documented new M-space emacs mode binding
-	- documented new character-search-backward command and default
-	  emacs mode binding to M-C-]
-
-shell.c
-	- only call posix_initialize if posixly_correct is set
-
-				   12/22
-				   -----
-cpp-Makefile
-	- make $(Program) depend on $(srcdir)/.distribution, for the
-	  benefit of systems where `make' does not have VPATH support
-	  [In 1.14.3]
-
-jobs.c
-	- if a foreground job is killed by SIGINT while job control is
-	  active, print a newline to compensate for the kernel printing
-	  ^C without one [in 1.14.3]
-
-bashline.c
-	- make sure bashline_reinitialize resets rl_completion_entry_function
-	  to NULL, as the comment says it should [In 1.14.3]
-
-				   12/23
-				   -----
-test.c
-	- fix a problem that caused core dumps if a `)' was missing in a
-	  parenthesized expression [In 1.14.3]
-
-jobs.c
-	- broke the code the manages the manipulation of the job table and
-	  process status out of flush_child into a new function waitchld()
-	- flush_child now just calls waitchld() with a parameter that tells
-	  it not to block
-	- wait_for calls waitchld() with the pid it's looking for and tells
-	  it to block (don't call it with WNOHANG)
-	- cleaned up wait_for considerably -- turned the wait_loop: label
-	  stuff into a do-while loop and removed the setting of job status
-	  (that's now done only by waitchld).  wait_for now calls
-	  waitchld continuously until the job it is interested in is
-	  marked JDEAD.
-
-				   12/28
-				   -----
-subst.c
-	- fixed expand_word_internal so that any word that expands into
-	  nothing and contains a double-quoted $@ is removed, like sh
-	  and ksh
-	- new function: expand_string_for_rhs, which calls expand_word_internal
-	  with a variable that lets it find out whether or not a $@ appeared
-	  in the WORD in ${paramOPword} when expanding it, so that "$@"
-	  and various other things are handled correctly on the rhs
-	- added params for parameter_brace_expand to tell expand_word_internal
-	  if a quoted $@ was processed as part of the rhs (or even the lhs);
-	  these new params are passed along to parameter_brace_expand_rhs
-	- pass the right value of quoted to parameter_brace_expand_rhs from
-	  parameter_brace_expand.  expand_string_for_rhs doesn't need to know
-	  whether the brace expression is quoted
-
-				   12/31
-				   -----
-support/printenv
-	- now an official part of the distribution, moved from CWRU/misc
-	  [in 1.14.3]
-
-cpp-Makefile
-	- copy support/printenv into the `tests' directory when making tests
-	  [in 1.14.3]
-	- change to understand GCC_STANDARD [in 1.14.3]
-
-support/bashbug.sh
-	- fixed a typo that caused it to not parse correctly [in 1.14.3]
-
-machines.h
-	- define GCC_STANDARD if the standard `cc' is gcc and you don't want
-	  to use the compiler named `gcc' for some reason [in 1.14.3]
-
-				    1/2
-				    ---
-general.h
-	- added FS_DIRECTORY to the list of flags that file_status returns
-
-execute_cmd.c
-	- changed find_in_path_element to return null if the flags argument
-	  specifies FS_EXEC_ONLY and the file is not executable
-	- return FS_DIRECTORY from file_status if the argument specifies a
-	  directory
-	- new function, is_directory (char *), which returns non-zero if the
-	  filename argument is a directory
-
-execute_cmd.h
-	- extern declaration for is_directory
-
-flags.c, flags.h
-	- hashing_disabled and locate_commands_in_functions were removed,
-	  hashing_enabled added
-
-execute_cmd.c, builtins/common.c. builtins/hash.def
-	- use hashing_enabled instead of hashing_disabled, and reverse
-	  the sense of tests of it
-
-documentation/bash.1, documentation/features.texi
-	- changed description of `set -h/set -o hashcmds', removed
-	  set -d/set -o nohash
-
-bashline.c
-	- changed command_word_completion_function to return matches if names
-	  are directories as well as if they are executable files
-
-support/mksysdefs
-	- look for `ranlib' in $PATH before searching the file system;
-	  look in /usr/gnu/bin for it; default to `:' if not found
-
-general.c
-	- change ansicstr to accept a second argument telling it whether to
-	  recognize \c and to pass back a non-zero value in it if \c is
-	  seen
-
-general.h
-	- changed extern declaration of ansicstr
-
-subst.c
-	- call ansicstr with an extra argument
+	- change to time_command so that a `real' value of 0 does not
+	  cause a divide-by-zero error when computing cpu percentage
 
-				    1/3
-				    ---
-builtins/echo.def
-	- rewrote to use ansicstr() with the new argument
-
-				    1/4
-				    ---
-trap.c
-	- changed instances of signal() to set_signal_handler() [in 1.14.4]
-	- combined reset_signals and restore_signals into a single function,
-	  since they were essentially identical
-
-subst.c
-	- if set -u is set, references to the positional parameters now
-	  generate errors if that parameter is not set [in 1.14.4]
-
-lib/*/Makefile, builtins/Makefile
-	- since RANLIB can be just `ranlib', just try to run it without
-	  checking that the file exists [in 1.14.4]
+lib/readline/signals.c
+	- if MUST_REINSTALL_SIGHANDLERS is defined, reinstall the SIGWINCH
+	  handler in rl_handle_sigwinch
 
 builtins/set.def
-	- changed `hashcmds' to `hashall'
-
-documentation/{bash.1,features.texi}
-	- changed `hashcmds' to `hashall'
+	- `unset' now rejects attempts to unset names that are not valid
+	  shell identifiers as errors
+	- add a description of `-o history' option to help text
 
-				    1/5
-				    ---
-trap.c
-	- make the loop that restores signal handlers run from signal 0 to
-	  make sure user subshells don't inherit traps on `exit' (to fix
-	  for 1.14.3, change restore_original_signals so that the loop
-	  starts from 0) [in 1.14.4]
-
-variables.c
-	- don't import exported function definitions at startup if the
-	  shell is restricted
-
-builtins/source.def
-	- don't allow use of pathname arguments containing `/' in a
-	  restricted shell
-
-execute_cmd.c
-	- when a shell is spawned to execute a shell script without a
-	  #! line, turn off the -r flag if the shell is restricted
-
-shell.c
-	- added a new long option `--restricted'
-
-documentation/bash.1
-	- added a section on the restricted shell, and documented the
-	  new `--restricted' long invocation option
-
-				    1/7
-				    ---
-shell.c
-	- when using bash -c command, make run_one_command return
-	  last_command_exit_value if a throw_to_top_level with value
-	  EXITPROG occurs [in 1.14.4]
-
-print_cmd.c
-	- make sure to initialize arg_index in the non-varargs implementation
-	  of cprintf [in 1.14.4]
-
-jobs.c
-	- don't try to change the state of the SIGCHLD handler before
-	  calling waitchld() from wait_for, since SIGCHLD is blocked
-	  while this code is executing [in 1.14.4]
-
-				   1/11
-				   ----
-lib/readline/rltty.c
-	- call control_keypad iff the value of a new variable,
-	  _rl_enable_keypad, is non-zero
-
-lib/readline/bind.c
-	- new readline variable `enable-keypad' to control whether readline
-	  tries to manipulate the application keypad
-
-documentation/{bash.1,readline.3}, lib/readline/doc/rluser.texinfo
-	- documented new `enable-keypad' variable
-
-				   1/12
-				   ----
-lib/readline/search.c
-	- make sure to call rl_unix_word_rubout and rl_unix_line_discard
-	  with the correct arguments [in 1.14.4]
-
-make_cmd.h
-	- make sure make_select_command is declared even if SELECT_COMMAND
-	  is not defined
-
-parse.y
-	- make sure the \[ and \] escape sequences are not recognized if
-	  READLINE is not defined [in 1.14.4]
-
-config.h
-	- make sure HISTORY is defined if READLINE is; code moved here
-	  from bashhist.c [in 1.14.4]
-
-bashhist.c
-	- removed check for READLINE being defined without HISTORY; now
-	  in config.h
-	- new function, bash_history_reinit
-
-flags.h, flags.c, builtins/set.def
-	- the -H/-o histexpand flag should be compiled into the shell
-	  only if BANG_HISTORY is defined [in 1.14.4]
-
-subst.c
-	- don't include sv_histchars unless BANG_HISTORY is defined
-	  [in 1.14.4]
-	- if QUOTED is true in parameter_brace_expand_rhs, pre-process the
-	  word on the rhs of the parameter expansion by a call to
-	  string_extract_double_quoted with the STRIPDQ parameter set to 1
-	- new arg for string_extract_double_quoted; causes it to strip
-	  double quotes and alter its backslash handling behavior; designed
-	  to be called from parameter_brace_expand_rhs
-	- changed all other instances of string_extract_double_quoted to
-	  call it with STRIPDQ set to 0, to get old behavior
-
-shell.c
-	- call bash_history_reinit rather than manipulating history
-	  variables directly
-
-variables.c
-	- don't auto-export $BASH [in 1.14.4]
-
-tests/rhs-exp.tests
-	- new test script to check for behavior fixed by changes to
-	  string_extract_double_quoted and parameter_brace_expand_rhs
-
-parse.y
-	- `for' and `select' must now take non-empty lists between
-	  `in' and `;'
-
-				   1/16
-				   ----
-subst.c
-	- fixed string_quote_removal to do double-quoted string processing
-	  itself rather than call string_extract_double_quoted, which
-	  assumes that a call to expand_word_internal or the equivalent
-	  will follow immediately and leaves some backslashes in place,
-	  inappropriately for quote removal
-
-				   1/23
-				   ----
-subst.c
-	- make sure to set `temp' to NULL after it's freed by sub_append_string
-	  in expand_word_internal to keep it from pointing to newly-allocated
-	  memory that will be subsequently freed, causing a `memory freed
-	  twice' error [in 1.14.4]
-
-trap.c
-	- handle the EXIT_TRAP specially in reset_or_restore_signal_handlers,
-	  since in both cases we simply want to free up the trap string and
-	  mark the signal as not trapped
-
-shell.h
-	- added \n to the list of characters in slashify_in_double_quotes
-
-				   1/26
-				   ----
-subst.c
-	- make string_extract_single_quoted and string_extract_double_quoted
-	  `inline'
-	- new function skip_single_quoted, used when we used to call
-	  string_extract_single_quoted and just throw the returned string
-	  away
-	- new function skip_double_quoted for the same purpose
-
-				   1/28
-				   ----
 subst.c
-	- fixed expand_word_internal so that if an assignment word
-	  is expanded, no word splitting is performed [in 1.14.4]
+	- in parameter_brace_patsub, we want backslash removal done on
+	  the replacement if (mflags & MATCH_QUOTED), since the code
+	  in expand_word_internal will not do it.  We need to call
+	  expand_string_unsplit directly, since maybe_expand_string does
+	  not do the right thing
 
-builtins/ulimit.def
-	- some systems lack RLIMIT_CPU; so `#ifdef' its use [in 1.14.4]
-	- some versions of cpp expand parameters like \n if `n' is an
-	  argument to the macro; change `n' to `num' in the definition
-	  of print_rlimtype to compensate [in 1.14.4]
-
-builtins/read.def
-	- make sure the read loop sets `saw_escape' to note that an
-	  escape character was read if CTLESC or CTLNUL is read [in 1.14.4]
-
-shell.c, sig.c
-	- only test interactive_shell before calling maybe_save_shell_history
-	  [in 1.14.4]
-shell.c
-	- include <locale.h> if HAVE_LOCALE_H is defined
-	- call setlocale(LC_ALL, "") at the beginning of main() if
-	  either _POSIX_VERSION or HAVE_SETLOCALE is defined
-
-support/mksysdefs
-	- look for <locale.h>, define HAVE_LOCALE_H if found
-
-cpp-Makefile
-	- pass HAVE_LOCALE_H through from sysdefs.h to the build process
-
-bashhist.c
-	- remove test against interactive_shell in maybe_save_shell_history
-	  [in 1.14.4]
-
-variables.c
-	- moved definition of DEFAULT_MAIL_PATH to config.h
-
-config.h, config.h.mini
-	- now has definition of DEFAULT_MAIL_PATH [in 1.14.4]
-	- changed default value of PATH to
-	  `/usr/gnu/bin:/usr/local/bin:/usr/ucb:/bin:/usr/bin:.'
-	  [in 1.14.4]
-
-documentation/bash.1
-	- updated the example of the default path in the description of
-	  the PATH variable
-
-lib/readline/readline.c
-	- set up a table of `legal' LC_CTYPE values and match the value of
-	  $LC_CTYPE against it when deciding whether or not to default to
-	  eight-bit input and output
-	- check for LC_ALL, LC_CTYPE, and LANG, in that order, for names to
-	  check against the legal_lc_ctype_values table
-
-lib/readline/bind.c
-	- moved strindex, stricmp, and strnicmp to lib/readline/util.c, 
-	  added _rl_ prefix to names
-	- some miscellaneous code cleanups and speedups
-
-lib/readline/rldefs.h
-	- define _POSIX_VDISABLE as _SVR4_VDISABLE if there is a define
-	  for the latter and not the former [in 1.14.4]
-	- moved defining _rl_stricmp and _rl_strnicmp as strcasecmp and
-	  strncasecmp, respectively, here from bind.c, since the functions
-	  are now defined in util.c and used in two files
-
-builtins/common.c
-	- removed the \n case in double_quote, so a backslash is not added
-	  before an existing newline.  The \<newline> pair is removed by
-	  the parser before anything else in the shell gets hold of it
-	  except within single quotes [in 1.14.4]
-
-				   1/30
+				   2/28
 				   ----
-general.c
-	- make sure the string index in canonicalize_pathname never goes
-	  < 0 (it can, in some cases, result in `start' being -1, and `i'
-	  being set to that) [in 1.14.4]
-
-				    2/1
-				    ---
 execute_cmd.c
-	- fix a typo in a call to `access' in AFS-specific code [in 1.14.4]
-
-				    2/3
-				    ---
-builtins/shopt.def
-	- finally added new `shopt' builtin
-
-builtins/umask.def
-	- converted to use builtin_getopt
-
-bashhist.c
-	- support for storing literal newlines in the history list when
-	  command_oriented_history is enabled, rather than using semicolons
-
-builtins/Makefile
-	- changed rule to make a .o file from a .def file to remove the
-	  .c file if the compilation fails
-
-				    2/6
-				    ---
-bashhist.c
-	- added new variable `force_append_history' that will force the
-	  history list to be appended to the history file at shell exit
-	- new function, maybe_append_history, to append any history lines
-	  from the current session to the history file.  Used by history -a
-
-builtins/bind.def
-	- changed the `-d' option to -p
-
-builtins/set.def
-	- added `allenv' as a -o synonym for `set -k'.  Now all of the
-	  single-letter shell options have -o equivalents
-
-builtins/shopt.def
-	- added -p option to display shell options, like other builtins
-	- added `interactive_comments' as a shell option, like set -o
-
-builtins/bind.def, builtins/history.def, builtins/enable.def
-	- changed use of multiple variables to hold option flag settings to
-	  one variable with bits representing flag values
-
-documentation/bash.1, documentation/features.texi
-	- changed the description of the `bind' builtin for the new -p option
-	- changed the description of the `set' builtin for the new
-	  `-o allenv' option
-	- updated the description of the `shopt' builtin
-
-builtins/history.def
-	- changed -a option to use maybe_append_history
-
-subst.c
-	- changed the substring/subarray code to make negative offsets
-	  count backwards from the end of the string or array
+	- if execute_for_command finds that the iteration variable is readonly,
+	  decrement loop_level before returning
 
-builtins/bashgetopt.c
-	- added option modifiers `;' (argument is optional) and `#'
-	  (argument is optional, but if present must be numeric)
+builtins/break.def
+	- if the break count is <= 0, display an error message and break out
+	  of all loops
 
-builtins/hash.def
-	- converted to use builtin_getopt
-	- broke code out into a new function: add_hashed_command
+builtins/command.def
+	- if PATH is unset, and we're using command -p, we don't want PATH
+	  to be set to the empty string when `command' completes
 
-				    2/7
-				    ---
-builtins/getopt.c
-	- changed to save state to avoid relying on `nextchar' staying the
-	  same across calls to sh_getopt [in 1.14.4]
-	- added a function to restore `nextchar' from this saved state
-	  [in 1.14.4]
-	- removed some dead code
+builtins/common.c
+	- POSIX.2 says `kill -l signum' prints the signal name without the
+	  leading `SIG' prefix, so check for this_shell_builtin == kill_builtin
+	  in display_signal_list
 
 builtins/getopts.def
-	- call function to restore sh_getopt state when parsing explicitly
-	  supplied arguments rather than the positional parameters [in 1.14.4]
+	- when invoked without any arguments, `getopts' now prints a usage
+	  message
 
-lib/readline/vi_mode.c
-	- fixed an off-by-one error in _rl_vi_done_inserting that put the
-	  \0 in vi_insert_buf at offset `len' instead of `len - 1'
-
-lib/readline/complete.c
-	- print_filename now filters out control characters and displays
-	  them in printable format, rather than relying on the tty driver
-	  to do the right thing [in 1.14.4]
-
-				    2/8
+				    3/3
 				    ---
-lib/posixheaders/stdc.h
-	- test for each special keyword being defined individually, rather
-	  than just testing on `const' [in 1.14.4]
-
-lib/readline/util.c
-	- new function _rl_abort_internal; rl_abort just calls this
-
-nojobs.c
-	- include error.h for extern function definitions [in 1.14.4]
-
-builtins/ulimit.def
-	- include <unistd.h> if HAVE_UNISTD_H is defined
-	- use HAVE_LIMITS_H to decide whether or not to include <limits.h>
-
-variables.c, subst.c, builtins/set.def
-	- removed special handling of `noclobber' variable
-
-builtins/set.def
-	- new function set_shellopts to set up the $SHELLOPTS variable
-	  based on the values of the set -o options; `set' calls this
-	  function whenever one of the options is changed
-	- new function parse_shellopts to take the value of SHELLOPTS and
-	  turn on each option found therein
-	- new function initialize_shell_options to parse any inherited
-	  value of $SHELLOPTS and set up $SHELLOPTS
-
-shell.c
-	- call initialize_shell_options at the end of shell_initialize()
-
-execute_cmd.c, general.c
-	- moved extract_colon_unit from execute_cmd.c to general.c
-
-execute_cmd.h, general.h
-	- moved extern declaration of extract_colon_unit from
-	  execute_cmd.h to general.h
-
-documentation/{bash.1,features.texi}
-	- documented SHELLOPTS
-
-				   2/18
-				   ----
-builtins/{shopt,pushd,jobs,umask}.def
-	- removed the `longjobs', `longdirs', `pushd_home' and `symbolic_umask'
-	  options
-
-lib/readline/{histexpand,histsearch,histfile}.c, lib/readline/histlib.h
-	- new files, split off from old history library history.c
-
-lib/readline/history.c, lib/readline/history.h
-	- new function clear_history() to clear the history list
-
-lib/readline/doc/hstech.texinfo
-	- documented clear_history ()
-
-builtins/pushd.def
-	- new function clear_directory_stack() to delete all elements of
-	  the dir stack
-	- new -c option for `dirs' to clear directory stack
-	- new function get_dirstack_index for dirs -N and dirs +N to use
-	- new function get_dirstack_element(i, dir) for use by other parts of
-	  the shell that want the functionality of dirs -N and dirs +N
-	  (dir is 1 for dirs +N, -1 for dirs -N)
-	- new function set_dirstack_element(i, dir, val) for use by other
-	  parts of the shell to change values in the dirstack (dir is -1
-	  for dirs -N, +1 for dirs +N).  Used when assigning to $DIRSTACK.
-
-builtins/history.def
-	- new -c option to clear the history list
-
-documentation/bash.1, documentation/features.texi
-	- removed no-longer-valid shopt options
-	- documented new dirs -c option
-	- documented new history -c option
-	- documented new \T and \H prompt escape sequences
-	- documented new $DIRSTACK dynamic array variable
-	- documented the new expand-glob and list-glob readline commands
-
-variables.c
-	- set PS4 to its default value of `+ ', so that unsetting it will
-	  disable the tracing characters [in 1.14.4]
-	- new framework for dynamic array variables -- each place an
-	  array index is assigned to, a test is made.  if a dynamic
-	  assignment function exists, it is called with args `self',
-	  the index being assigned to, and the new value
-	- new variable `$DIRSTACK', a dynamic array variable that holds
-	  the current contents of the directory stack.  You can even
-	  change the stack by assigning to thie variable
-
-shell.c
-	- changed indirection_level_string to return the null string if
-	  $PS4 is unset or null [in 1.14.4]
-
-parse.y
-	- new \H prompt escape for hostname up to first `.'; changed \h
-	  to return full hostname (like \w/\W)
-	- new \T prompt escape for 12-hour time
-
-bashline.c
-	- new readline functions to expand glob patterns and insert or
-	  list the expansions (special completion functions)
-
-
-				   2/20
-				   ----
-builtins/pushd.def
-	- new code for `pushd -n' and `popd -n' to inhibit cd when
-	  adding or removing directories from the stack
-	- broke functionality off into separate functions for use by the
-	  $DIRSTACK manipulation functions
-
-				   2/21
-				   ----
-lib/readline/kill.c
-	- new functions to copy words to the kill ring, backward or
-	  forward
-
-lib/readline/readline.h
-	- extern declarations for rl_copy_{backward,forward}_word
+builtins/common.c
+	- add a second argument to get_numeric_arg: if non-zero, the shell
+	  exits on a bad argument; if not, the shell jumps to top_level
+	  with a DISCARD argument, which aborts the current command
 
+builtins/{break,exit,history,return,shift}.def
+	- change calls to get_numeric_argument
 
 lib/readline/funmap.c
-	- new bindable readline commands: copy-backward-word and
-	  copy-forward-word to copy portions of the line to the kill
-	  ring without deleting them
-
-documentation/{bash.1,readline.3}, lib/readline/doc/hsuser.texinfo
-	- documented the new copy-backward-word and copy-forward-word
-	  readline commands
-
-mailcheck.c, config.h
-	- the declaration of DEFAULT_MAIL_PATH is now only in config.h
-	  and used by mailcheck.c and variables.c.  There is no need to
-	  use DEFAULT_MAIL_PATH_LEN; `sizeof' does the job [in 1.14.4]
-
-				   2/22
-				   ----
-support/mksysdefs
-	- changes to fix `RELEASE' if it ends up being set to the empty string
-	  [in 1.14.4]
-	- check for amdahl UTS [in 1.14.4]
-	- check for SGI Irix version 6.x [in 1.14.4]
-
-machines.h
-	- new entry for Amdahl UTS [in 1.14.4]
-	- changes to the SGI entry for Irix 6.x [in 1.14.4]
-
-				   2/23
-				   ----
-machines.h
-	- changes for BSD/OS 2.0 (M_OS now set to `BSD_OS') [in 1.14.4]
-
-support/mksysdefs
-	- change to recognize BSD/OS 2.0 and set SYSDEF to BSDI2 [in 1.14.4]
-
-siglist.h
-	- don't declare sys_siglist on BSD/OS 2.0 [in 1.14.4]
-
-				   2/24
-				   ----
-parse.y
-	- if reset_parser is called while the prompt command is being
-	  executed because of a syntax error in $PROMPT_COMMAND, an
-	  infinite loop results.  set token_to_read back to 0 at the
-	  end of execute_prompt_command to stop the looping [in 1.14.4]
-
-				   2/25
-				   ----
-mailcheck.h
-	- new header file for mail-checking and related definitions
-	- declaration of DEFAULT_MAIL_PATH is now here
-
-mailcheck.c
-	- new function, make_default_mailpath, which constructs a default
-	  $MAILPATH string from DEFAULT_MAIL_PATH [in 1.14.4]
-	- change remember_mail_dates to call make_default_mailpath
-	  [in 1.14.4]
-
-externs.h
-	- moved function declarations for functions in mailcheck.c to
-	  mailcheck.h
-
-variables.c
-	- change initialize_shell_variables to call make_default_mailpath
-	  [in 1.14.4]
-	- removed some unneeded variables in initialize_shell_variables
-
-				   2/26
-				   ----
-lib/readline/callback.c
-	- new file with readline callback function interface
-
-lib/readline/rlconf.h
-	- new READLINE_CALLBACKS define for the readline callback code to
-	  be compiled in and available
-
-lib/readline/readline.c
-	- broke readline_internal into three functions: readline_internal_setup,
-	  readline_internal_charloop, and readline_internal_teardown
-	- changes for READLINE_CALLBACKS
+	- add `dump-macros' to list of bindable names
 
 lib/readline/readline.h
-	- extern declarations for the readline callback code
-
-cpp-Makefile
-	- added lib/readline/callback.c to the list of readline files
-
-lib/readline/input.c
-	- added a layer of indirection to allow the user to specify the
-	  function that reads a character from rl_instream.  The variable
-	  name is rl_getc_function, set by default to rl_getc
-
-				   2/27
-				   ----
-lib/readline/display.c
-	- added a variable rl_redisplay_function to allow an application-
-	  specified redisplay function, for those apps that want to control
-	  redisplay
+	- added extern declaration for rl_prompt (it was apparently missing)
 
 lib/readline/readline.c
-	- extern declarations for rl_getc_function and rl_redisplay_function
-
-lib/readline/{readline,display,isearch,search,parens}.c
-	- changed to call through rl_redisplay_function rather than
-	  rl_redisplay directly
-
-parse.y
-	- try to avoid some work in reset_readline_prompt and prompt_again
-	  if the prompt is the empty or null string [in 1.14.4]
+	- new internal function, _rl_init_line_state, which sets rl_point
+	  and rl_end to 0, sets the_line to point to _rl_line_buffer, and
+	  clears the line
 
-lib/readline/readline.c
-	- call rl_expand_prompt unconditionally [in 1.14.4]
+lib/readline/callback.c
+	- if a user's callback function does not clear the line, clear it
+	  for him by calling _rl_init_line_state
 
-lib/readline/display.c
-	- short-circuit out of rl_expand_prompt if the prompt string is null
-	  after clearing out the saved local prompt values.  This allows
-	  $PS2 to be set to "" [in 1.14.4]
+				    3/4
+				    ---
+alias.c
+	- made the readline support functions #ifdef READLINE, so they're
+	  not compiled into the shell unless readline is
 
-lib/readline/doc/rltech.texinfo
-	- documented the callback code and functions
-	- documented rl_getc_function and rl_redisplay_function
+lib/readline/bind.c
+	- new function _rl_untranslate_macro_value, to expand meta-prefixes
+	  and other special characters in a macro value for printing by
+	  _rl_macro_dumper_internal
+	- call _rl_untranslate_macro_value in _rl_macro_dumper_internal to
+	  get a printable version of the macro value
 
-lib/readline/{callback,readline,rltty,signals}.c
-	- added layer of indirction for terminal prep and deprep, with
-	  rl_term_prep_function and rl_term_deprep_function.  These are
-	  set by default to rl_prep_terminal and rl_deprep_terminal,
-	  respectively
+lib/readline/readline.c
+	- new variable, rl_dispatching, set to 1 when we call a function
+	  from _rl_dispatch
 
 lib/readline/readline.h
-	- make rl_term_prep_function and rl_term_deprep_function available
-	  to callers
+	- extern declaration for rl_dispatching
 
-				    3/1
-				    ---
 lib/readline/complete.c
-	- broke the (long, complicated) rl_complete_internal code into a
-	  number of separate functions:
-
-		find_completion_word
-		gen_completion_matches
-		remove_duplicate_matches
-		display_matches
-		insert_text
-		insert_match
-		append_to_match
-		insert_all_matches
-
-	- made some efficiency improvments to filename_completion_function
-	- the completion ignore function is now called no matter what type
-	  of completion is being performed (as it should have been all along)
-
-lib/readline/rldefs.h
-	- use #defines for the possible values ORed into `found_quote' by
-	  the completion code
-
-				    3/2
-				    ---
-make_cmd.c, general.c
-	- moved make_word_array from make_cmd.c to general.c, renamed to
-	  word_list_to_argv, extended it to optionally not malloc all of
-	  the strings and to reserve space at the start of the array
-
-execute_cmd.c
-	- changed to use word_list_to_argv, not mallocing space for the
-	  strings
-
-builtins/common.c, builtins/common.h
-	- new function make_builtin_argv, which uses word_list_to_argv
-	  and reserves 1 slot at the beginning for the command name
+	- make sure S_ISCHR and S_ISBLK are defined before using them
 
-builtins/exec.def
-	- changed to use word_list_to_argv
+lib/readline/terminal.c
+	- add a new #define NEED_EXTERN_PC.  Define this if the termcap
+	  or curses libraries need `extern' before declarations of PC,
+	  BC, and UP
 
-builtins/{getopts,test}.def
-	- changed to use make_builtin_argv
+lib/readline/{readline,terminal,histfile}.c
+	- changes to compile on OS/2 with OS/2 `EMX' port of gcc, originally
+	  sent by ilya@math.ohio-state.edu
 
-subst.c
-	- new function match_pattern_char, to see if the first char of
-	  a string has a chance to match a given pattern (test against
-	  the first char of the pattern); used by match_pattern in the
-	  MATCH_ANY case
+builtins/set.def
+	- fixed a bug in set_shellopts that caused the shell to crash if
+	  there were no shell options set
 
-				    3/3
+				    3/5
 				    ---
-jobs.c
-	- renamed flush_child to sigchld to capture the functionality better
+configure.in,Makefile.in
+	- choose run-all or run-minimal as the test script based on whether
+	  the --enable-minimal-config option was given to configure
 
-array.c
-	- new function array_pat_subst, to do pattern substitution on each
-	  element in an array
-
-array.h
-	- extern declaration for array_pat_subst
-
-subst.c
-	- new function pat_subst to do pattern matching and substitution on
-	  a string
-	- new function parameter_brace_pat_subst to implement
-	  ${v/[/]pat[/sub]}; calls pat_subst for simple vars,
-	  pos_params_pat_subst to do substitution on the positional params,
-	  and array_pat_subst for things like ${v[@]/p/r}
-
-subst.h
-	- extern declaration for pat_subst so array.c can find it
+builtins/setattr.def
+	- fixed `export -p' and `readonly -p' so that they output `export'
+	  or `readonly' when in POSIX mode, as POSIX.2 specifies
 
 				    3/6
 				    ---
-parse.y
-	- <>filename now dups filename to file descriptor 0 for both input
-	  and output even when not in posix.2 mode
+builtins/setattr.def
+	- make `readonly -a var=(...)' work just like `declare -ar var=(...)',
+	  since the two logically mean the same
+	- `readonly -f' and `export -f' don't print the function definition
+	  for each readonly or exported function, respectively, when in
+	  POSIX mode
 
-bashline.c
-	- add the globbing characters to the list of characters that need
-	  to be quoted by filename completion
+jobs.c, nojobs.c
+	- don't report status for processes killed by SIGPIPE if
+	  DONT_REPORT_SIGPIPE is defined
 
-jobs.h
-	- new convenience macros:
-		RUNNING, STOPPED, DEADJOB - to test a job's state
-		IS_FOREGROUND, IS_NOTIFIED, IS_JOBCONTROL - flags
+config.h.top
+	- added a commented-out define for DONT_REPORT_SIGPIPE
 
-jobs.c, builtins/{kill,fg_bg}.def
-	- changed to use new jobs.h macros
+execute_cmd.c
+	- `time' can now be used to time background pipelines, and reports
+	  the timing statistics when the pipeline completes
 
-				    3/7
-				    ---
-array.c, array.h
-	- new function dup_array_subrange(a, s, e) to make a new array
-	  out of the elements of array A between S and E, inclusive
-	- add `quoted' parameters to array_subrange and array_pat_subst
-	  to preserve proper quoting of elements when expanding things
-	  like "${av[@]/xx/yy}"
-	- new function array_quote to quote the members of an array like
-	  the functions in subst.c
+[bash-2.01-alpha1 frozen]
 
+				   3/12
+				   ----
 subst.c
-	- pass the quoted flag to array_subrange and array_pat_subst
-	- quote_string is no longer static
-
-subst.h
-	- extern declaration for quote_string
-
-builtins/hash.def
-	- added a -p pathname option to specify a pathname for the command
-	  name to be hashed.  With -p, no path search is performed.
-
-lib/readline/histexpand.c
-	- broke history_tokenize off into history_tokenize_internal and
-	  added two arguments: a character index and a word index.  If
-	  the char index is >= 0, the word index will be modified to point
-	  into the returned array of strings to the word surrounding that
-	  particular character index
-	- new function history_find_word(line, ind) to return the word 
-	  containing the character at index IND in LINE
-	- new variable search_match, found by history_find_word, to hold the
-	  word last matched by a !?string? search
-	- corrected a problem with the `%' modifier: it should insert the
-	  word last matched, not the last search string
-
-				    3/8
-				    ---
-cpp-Makefile
-	- changed INSTALL_PROGRAM and INSTALL_DATA to use support/install.sh
-	- changed `install' target to not explicitly save the old version of
-	  bash in `bash.old'; let install take care of it
+	- move the parent end of the pipe file descriptor used for process
+	  substitution to a high, unused file descriptor to avoid clashes
+	  with redirections performed explicitly by a script
 
-shell.c
-	- new static variable `running_under_emacs', set to 1 if the
-	  variable `EMACS' is in the startup environment, and to 2 if
-	  we're running under the `eterm' terminal emulator
-	- send an escape sequence to eterm if running_under_emacs is 2
-	  after executing any $PROMPT_COMMAND to tell it the current
-	  directory
-
-				    3/9
-				    ---
-builtins/ulimit.def
-	- made getting -u work for systems that have a MAXUPRC define.
-	  it still cannot be set without RLIMIT_NPROC
-
-shell.c
-	- include trap.h for sig definitions
-
-builtins/common.h, subst.h, general.h, externs.h
-	- more extern function declarations
-
-bashhist.c
-	- new function, bash_history_disable(), to turn off history and
-	  history expansion
-	- bash_history_disable now sets history_expansion_inhibited
-
-bashline.h
-	- new file with extern declarations from bashline.c
+configure.in
+	- added a `--with-curses' argument so curses can be forcibly chosen
+	  over libtermcap (some Unix versions ship lousy termcap databases)
 
-parse.y, bashhist.c, builtins/bind.def, subst.c, sig.c
-	- include bashline.h
+support/mkconffiles
+	- new script to create _distribution and _patchlevel from values
+	  contained in `configure'
 
-				   3/10
-				   ----
-lib/glob/glob.h
-	- new file with extern declarations for local glob library
-
-shell.c
-	- include glob.h, mailcheck.h
-
-cpp-Makefile
-	- changes to dependencies due to mailcheck.h, bashline.h, glob.h
+doc/bashref.texi
+	- updated installation instructions
 
 				   3/13
 				   ----
-execute_cmd.c
-	- new function execute_connection to execute commands of type
-	  cm_connection
-	- new function execute_pipeline, to execute pipelines
-
-lib/readline/rltty.c
-	- add a `tty' argument to the set_winsize function/macro, the
-	  non-shell version needs it
+general.c
+	- if `getgroups' returns 0, make sure we add the primary group id
+	  as GROUPS[0].
+	- if we have getgroups, and the primary gid is not in the array
+	  getgroups returns, add it as group_array[0] and shuffle everything
+	  up one element.  This ensures that current_user.gid == group_array[0]
+	  all the time
+
+tests/builtins.tests
+	- changes to avoid stray variables in environment when the shell
+	  version of printenv is used with bash as /bin/sh, running the
+	  `exec -c' tests.
 
 				   3/14
 				   ----
-parse.y
-	- added \a (bell) and \e (escape) prompt string escape sequences
-
-				   3/16
-				   ----
-lib/readline/display.c
-	- fixed bug in update_line that caused stray characters to be left on
-	  lines after the first if characters are deleted so that the first
-	  line becomes exactly as long as the screen width [in 1.14.4]
-
-lib/readline/doc-support
-	- brought in the GNU getopt because texindex needs it
+builtins/cd.def
+	- spelling correction is no longer enabled by default
 
-documentation/Makefile
-	- adopted the BSD convention of a suffix of `.0' for formatted
-	  manual pages
+support/bashbug.sh
+	- if the shell's release status is alpha or beta, offer the option
+	  of sending the bug report to the bash-testers mailing list as
+	  well as to chet
 
 				   3/17
 				   ----
-support/inpath
-	- a script to find out if a particular command name appears in
-	  a directory in $PATH [in 1.14.4]
-
-support/mksysdefs
-	- changed to use inpath to find `ranlib' [in 1.14.4]
-
-				   3/18
-				   ----
-bashline.c
-	- include bashline.h
-	- removed the DYNAMIC_HISTORY_COMPLETION define; that code is now
-	  included unconditionally
-	- renamed ETCHOSTS to DEFAULT_HOSTS_FILE; moved definition to
-	  bashline.h
-	- moved definition of BRACE_COMPLETION to config.h
-
-config.h
-	- conditional definition of BRACE_COMPLETION is now here
-
-pathnames.h
-	- new file with defines that are absolute pathnames
-
-shell.h
-	- include pathnames.h
-
-mailcheck.h
-	- def of DEFAULT_MAIL_PATH now moved to pathnames.h
-
-shell.c
-	- def of SYS_PROFILE moved to pathnames.h
-
-alias.h
-	- removed a bunch of extra definitions
-
-lib/readline/{vi_mode.c,bind.c}
-	- removed superfluous STATIC_MALLOC code
-
-variables.c
-	- removed SHADOWED_ENV code
-
-				   3/20
-				   ----
-machines.h
-	- don't define USG or USGr3 for linux machines [in 1.14.4]
-	- change REVERSED_SETVBUF_ARGS to SETVBUF_REVERSED for autoconf
-	  compatibility
-
-shell.c
-	- change REVERSED_SETVBUF_ARGS to SETVBUF_REVERSED for autoconf
-	  compatibility
-
-support/mksysdefs, maxpath.h
-	- change HAVE_SYS_PARAM to HAVE_SYS_PARAM_H for autoconf
-	  compatibility
-
-support/mksysdefs, jobs.h
-	- use HAVE_SYS_WAIT_H
-
-machines.h, error.c
-	- change HAVE_VFPRINTF to HAVE_VPRINTF for autoconf compatibility
-
-test.c
-	- replace UID_T and GID_T with GETGROUPS_T, the size of the elements
-	  of the array returned by getgroups
-
-general.c
-	- define HAVE_KILLPG if killpg() is present; compile in a replacement
-	  killpg if not
-
-machines.h, jobs.c
-	- change BSD_GETPGRP to HAVE_BSD_PGRP
+configure.in
+	- configure --without-gnu-malloc by default on *-sgi-irix6* because
+	  their code needs 8-byte alignment
 
-sig.c
-	- don't call initialize_siglist if HAVE_SYS_SIGLIST is defined, no
-	  longer use INITIALIZE_SIGLIST
+support/bashbug.sh
+	- ``' needs to be quoted with a backslash in double-quoted strings
 
-machines.h, lib/malloc/malloc.c
-	- change NO_SBRK_DECL to SBRK_DECLARED
+aclocal.m4
+	- slight changes to the strcoll test, since AIX 4.2 returns -1, 0, or
+	  1 from strcmp(3) but a numeric collation order difference from
+	  strcoll(3)
 
-				   3/22
+				   3/18
 				   ----
-machines.h
-	- BSD/OS 2.0 does not need INT_GROUPS_ARRAY in SYSDEP_CFLAGS
-
-lib/readline/rldefs.h
-	- don't check `Linux'; just include <termcap.h> if HAVE_TERMCAP_H
-	  is defined
-
-builtins/command.def
-	- changed get_standard_path to use HAVE_CONFSTR 
+command.h
+	- new redirection error code:  HEREDOC_REDIRECT
 
 execute_cmd.c
-	- execute_simple_command no longer uses alloca
-	- don't check RISC6000 anymore when deciding whether to use
-	  #pragma alloca
-
-execute_cmd.c, shell.c
-	- check for ultrix instead of Ultrix when calling alloca(0)
-
-jobs.h
-	- check HAVE_UNISTD_H to see whether to declare fork, getpid,
-	  and getpgrp
+	- return HEREDOC_REDIRECT from do_redirection_internal when
+	  here_document_to_fd cannot create the temp file for a here document
+	- changed redirection_error to print a meaningful message when
+	  here document temp file creation fails (HEREDOC_REDIRECT)
 
-builtins/common.c
-	- include <signal.h> for NSIG
-
-sig.h
-	- don't define SIGABRT as SIGIOT unless SIGIOT is defined
+				   3/19
+				   ----
+subst.c
+	- changes to match_pattern_char: return 1 if the first char of the
+	  pattern is `?' only if the string is non-null; just return 1 if
+	  the first char of the pattern is `[' and the string is non-empty
+	  rather than try to re-implement the brace matching code from fnmatch
 
-jobs.c
-	- check for ultrix instead of Ultrix
+lib/glob/fnmatch.c
+	- some changes from glibc-2.0.1 posix/fnmatch.c
 
-				   3/23
+				   3/21
 				   ----
-general.c, general.h
-	- new function check_dev_tty, which makes sure we can open
-	  /dev/tty
+variables.c
+	- only do the initialization of `ignoreeof' if the shell is
+	  interactive
+	- reset values of $SHLVL > 1000 to 1 in adjust_shell_level, and
+	  don't call itos, since we don't need its generality
+	- new function, initialize_shell_level, just calls adjust_shell_level
+	  with argument of 1.  If $SHLVL is unset, adjust_shell_level will
+	  deal with it correctly
+	- change initialize_shell_variables to not malloc a copy of each
+	  environment variable, just keep two pointers into the env string:
+	  one for the name, one for the value
+	- broke the code that computes the value of $BASH out into a separate
+	  function: get_bash_name
+	- get_bash_name special-cases shell_name with a `./' prefix when
+	  initializing $BASH
+	- new function: set_home_var, sets $HOME to current_user.home_dir if
+	  it's not already set, calling get_current_user_info if
+	  current_user.home_dir is NULL
+	- new function: set_shell_var, sets $SHELL to current_user.shell if
+	  it's not already set, calling get_current_user_info if
+	  current_user.shell is NULL
+	- changed places that reference information in current_user to check
+	  for NULL values of the member they're interested in and call
+	  get_current_user_info if necessary
 
 shell.c
-	- call check_dev_tty instead of having the code inline
-
-posixstat.h
-	- removed references to isc386
-
-general.h
-	- code to define one of TERMIOS_TTY_DRIVER, TERMIO_TTY_DRIVER,
-	  or NEW_TTY_DRIVER
+	- moved the code that sets up the information in current_user that
+	  comes from the password file into a new function,
+	  get_current_user_info
+	- shell_initialize calls get_current_user_info only if the shell is
+	  interactive
 
-jobs.c, nojobs.c
-	- use the general.h code to define the tty driver types
+externs.h
+	- new extern declaration for get_current_user_info(), so variables.c
+	  can use it
 
 				   3/24
 				   ----
-builtins/declare.def
-	- new -p option to display variables and their values and attributes
-	  `declare -p xxx' displays attribs and value of var `xxx'
-
-builtins/setattr.def
-	- new function to display the attributes and value of a particular
-	  variable; used by set_or_show_attributes
-	- new function to look up a variable by name and show that name's
-	  attributes and value
-
-builtins/common.c
-	- declarations for new functions in setattr.def
-
-variables.c
-	- don't set a default value for MAILPATH in initialize_shell_variables;
-	  let remember_mail_dates take care of it [in 1.14.4]
-
-mailcheck.c
-	- fixed an off-by-one bug in make_default_mailpath [in 1.14.4]
+lib/tilde/tilde.c
+	- if SHELL is defined, user the current_user struct info to find
+	  the user's home directory rather than calling getpwuid
 
-				   3/28
+				   3/25
 				   ----
-[changes for autoconf-generated config files]
-
-lib/malloc/malloc.c
-	- use HAVE_GETPAGESIZE, HAVE_BSD_SIGNALS, HAVE_POSIX_SIGNALS
-
-trap.c
-	- remove tests for USG and USGr4
-	- test on HAVE_POSIX_SIGNALS rather than _POSIX_VERSION
-	- use MUST_REINSTALL_SIGHANDLERS define to decide whether trap_handler
-	  should call signal again
-
-getcwd.c
-	- redid the directory includes for autoconf compatibility
-	- use HAVE_LSTAT instead of testing for S_ISLNK
-	- use STRUCT_DIRENT_HAS_D_INO
-	- include memalloc.h for alloca define
-
-general.h
-	- cleaned up strchr, strrchr definitions
-	- removed tests against USG
-	- use HAVE_MEMMOVE instead of MEMMOVE_MISSING
-
-jobs.c
-	- use HAVE_WAIT3, MUST_REINSTALL_SIGHANDLERS, GETPGRP_VOID
-
-mailcheck.h
-	- removed definition of DEFAULT_MAIL_DIRECTORY; now set by
-	  autoconf in config.h
-
-shell.c
-	- remove checks on USG, just check for HAVE_GETPW_DECLS
-	- use C_ALLOCA define
-	- redid the isnetconn() code using HAVE_SYS_SOCKET_H,
-	  HAVE_GETPEERNAME, SVR4 and SVR4_2
-
-general.c
-	- use HAVE_KILLPG, HAVE_RESTARTABLE_SYSCALLS, HAVE_UNAME,
-	  ULIMIT_MAXFDS, HAVE_TIMEVAL, HAVE_TIMES
-
-lib/readline/rldefs.h
-	- redid the tty driver definitions using HAVE_TERMIOS_H, etc.
-	- don't define anything having to do with signal type
-
-lib/glob/glob.c
-	- redid the DIRENT defines and includes
-	- redid other includes to remove dependencies on USG and system
-	  type (e.g., NeXT)
-
-builtins/times.def
-	- changed to use HAVE_GETRUSAGE, HAVE_TIMEVAL, HAVE_TIMES, and
-	  the autoconf way to include <sys/time.h> and <time.h>
-
-builtins/ulimit.def
-	- changed to use HAVE_GETRLIMIT
-	- removed test of USG being defined
-
-siglist.h
-	- changed to use SYS_SIGLIST_DECLARED, HAVE_UNDER_SYS_SIGLIST, and
-	  HAVE_STRSIGNAL
-
-print_cmd.c
-	- use PRINTF_DECLARED
-
-builtins/command.def
-	- use HAVE_CONFSTR along with _CS_PATH to get the standard path
-
-execute_cmd.c
-	- only compile in execute_shell_script if HAVE_HASH_BANG_EXEC is not
-	  defined
-
 nojobs.c
-	- use HAVE_SIGINTERRUPT, HAVE_KILLPG, HAVE_POSIX_SIGNALS,
-	  MUST_REINSTALL_SIGHANDLERS, HAVE_WAITPID, *_TTY_DRIVER defines
+	- don't try to open /dev/tty when getting or setting the tty state
+	  and window size; use shell_tty instead
+	- initialize shell_tty to standard error in initialize_job_control
+	- only fetch the tty state initially if the shell is interactive
 
-test.c
-	- set up a new `getmaxgroups' define, moving the code out of inline
-	- eliminate use of GETGROUPS_T
+general.c
+	- open /dev/tty with the O_NONBLOCK flag
 
 variables.c
-	- use CAN_REDEFINE_GETENV
-
-sig.c, sig.h
-	- use HAVE_POSIX_SIGNALS, MUST_REINSTALL_SIGHANDLERS
-
-pathnames.h
-	- removed the default mail directory defines; now set by autoconf
-
-oslib.c
-	- new file, functions from general.c that are unix-version variable
-
-general.h
-	- slightly changed function declarations for use by oslib.c
-
-support/bashbug.sh
-	- changed @xxx@ to !xxx! for sed substitutions to avoid conflicts
-	  with autoconf substitutions in the Makefile
-	- added MACHTYPE variable
-
-mailcheck.c
-	- changed DEFAULT_MAIL_PATH to DEFAULT_MAIL_DIRECTORY, since that
-	  more clearly defines its function
-
-lib/readline/rltty.h
-	- new file, to include the correct tty driver #include file
-
-lib/readline/rltty.c
-	- include rltty.h
-
-lib/malloc/malloc.c
-	- make systems with Posix signals block all signals while malloc
-	  is executing
-
-				   3/29
-				   ----
-input.c
-	- use off_t as the type of a seek offset, rather that int or long
-
-variables.c, oslib.c
-	- moved `getenv' from variables.c to oslib.c
-
-				   3/31
-				   ----
-sig.c, sig.h
-	- new function, jump_to_top_level, which just calls longjmp with
-	  top_level as an argument -- here to isolate calls to longjmp
-
-subst.c
-	- replaced calls to longjmp with jump_to_top_level
-
-bashjmp.h
-	- new file with setjmp/longjmp defines and declarations
-
-shell.h, nojobs.c
-	- include bashjmp.h in place of setjmp.h
-
-shell.c, sig.c, execute_cmd.c, unwind_prot.h, expr.c,
-builtins/{return,source}.def
-	- change to use new defs in bashjmp.h
-
-subst.c
-	- make sure PAT and REP in pattern substitution are run through
-	  expand_string_unsplit
+	- changed all_vars so that it sorts its output all the time, not
+	  just when in POSIX mode.  This means that the output of `set'
+	  and `export' will be sorted
 
 builtins/set.def
-	- don't try to blindly dereference the value returned by
-	  find_flags in set_shellopts
-
-				    4/2
-				    ---
-aclocal.m4, config.h.in, config.h.top, config.h.bot, configure.in,
-support/install.sh, support/config.guess, support/config.sub
-	- new files for (now official) autoconf-based configuration
-
-Makefile.in, builtins/Makefile.in,
-lib/{doc-support,malloc,glob,termcap,tilde,readline}/Makefile.in
-	- new Makefiles for autoconf
-
-[additionally, all source files now include config.h]
-
-				    4/3
-				    ---
-array.c, array.h
-	- index_t --> arrayind_t, because some systems define index_t in
-	  <sys/types.h>
-
-				    4/4
-				    ---
-aclocal.m4
-	- fix test for broken dup2
-
-lib/*/Makefile.in
-	- make all object files depend on $(BUILD_DIR)/config.h
-
-				    4/5
-				    ---
-array.c
-	- fixed problems in empty_array: need to reset max_index
-	  and max_size, and remove the links in the element chain
-	  after freeing them
-
-jobs.c
-	- new functions to save an array of status values for each
-	  foreground job that exits (degenerate case is to have a
-	  single-process job exit and have a 1-element array)
-	- new function to set a shell array variable `PIPESTATUS'
-	  which holds the status values from each member of the
-	  last-executed pipeline that spawned children
-
-documentation/bash.1
-	- documented $PIPESTATUS
-
-				    4/6
-				    ---
-configure.in
-	- added new --with-afs argument to #define AFS for execute_cmd.c
-
-				   4/10
-				   ----
-builtins/ulimit.def
-	- use sysconf(_SC_CHILD_MAX) to find the maximum number of child
-	  proceeses per user if HAVE_SYSCONF and _SC_CHILD_MAX are both
-	  defined and RLIMIT_NPROC is not
-
-				   4/12
-				   ----
-lib/readline/undo.c
-	- new variable, local to library, to keep track of the number of
-	  `open' undo groups (UNDO_BEGIN without corresponding UNDO_END)
-	- new function _rl_fix_last_undo_of_type to modify start and end
-	  bounds of last undo record of a specified type
-
-lib/readline/vi_mode.c
-	- _rl_vi_done_inserting now calls rl_end_undo_group if the count
-	  of unclosed groups is > 0
-	- fixed rl_vi_change_to to save an undo record when redoing and
-	  to fix the buffer corruption when doing `u' undo after a `.'
-	  redo of `C'
-
-				   4/13
-				   ----
-unwind_prot.h
-	- changed to use a union { char *s; int i; } when unwind-protecting
-	  integers to force correct alignment on machines where ints and
-	  pointers differ in size
+	- in initialize_shell_options, only call parse_shellopts if we
+	  inherited $SHELLOPTS from the environment
+	- make sure we call parse_shellopts from initialize_shell_options
+	  with a copy of the value of SHELLOPTS, in case one of the functions
+	  called while setting one of the variables modifies $SHELLOPTS
 
 lib/readline/readline.c
-	- new variable Keymap rl_executing_keymap, which is set to the
-	  keymap the last function was invoked out of
+	- make sure that digit arguments don't change the state of
+	  rl_last_func
 
-lib/readline/bind.c
-	- new variable Keymap rl_binding_keymap, which is set to the last
-	  keymap a function and key sequence were bound in
+support/printenv.c
+	- new file, printenv(1) clone, used to avoid environment variables
+	  that might be set automatically when using printenv.sh
 
-				   4/17
+lib/tilde/tilde.c
+	- if SHELL is defined, don't call getenv to get the value of $HOME,
+	  call get_string_value () directly
+
+				   3/26
 				   ----
-general.c
-	- removed xmalloc, xrealloc, xfree to xmalloc.c
+lib/readline/histexpand.c
+	- abstracted the `#ifdef SHELL' stuff that checked for special cases
+	  that should not be history expanded ([!...], ${!...}) into a call
+	  to a function that is the value of the
+	  new history_inhibit_expansion_function variable
 
-lib/malloc/Makefile.in
-	- MALLOC and ALLOCA are both set by autoconf
-	- need to provide some empty stub file to make up the library
-	  in case neither malloc.c and alloca.c are compiled into the
-	  shell
+lib/readline/history.h
+	- extern declaration for history_inhibit_expansion_function
 
-lib/malloc/gmalloc.c
-	- new file, GNU libc malloc code
+bashhist.c
+	- new function, bash_history_inhibit_expansion, which checks for
+	  the special cases in which history expansion should be inhibited
+	- changes to the various history initialization functions to
+	  set history_inhibit_expansion_function
 
-Makefile.in
-	- new source file, xmalloc.c, new object file, xmalloc.o
+lib/readline/doc/hstech.texinfo
+	- documented history_inhibit_expansion_function
 
-general.h
-	- changed type of size argument to xmalloc, xrealloc to `size_t'
+lib/readline/shell.c
+	- new file, containing versions of the functions that are provided
+	  by bash when readline is linked as part of bash
+	- new function: get_env_value().  If SHELL is defined, this calls
+	  get_string_value().  If SHELL is not defined, this calls getenv()
 
-configure.in
-	- changes for new argument `--with-glibc-malloc' that includes
-	  gmalloc.o in libmalloc.a
+lib/readline/histexpand.c
+	- moved single_quote() to shell.c
 
-				   4/18
-				   ----
-xmalloc.c
-	- if malloc or realloc return null, report in the error message
-	  how many bytes have been allocated
+lib/readline/util.c
+	- moved savestring() to shell.c
 
-lib/readline/display.c
-	- changed redisplay code to use an array of positions in the visible
-	  and invisible lines at which to break lines instead of simply
-	  calculating based on the screenwidth and number of invisible
-	  characters.  In the future, this will allow newlines embedded in
-	  the lines to display to be handled better
+lib/readline/terminal.c
+	- moved set_lines_and_columns() to shell.c
 
-[Bash-1.14.4 released to net]
+lib/readline/Makefile.in, Makefile.in
+	- added shell.c and shell.o to the appropriate variables that contain
+	  the files comprising the readline and history libraries
 
-				   4/19
-				   ----
 lib/readline/signals.c
-	- changed last call to signal() to call rl_set_sighandler()
-
-				   4/20
-				   ----
-lib/readline/display.c
-	- changed some ascii-specific code to use CTRL_CHAR and UNCTRL
-	- finished up the changes that keep an array of line breaks
-
-shell.h
-	- moved #define constants for parameter pattern substitution here
-	  from subst.c
+	- introduced two new cpp defines:  HANDLE_JOB_SIGNALS and
+	  HANDLE_SIGTERM.  When HANDLE_JOB_SIGNALS is defined, SIGTSTP,
+	  SIGTTIN, and SIGTTOU are caught and handled.  When HANDLE_SIGTERM
+	  is defined, SIGTERM is caught and handled.  These are both
+	  defined automatically if SHELL is not defined
 
-subst.c
-	- changed the pattern substitution functions to take a `flags'
-	  parameter, which subsumes the match type, global replacement
-	  flag, and quoted variables
-	- changed the pattern substitution functions to handle the `#'
-	  and `%' match qualifiers, which anchor the match at the
-	  beginning and end of the string, respectively
+lib/readline/{bind,histfile,nls,readline,terminal}.c
+	- call get_env_value instead of getenv().  This should remove the
+	  dependency on being able to redefine getenv() in oslib.c
 
 shell.c
-	- moved the code that turns off privileged mode into a function
-	  named `disable_priv_mode'
-	- if the shell is running setuid or setgid and `-p' is not
-	  supplied, turn off privileged mode and reset the effective
-	  uid/gid
+	- added a missing argument of -1 to the call to list_minus_o_opts.
+	  Now `bash -o' lists all options, not just random ones depending
+	  on what's on the stack
 
-				   4/21
-				   ----
-subst.c
-	- added a `quoted' paramter to parameter_brace_remove_pattern;
-	  Posix.2 says that the pattern is parsed differently if the
-	  entire expression is double-quoted
-	- `getpattern' now takes a second argument, `quoted'
-	- fixed `getpattern' to correctly handles a pattern spec when
-	  the whole expression is double-quoted.  Posix.2 says that
-	  quote characters inside the pattern spec don't quote any
-	  special pattern chars if the whole thing is double-quoted.
-	  For example, the `*' in "${foo#'*'}" is not quoted, and the
-	  single quotes must appear literally.
-
-Makefile.in
-	- add `documentation' as a dependency of `.made'
-
-				   4/24
-				   ----
-Makefile.in
-	- added `installdirs' target that makes bindir, infodir, mandir,
-	  and man3dir [in 1.14.5 cpp-Makefile]
-
-				   4/25
+				   3/28
 				   ----
-builtins/fc.def
-	- fixed problem with `fc -l' that occurred when fewer than 16 lines
-	  were in the history list [in 1.14.5]
+builtins/ulimit.def
+	- change RLIM_INFINITY to the hard limit only if the hard limit is
+	  greater than the current (soft) limit
 
-				    5/1
-				    ---
-Makefile.in, {builtins,documentation}/Makefile.in, lib/*/Makefile.in
-	- fixed up the various `clean' targets to agree with Gnu coding
-	  standards
+hashlib.c
+	- return immediately from flush_hash_table if the hash table passed
+	  is NULL
 
-				    5/2
+				    4/1
 				    ---
-lib/readline/complete.c
-	- made insert_all_matches correctly quote each of the filenames
-	  inserted into the line, if necessary
+shell.c
+	- remove call to initialize_filename_hashing -- initialize the hash
+	  table the first time a hashed command has to be remembered
 
-config.h.top
-	- surround definitions of DEFAULT_PATH_VALUE and STANDARD_UTILS_PATH
-	  with #ifndef/#endif to allow them to be overridden from the command
-	  line
+hashcmd.c
+	- new file, with functions to perform filename hashing and lookup
+	  taken from builtins/hash.def and builtins/common.c
+	- change to remember_filename -- call initialize_filename_hashing
+	  if hashing_initialized is 0
 
-builtins/set.def
-	- changed set -o allenv to set -o keyword for ksh88 compatibility
+hashcmd.h
+	- new file, mostly from builtins/hashcom.h, with extern function
+	  declarations added
 
-documentation/{bash.1,features.texi}
-	- changed set -o allenv to set -o keyword
+execute_cmd.c, builtins/{hash,type}.def
+	- include hashcmd.h for hash function and type definitions
 
-builtins/setattr.def
-	- added a `nodefs' attribute to set_or_show_attributes,
-	  show_var_attributes, and show_name_attributes to inhibit printing
-	  of definition as well as name
+builtins/{common.{c,h},hash.def}
+	- moved hashing functions and declarations to hashcmd.c/hashcmd.h
 
-builtins/declare.def
-	- changed calls to set_or_show_attributes, and show_var_attributes
-	  accordingly
+Makefile.in, builtins/Makefile.in
+	- changed source and object file definitions and dependencies because
+	  of addition of hashcmd.h and hashcmd.c
 
-				    5/3
-				    ---
-Makefile.in
-	- LIBPATH -> LIBSUBDIR
+builtins/hash.def
+	- return immediately from print_hashed_commands if hashed_commands
+	  is empty, indicating that the hash table has not been initialized
 
-				    5/4
+				    4/2
 				    ---
 lib/readline/bind.c
-	- renamed readline variable meta-flag to be input-meta
-
-documentation/bash.1, lib/readline/doc/rluser.texinfo
-	- changed meta-flag to input-meta
-
-documentation/Makefile.in
-	- use groff -Tascii to convert .1 -> .0
-
-subst.c
-	- new function, strip_trailing_ifs_whitespace, does the obvious
-
-builtins/read.def
-	- call strip_trailing_ifs_whitespace before assigning last
-	  variable to remainder of input string [in 1.14.5]
-
-				    5/5
-				    ---
-builtins/hashcom.h
-	- changed check_dot member of the PATH_DATA struct to flags, and
-	  added a HASH_CHKDOT define to replace the check_dot semantics
-	- add HASH_RELPATH define for flags value
+	- fixed translation of ESC in rl_untranslate_keyseq and
+	  rl_untranslate_macro_value
 
-builtins/hash.def, builtins/common.c
-	- instead of xxx->check_dot, use (xxx->flags & HASH_CHKDOT)
-
-builtins/hash.def
-	- if the full pathname to which a command is being hashed does
-	  not begin with a `/', set the HASH_RELPATH flag for it
-
-builtins/common.c
-	- if a hashed filename as HASH_RELPATH set, check ./filename,
-	  returning null if that filename is not executable
-
-execute_cmd.c, general.c, execute_cmd.h, general.h
-	- moved same_file and check_binary_file from execute_cmd.c to
-	  general.c
-
-hashlib.c, Makefile.in
-	- renamed hash.c to hashlib.c
-
-hashlib.h, Makefile.in, builtins/Makefile.in, alias.h, variables.h,
-execute_cmd.c, hashlib.c, builtins/hashcom.h
-	- renamed hash.h to hashlib.h
-
-variables.c, alias.c
-	- removed inclusion of `hash.h'; header files already include
-	  correct file
+lib/readline/{readline,kill}.c
+	- added an argument to _rl_fix_point telling it whether or not to
+	  fix up the mark also; changed calls to _rl_fix_point to add the
+	  appropriate argument
 
 Makefile.in
-	- new rule to remake all the Makefiles (`make Makefiles')
-	- more changes to adhere to GNU coding standards for the various
-	  flavors of `clean' targets
-
-				    5/8
-				    ---
-documentation/Makefile.in
-	- use `texi2dvi' to make dvi files rather than tex and texindex
-	  [in 1.14.5]
-	- don't install `bash_builtins.1' [in 1.14.5]
+	- changed the substitution delimiter in the `sed' commands that
+	  create bashbug from `:' to `%' to avoid conflicts with options
+	  containing `:'
 
-Makefile.in
-	- removed instances of doc-support/texindex
-
-				    5/9
+				    4/3
 				    ---
-make_cmd.c
-	- new function make_bare_word, to make a WORD_DESC from a string but
-	  not to set any of its flags
-	- broke make_word into make_bare_word and make_word_flags
-	- fixed a bug in make_word_flags to make backslash-quoting a quoting
-	  character work right to not set the W_QUOTED flags
-
-subst.c, array.c, execute_cmd.c
-	- call make_bare_word instead of make_word in situations where we
-	  don't want the flags set inadvertently
-
-				   5/11
-				   ----
-subst.c
-	- removed special handling of $POSIX_PEDANTIC
-
-				   5/12
-				   ----
-shell.c
-	- broke the code that fetches the uids and gids into a new
-	  function, uidget()
-
-subst.c
-	- EUID and UID are no longer handled specially, since they're
-	  readonly
-	- removed sv_uids function
-	- merge sv_histfilesize into sv_histsize
-
-variables.c
-	- instead of calling sv_uids from initialize_shell_variables,
-	  call uidset() instead
-	- sv_uids -> uidset with mods to make it faster and avoid an
-	  extra call to free, malloc, and itos
-
-bashhist.c
-	- call sv_histsize instead of sv_histfilesize
-
-				   5/13
-				   ----
-pathexp.c
-	- moved setup_ignore_patterns here from bashline.c, so GLOBIGNORE
-	  works even when readline is compiled out of the shell
-	- changed the ignore data structure to a `struct ignorevar',
-	  defined in pathexp.h
-	- added functions to implement GLOBIGNORE
-
-bashline.c
-	- setup_ignore_patterns now in pathexp.c
-	- renamed _ignore_names to ignore_completion_names
-
-subst.c
-	- new function sv_globignore to be called when GLOBIGNORE changes
-	  value
-
-documentation/{bash.1,features.texi}
-	- documented GLOBIGNORE
-
-				   5/15
-				   ----
-documentation/texinfo.tex
-	- upgraded to version 2.145 from autoconf-2.3 distribution
-
-				   5/16
-				   ----
-lib/readline/display.c
-	- delicate surgery on rl_redisplay, update_line, and _rl_update_final
-	  to convert to using the line breaks array instead of assuming that
-	  lines wrap because they're too long and using absolute buffer
-	  positioning calculated from the screen width.  Now the `lithist'
-	  shopt option works right
-
-lib/readline/complete.c
-	- make sure insert_match doesn't double an opening quote character
-	  after make_quoted_replacement adds an opening quote [in 1.14.5]
-	- make sure append_match doesn't double a closing quote character
-	  [in 1.14.5]
-
-quit.h
-	- new macros: SETINTERRUPT, CLRINTERRUPT, ADDINTERRUPT, DELINTERRUPT
-	  to manipulate the value of interrupt_state
-
-trap.c, sig.c, jobs.c
-	- changes to use the new macros from quit.h
-
-jobs.c
-	- make an a job that exits due to SIGINT make the shell act as if
-	  it received the interrupt itself, but only if SIGINT is not
-	  trapped [in 1.14.5]
-
-				   5/18
-				   ----
-builtins/common.c
-	- fix up find_hashed_filename and the HASH_RELPATH code.  still need
-	  to use `shopt -s checkhash' to check the hashed pathnames.  could
-	  fix this up more to make that unnecessary in the HASH_RELPATH case
-
-subst.c
-	- new function get_array_value that does array subscripting for
-	  things like aa[1], for use by other parts of the shell like the
-	  expression evaluator
-
-expr.c
-	- changes to make things like $(( aa[1] + aa[2])) work without
-	  using ${aa[1]}
-
-bashhist.c
-	- converted the HISTIGNORE code to use the `struct ignorevar'
-	  framework, with a callback function histignore_item_func to
-	  set the HIGN_EXPAND flag if needed
-
-pathexp.c
-	- made the `globignore' variable static
-	- redid the code that removes ignored names from the `names' array
-	  in ignore_globbed_names to make it more efficient
-
-bashline.c
-	- made the `fignore' variable static
-
-				   5/20
-				   ----
-Makefile.in
-	- made the `distclean' target remove the Makefiles in subdirectories
-	  created by autoconf
-
-				   5/22
-				   ----
-builtins/command.def
-	- don't allow `command -p' if the shell is restricted
-
-documentation/bash.1
-	- documented the restriction on command -p for rbash
-
-				   5/23
-				   ----
-aclocal.m4
-	- add a new macro BASH_CHECK_TYPE, based on AC_CHECK_TYPE, that
-	  allows the caller to specify the header files to be included
-	  in the test program and provides for a default value to be
-	  defined if the type is found in the system files
-
-configure.in
-	- check for getrusage and gettimeofday functions/syscalls
-	- use BASH_CHECK_TYPE to check for clock_t in sys/types.h and
-	  sys/times.h
-	- use BASH_CHECK_TYPE to check for sigset_t instead of a special
-	  BASH_TYPE_SIGSET_T
-	- use BASH_CHECK_TYPE to check for quad_t instead of BASH_QUAD_T
-	- new argument --enable-command-timing to compile in the `time'
-	  reserved word and command timing
-
-general.c
-	- print_time_in_hz takes a `clock_t', not a `time_t'
-
-execute_cmd.c
-	- support for timing pipelines with a new function time_command
-	- execute_command_internal calls time_command if it is passed a
-	  command with the CMD_TIME_PIPELINE bit set in command->flags
-	- new functions difftimeval and addtimeval to do arithmetic on
-	  timeval structs with overflow
-
-parse.y
-	- new production: pipeline_command, used by list1 and simple_list1
-	- pipeline_command includes rules to handle `!' and time
-	- code to recognize `time' as a reserved word
-
 print_cmd.c
-	- new code to print `time ' before a command if the CMD_TIME_PIPELINE
-	  flag bit is turned on
-
-builtins/test.def
-	- added description of string1 < string2 and string1 > string2 to
-	  the long doc
-
-test.c
-	- added `<' and `>' string binary operators
-
-documentation/bash.1
-	- documented the `time' reserved word and command timing
-	- documented the new test `<' and `>' binary operators
-
-documentation/features.texi
-	- documented the `time' reserved word and command timing
-	- added more to the section detailing the differences between bash
-	  and sh
-	- added descriptions of LINENO and ENV to the Ksh variables section
-	- added description of echo to bash builtins section
-	- added PPID, BASH, SHLVL to the Bash variables section
+	- made the initial value and the default growth value for the
+	  printed command somewhat smaller -- we don't really need to
+	  allocate 4096 bytes for the printed command
+	- added stdarg support to xprintf if PREFER_STDARG is defined
 
-subst.c
-	- added a `quoted' parameter to extract_dollar_brace_string; changed
-	  all calls to it
-
-				   5/24
-				   ----
-builtins/let.def
-	- wrote code for an `exp' builtin that treats all of its arguments
-	  as an expression, concatenates them like `eval', and runs the
-	  expression evaluator
-
-expr.c
-	- added code to do the Posix.2 conditional operator: expr?expr:expr
-	- added a `noeval' flag to suppress evaluation.  currently it only
-	  suppresses assignment
-	- added code to the && and || functions so that evaluation is
-	  suppressed in the part of the statement that is not supposed to
-	  be executed (a && b: set noeval if a is false; a || b; set
-	  noeval if a is true)
-
-				   5/25
-				   ----
-documentation/{bash.1,features.texi}
-	- documented new `expr?expr:expr' conditional expression syntax
-	  now arithmetic evaluation is Posix.2-conformant
-
-lib/readline/readline.c
-	- added several more locale names to the list of legal $LANG
-	  values
+stringlib.c
+	- changed strip_trailing to take the index of the last character
+	  as the second argument, saving a (useless) call to strlen, since
+	  the caller already knows where the end of the string is
 
 subst.c
-	- fixed a bug in parameter_brace_patsub: when replacing a string 
-	  with nothing, rep was set to "", and the code attempted to
-	  free "", which the GNU malloc upchucked on
-
-bashhist.c
-	- just add a line to the history if command_oriented_history is
-	  set to 1 and the current line in the command is > 1; don't
-	  even bother checking history_ignore
+	- change call to strip_trailing in command_substitute to add the
+	  new second argument
 
-parse.y
-	- new variable `two_tokens_ago' to remember the token read before
-	  `token_before_that'
-	- fix to history_delimiting_chars to make sure that no semicolon
-	  is added after `()' (assume its a function definition), but that
-	  a semicolon is added after other `)' (assume its a parenthesized
-	  command)
-
-				   5/31
-				   ----
-tests/run-all
-	- put `.' first in $PATH and don't export ENV
-	- set THIS_SH to ../bash if it's unset
-
-lib/readline/complete.c
-	- only try to find a word break character in rl_complete_internal
-	  if we hit the end of the input string and found_quote == 0
-	  (the opening quote could have been the last character in the
-	  string)
+externs.h
+	- changed extern declaration for strip_trailing
 
-				    6/2
+				    4/4
 				    ---
-subst.c
-	- changed make_quoted_char to return CTLNUL\0 if passed a \0
-	  (this is what quote_string does, too) [in 1.14.5]
-	- changed list_string to use make_quoted_char when adding a quoted
-	  null argument due to a null field when ifs != ' \t\n'
-	- added an `expandpat' argument to getpattern, which tells it to
-	  call string_extract_double_quoted if the pattern expression is
-	  double-quoted, and changed all calls to initially pass `1' as
-	  its value
-
-jobs.c
-	- changed start_job so that an attempt to start a job marked as
-	  JDEAD elicits an error message [in 1.14.5]
-
-Makefile.in
-	- fixed `distclean' target so that it removes Makefiles in subdirs
-	  *after* descending into them to do submakes (!)
-	- fixed `realclean' target to remove everything that distclean does
+Makefile.in, configure.in, lib/malloc/Makefile.in
+	- changed the strategy for picking which `malloc' to include by
+	  having configure define a `malloc target' and the Makefile in
+	  lib/malloc implementing rules for that target
 
-				    6/5
+				    4/5
 				    ---
-builtins/declare.def
-	- fixed a typo that made `declare +r var' turn off read-only status
-	  for a variable [in 1.14.5]
-	- added -p option to short doc and long doc
-	- added code to allow `declare -f -options name' to set and unset
-	  attributes for the named functions.  Only when no other options
-	  are supplied will the named and value of the function be displayed
-	  [in 1.14.5]
-
-variables.h
-	- new SETVARATTR macro to set or unset attributes for a specific
-	  SHELL_VAR *
-
-builtins/setattr.def
-	- use SETVARATTR
+Makefile.in
+	- slightly changed the rules for remaking `parser-built':  it is
+	  now a copy of y.tab.h, updated only when the contents of y.tab.h
+	  change
+	- everything that used to depend on y.tab.h now depends on
+	  parser-built
 
-				    6/7
+				    4/6
 				    ---
-execute_cmd.c
-	- moved the retrieval of $PS3 inside the loop in execute_select_command
-	  so that PS3 can be modified in the select command's body
-	  [in 1.14.5]
-
-execute_cmd.c
-	- changed execute_builtin and execute_function to not set
-	  builtin_env and function_env, respectively, to NULL if there is
-	  no temporary env.  This makes the temp env persist across calls
-	  to functions from other functions [in 1.14.5]
-
-configure.in
-	- Linux needs LOCAL_LDFLAGS set to -rdynamic
+execute_cmd.c, print_cmd.c
+	- use #include <y.tab.h> so we pick up y.tab.h from the build
+	  directory instead of the source directory if it happens to be
+	  recreated in the build directory
 
-				    6/8
+				    4/7
 				    ---
-general.c
-	- canonicalize_pathname should not attempt to interpret backslash
-	  quoting `/', since Unix doesn't really allow it [in 1.14.5]
-
-bashline.c
-	- added `\', `!', and `)' to the list of characters which
-	  cause filenames to be quoted
-	- changed bash_quote_filename to use any of the three shell
-	  quoting styles based on the value of a variable,
-	  completion_quoting_style
-	- if *qcp is not 0 when passed to bash_quote_filename, adjust the
-	  type of completion we're doing based on its value (i.e.,
-	  *qcp == '"' forces double quoting, *qcp == '\'' forces single)
-	- bash_quote_filename now leaves the quotes intact in the filename
-	  it returns.  The readline completion code takes care of avoiding
-	  doubled open quotes
-	- if a filename containing a `!' is passed to bash_quote_filename
-	  without an opening quote character, and we are performing history
-	  expansion, use single quoting as the quoting style
-
-bashhist.c
-	- changed bash_history_disable to not call bash_history_reinit(0),
-	  but do what it needs to directly
-	- new function bash_history_enable
-	- make history_expansion_inhibited exist only if BANG_HISTORY
-	  is defined
-
-bashhist.h
-	- extern declaration for bash_history_enable
-
-lib/readline/complete.c
-	- make make_quoted_replacement set should_quote to 1 if the quote
-	  character is `'' as well as if it's `"'
-
-test.c
-	- added the csh-like `=~' and `!~' pattern-matching binary operators
-	  which match the string on the lhs against the shell pattern on
-	  the rhs.  PATTERN_MATCHING must be defined for this to work; it is
-	  undefined and undocumented by default
-
-jobs.c
-	- broke the code that gets the new window size and sets $LINES and
-	  $COLUMNS out of sigwinch_sighandler into a new function,
-	  get_window_size, which sigwinch_sighandler calls
-
-				   6/12
-				   ----
-parse.y
-	- new function, pop_expansion, to remove the top string on the
-	  expanded token stack
-	- renamed save_expansion to push_expansion
-
-lib/readline/complete.c
-	- fixed a bug in find_completion_word that tested found_quote
-	  instead of quote_char when trying to decide if we have an unclosed
-	  quoted string [in 1.14.6]
-
-				   6/26
-				   ----
-subst.c
-	- fixed expand_word_internal to remove all traces of $*, even if
-	  it's quoted, if there are no positional parameters and there are
-	  other characters in the expansion
-
 bashline.c
-	- don't attempt hostname completion if multiple consecutive `@'
-	  characters appear
+	- fixed another problem with `pwd`/[TAB] thinking that the `/
+	  started an unclosed command substitution, generating errors
 
-				   6/27
-				   ----
-shell.c
-	- initialize top_level early, and exit if a longjmp sends us there
-	  before we reinitialize
-
-subst.c
-	- more fixes to expansion of quoted $* when no positional parameters
-	- broke the code that assigns a value to an array element (and parses
-	  the array element reference) into a new function,
-	  do_array_element_assignment
-
-builtins/read.def
-	- new function, bind_read_variable, to allow binding simple variables
-	  and array elements to strings read (uses do_array_element_assignment)
-	- changed occurrences of bind_variable to bind_read_variable where
-	  it matters
-
-variables.c
-	- tentative change to bind_variable to make x=y the same as x[0]=y
-	  if x is already an array variable.  This works for `read' as well.
-	  This is what ksh does.
-
-				   6/28
-				   ----
-alias.h
-	- added a `flags' member to the ASSOC struct and renamed it to
-	  `alias_t'
-
-alias.c, bashline.c, builtins/alias.def, builtins/type.def
-	- changed ASSOC to alias_t
-
-alias.c
-	- changed add_alias to set the AL_EXPANDNEXT flag when the alias is
-	  inserted into the hash table
-
-parse.y
-	- added a third parameter to push_string: a pointer to the alias_t
-	  that is being expanded
-	- push_string marks the alias it's passed as being expanded
-	  (AL_BEINGEXPANDED)
-	- pop_string marks the alias being popped as no longer being
-	  expanded
-	- changed shell_getc to defer popping an alias expansion until
-	  the parser has had a chance to catch up, since the parser reads
-	  ahead and would cause the expansion to be popped before it
-	  fully parsed the expanded string
-	- changed alias_expand_word to check the AL_BEINGEXPANDED flag
-	  instead of the expanded_token_stack when checking whether an
-	  alias is already being expanded
-
-oslib.c
-	- make the definition of bzero be surrounded by #ifdef HAVE_BZERO
-	  rather than lumping it in with bcopy
-
-configure.in
-	- add a test for bzero
-
-config.h.in
-	- add a template for HAVE_BZERO
-
-				    7/3
-				    ---
-builtins/set.def
-	- added new meaning for `set +o' without options, as per the latest
-	  draft of Posix.2.  It means to list -o options as a series of set
-	  commands to recreate the current settings
-
-documentation/bash.1
-	- augmented description of the `set' builtin
-
-				    7/6
-				    ---
-jobs.c
-	- make sure that temp_handler is not set to SIG_DFL before calling
-	  it from waitchld
-
-builtins/cd.def
-	- rewrote cd_builtin to make the code flow clearer
-	- broke full directory name construction out into a separate
-	  function: mkpath
-	- cd_builtin now tests that a directory constructed from a $CDPATH
-	  entry is actually a directory before trying to chdir to it
-	- added an error message if $HOME is not set
-	- added a descriptive error message if `cd -' used and OLDPWD unset
-	- changed the error messages to use builtin_error, not file_error
-	- moved the code at the bind_and_exit label into a function,
-	  bindpwd (no_symlinks)
-	- Posix.2 says that when using $CDPATH, the resultant value of PWD
-	  should have no symlinks
-
-				    7/7
-				    ---
-input.c
-	- if fd_to_buffered_stream fails in some way and returns a NULL
-	  buffered stream, with_input_from_buffered_stream uses a function
-	  that does nothing but return EOF as the `getter'.  This keeps
-	  the shell from crashing if invoked with fd 0 closed
-
-subst.c
-	- made the `quoted' parameter to expand_word_internal into a flags
-	  word with flag values defined in shell.h.  Each place where
-	  `quoted' is tested for a non-zero value was changed to test
-	  (quoted & (Q_HERE_DOCUMENT|Q_DOUBLE_QUOTES)) != 0 instead
-
-mailcheck.c, shell.c, parse.y
-	- change calls to the expand_string* functions to pass an explicit
-	  Q_DOUBLE_QUOTES where appropriate
-
-				    7/9
+				    4/8
 				    ---
-subst.c
-	- getpattern now passes Q_NOQUOTE to expand_word_internal, indicating
-	  that quotes in the pattern spec are not to be treated specially
-
-				   7/12
-				   ----
-jobs.c
-	- fixed up the MUST_UNBLOCK_CHILD code in wait_for so that it now
-	  unblocks all signals and sets the signal handler for SIGCHLD to
-	  SIG_DFL.  pre-v4 SCO machines seem to require this [in 1.14.5]
-
-[Bash-1.14.5 released to net 7/16]
-
-				   7/17
-				   ----
-documentation/features.texi
-	- changed the Bourne Shell builtins section to have the same format
-	  as the bash builtins section, with usage synopses
-	- added a section on the restricted shell
-
-				   7/18
-				   ----
-documentation/features.texi
-	- changed the name of this document to the Bash Reference Manual
-	- added sections on POSIX mode and restricted shell
-
-lib/readline/doc/hsuser.texinfo
-	- added the text from the bash manual page about the bash history
-	  facilities if BashFeatures is set
-
-				   7/19
-				   ----
-documentation/features.texi
-	- added more stuff to the bash basics section
-
-lib/readline/doc/rluser.texinfo
-	- added a sample inputrc file
-
-				   7/20
-				   ----
-documentation/features.texi
-	- added section on shell expansions to bash basics section
-
-builtins/Makefile.in, lib/glob/Makefile.in, lib/malloc/Makefile.in,
-lib/readline/Makefile.in, lib/termcap/Makefile.in, lib/tilde/Makefile.in
-	- ar is now called with flags `cr' when building libraries
-
-				   7/21
-				   ----
-lib/readline/complete.c
-	- fixed a bug in find_completion_word that resulted in the found_quote
-	  flag and the delimiter character not being passed back to the
-	  caller (tested (*fp) instead of (fp) before assigning found_quote;
-	  similarly for dp).  With this fix, the filename dequoting function
-	  is properly called
-
-				   7/25
-				   ----
 general.c
-	- renamed bash_tilde_expand to bash_tilde_expansion_failure_hook
-	- new function, bash_tilde_expand, that just sets interrupt_immediately
-	  and calls tilde_expand, returning what tilde_expand returns
+	- renamed bash_tilde_expansion_failure_hook to be
+	  bash_special_tilde_expansions, since that more accurately reflects
+	  its function
+	- changed tilde_initialize so that there is no failure hook -- the
+	  special expansions are handled first with the preexpansion hook
 
-subst.c, general.c, shell.c, execute_cmd.c, variables.c, bashline.c,
-builtins/cd.def
-	- changed calls to tilde_expand to calls to bash_tilde_expand instead
-
-				   7/27
-				   ----
-builtins/set.def
-	- added a -o history option to enable and disable command history
-	  saving.  This uses bash_history_enable() and bash_history_disable()
-	  It's keyed off of remember_on_history.  With this, shell scripts 
-	  can use the command history
-
-bashhist.c
-	- bash_history_enable now calls sv_history_control and sv_histignore
+lib/tilde/tilde.c
+	- new variable: tilde_expansion_preexpansion_hook -- if non-null, it
+	  points to a function that is called before standard tilde expansion
+	  is attempted
 
-builtins/fc.def
-	- if history_list() returns a null pointer, immediately return
-	  [in 1.14.6]
+lib/tilde/tilde.h
+	- extern declaration for tilde_expansion_preexpansion_hook
 
-documentation/{bash.1,features.texi}, lib/readline/doc/hsuser.texinfo
-	- added a description of the `set -o history' option and changed
-	  text to refer to it rather than strictly interactive shells
+doc/{bash.{1,html},bashref.texi}
+	- added optional open paren to description of `case' command syntax
 
-				    8/1
+				    4/9
 				    ---
 variables.c
-	- new builtin variable $HOSTNAME, initialized to current_host_name
-	  as set in shell.c
-
-documentation/bash.1, documentation/features.texi
-	- documented $HOSTNAME
-
-				    8/3
-				    ---
-support/texi2html
-	- new program, a texinfo-html converter
-
-documentation/Makefile.in
-	- added directives to produce features.html from features.texi
-
-MANIFEST.doc
-	- added features.html and features_toc.html to the doc distribution
-
-				   8/10
-				   ----
-lib/readline/bind.c
-	- new functions rl_variable_dumper and rl_dump_variables to print
-	  readline variables and their values to rl_outstream
-	- new functions rl_get_keymap_name and rl_get_keymap_name_from_edit_mode
-	  to get the name of the current keymap for the variable dumper
-
-lib/readline/readline.h
-	- extern declarations for rl_variable_dumper, rl_dump_variables, and
-	  rl_get_keymap_name
-
-builtins/bind.def
-	- added -V flag to print variable names and bindings
-	- added -P flag to print variable names and bindings in the syntax
-	  of the inputrc file
-
-lib/readline/funmap.c
-	- added new bindable `dump-variables' readline command
-
-documentation/{bash.{1,html},readline.3,features.html},
-lib/readline/doc/rluser.texinfo
-	- documented new `dump-variables' readline command
-
-documentation/{bash.{1,html},features.{texi,html}
-	- documented new bind -P and -V options
-
-documentation/bashbug.1
-	- new manual page
-
-documentation/Makefile.in
-	- changes to build and install bashbug.1
-
-				   8/15
-				   ----
-print_cmd.c
-	- new function `xprintf' that just calls vfprintf(stdout, ...)
-	  to avoid having to declare printf
-
-parse.y
-	- slight change in how read_token_word computes `all_digits'
-
-				   8/21
-				   ----
-
-subst.c
-	- new function dequote_escapes to remove CTLESC escaping
-	  CTLESC and CTLNUL in a string, returning a new string
+	- on qnx, set and export a variable `NODE' which contains the QNX
+	  `node id'
 
-				   8/25
-				   ----
-shell.c
-	- execute the file SYS_BASHRC for interactive shells if SYS_BASHRC
-	  is defined, before executing ~/.bashrc
-
-config.h.top
-	- add a dummy commented definition for SYS_BASHRC, defaulting to
-	  /etc/bash.bashrc
-
-lib/readline/input.c
-	- added code to call select() in rl_gather_tyi() and return
-	  immediately if it indicates that there is nothing to read on
-	  the readline input fd (#ifdef HAVE_SELECT)
-
-lib/posixheaders/posixdir.h
-	- new file to localize the <dirent.h>/<sys/dir.h> and dirent/direct
-	  mess
-
-posixdir.h, lib/readline/posixdir.h
-	- symlinks to lib/posixheaders/posixdir.h
-
-lib/readline/rldefs.h
-	- removed the posix dir includes and defines
-
-lib/readline/complete.c, getcwd.c
-	- include posixdir.h rather than having the code inline
-
-builtins/cd.def
-	- new code to do spelling correction on the directory name
-	  modified from patch sent by Neil Russell (caret@c-side.com)
-
-				   8/29
-				   ----
-builtins/shopt.def
-	- new option `cdspell' to toggle cd directory name spelling
-	  correction
-
-documentation/{bash.{1,html},features.texi}
-	- documented new shopt `cdspell' option
-
-				    9/5
-				    ---
-subst.c
-	- $'' should expand to the same thing as '' (a quoted null string)
-
-shell.h
-	- changed CTLNUL to '\177'; some scripts use ^B for things like IFS
-	  [in 1.14.6]
-
-				    9/7
-				    ---
-support/config.guess
-	- added code to identify a PPC running Solaris 2
-
-doc
-	- new directory replacing documentation
-
-Makefile.in, MANIFEST, MANIFEST.doc, configure.in
-	- replaced `documentation' with `doc'
-
-				    9/8
-				    ---
-lib/readline/display.c
-	- fixed the code in _rl_update_final that decides whether or not the
-	  cursor is at character position 0 on an otherwise-empty line and
-	  adjusts _rl_vis_botlin accordingly
-
-				   9/13
-				   ----
 general.c
-	- fixed canonicalize_pathname to make sure that intermediate
-	  results correspond to directories before blindly chopping
-	  off the last component when we see a `..'.  This fixes the
-	  `cd ../.../..' being equivalent to `cd ..' problem
-
-subst.c
-	- fixed make_named_pipe so that mkfifo is called with mode 0600
-	  for security reasons [in 1.14.6]
-	- changed the /dev/fd version of add_fifo_list to zero out new
-	  entries to the fifo_list when it resizes it larger [in 1.14.6]
-	- changed make_dev_fd_filename to use DEV_FD_PREFIX
-
-aclocal.m4
-	- changed BASH_HAVE_DEV_FD to check for /proc/self/fd and define
-	  DEV_FD_PREFIX to either "/dev/fd/" or "/proc/self/fd/" as
-	  appropriate
-
-				   9/14
-				   ----
-lib/readline/display.c
-	- when displaying the first line of a multiline prompt, make sure
-	  that the final \n is followed by a \r (outputting one if necessary)
-	  so that we know we are starting at column 0
-
-variables.h
-	- new attribute: att_local
-
-variables.c
-	- changed make_local_variable to set the att_local attribute
-	- changed makunbound to just make variables marked as `local' in
-	  the current context invisible.  This makes the local variable
-	  persist throughout the function even if it is unset, so a
-	  subsequent assignment preserves the `local' attribute
-	- changed kill_all_local_variables to unset the att_local attribute
-	  before calling makunbound
-
-				   9/18
-				   ----
-lib/readline/complete.c
-	- make sure the cursor is on the last line of a possibly-multiple-
-	  line command line before listing the completions in
-	  display_matches
-
-				   9/19
-				   ----
-braces.c
-	- fixed the non-SHELL case in brace_gobbler so the loop exits when
-	  it should
-
-lib/readline/bind.c, lib/readline/readline.h
-	- added functions to dump key sequences bound to macros and their
-	  values
-
-builtins/bind.def
-	- added -S and -s options to dump the readline macros and their values
-	- changed the options so that -p and -P dump functions, -v and -V
-	  dump variables, and -s and -S dump macros (s for string)
-
-doc/{bash.{1,html},readline.3,features.texi}, lib/readline/doc/rluser.texinfo
-	- added documentation for the new `bind' options
-	- added documentation for the new readline functions to dump macros
-	  and the key sequences that output them
-
-				   9/22
-				   ----
-print_cmd.c
-	- new function xtrace_print_word_list (WORD_LIST *) to print the
-	  words of a simple command when set -x is on.  This prints ''
-	  when it encounters an empty string
-
-externs.h
-	- new extern declaration for xtrace_print_word_list
-
-execute_cmd.c
-	- call xtrace_print_word_list in execute_simple_command
-
-				   9/25
-				   ----
-builtins/getopts.def
-	- make sure that the loop counter stops at 10 when stepping through
-	  the dollar_vars array to count the number of positional parameters
-	  [in 1.14.6]
-
-				   9/30
-				   ----
-lib/readline/histsearch.c
-	- fixed history_search_internal to bail immediately if it gets a
-	  null or empty search string.  This fixes the !? core dumps.
-	  [in 1.14.6]
-
-				   10/3
-				   ----
-lib/readline/histexpand.c
-	- if there is a null string given with a !? search specifier, use
-	  a previous search string if one exists, else fail immediately
-
-trap.c
-	- made run_exit_trap preserve $? around the call to execute the
-	  trap string, unless the trap string contains an `exit'
-	  command, in which case it can set the shell's exit status
-	  [in 1.14.6]
-
-	  In other words:
-
-		touch /tmp/z
-		trap 'rm /tmp/z ; exit 5' 0
-		exit 1
-
-	  exits with status 5; while
-
-		touch /tmp/z
-		trap 'rm /tmp/z' 0
-		exit 1
-
-	  exits with status 1
-
-subst.c, parse.y
-	- moved the $'...' code from subst.c to parse.y, more like ksh
-	  does it
-
-doc/bash.1, doc/bash.html, doc/features.texi
-	- moved the description of $'...' from the expansion section to
-	  the quoting section
-
-				   10/4
-				   ----
-command.h
-	- added a `line' member to the function struct for the source line
-	  the function definition starts on
-
-make_cmd.c
-	- initialize the `line' member in Function_def to 0
-	- make_function_def takes a third parameter telling which line the
-	  function definition started on and a fourth telling which line
-	  the function body started on
-
-make_cmd.h
-	- changed prototype for make_function_def
-
-parse.y
-	- new variable `function_dstart', set by read_token and read_token_word
-	  to keep track of where a function definition begins
-	- new variable `function_bstart' to keep track of where function
-	  body begins
-	- pass function_dstart and function_bstart to make_function_def
-	- new function strtrans to do the $"..." locale-specific translation
-	  of `...'
-	- call strtrans() to translate $"string".  The translated string is
-	  double-quoted
-
-doc/bash.{1,html}
-	- updated the description of LINENO now that line numbers within
-	  functions are correct
+	- QNX system can now handle pathnames with a leading `//'
 
 configure.in
-	- look for the `gettext' library function
+	- added `-Dqnx' to LOCAL_CFLAGS on QNX machines
 
-config.h.in
-	- define HAVE_GETTEXT if gettext(3) exists in a findable library
+lib/malloc/getpagesize.h
+	- some systems need sysconf(_SC_PAGE_SIZE) to obtain the page size;
+	  added code to check for it
 
-				   10/5
+				   4/10
 				   ----
-builtins/common.c
-	- changed backslash_quote so that `#' is only quoted at the start
-	  of a word
-	- new function contains_shell_metas returns 1 if the argument
-	  string contains one or more shell meta-characters that require
-	  quoting
-
-builtins/common.h
-	- extern declaration for contains_shell_metas
-
 print_cmd.c
-	- changed xtrace_print_word_list to print words containing
-	  shell metacharacters within single quotes
+	- include the prototype for cprintf only if PREFER_STDARG is defined,
+	  otherwise just have a K&R-style forward function declaration
 
-lib/termcap
-	- upgraded to GNU termcap version 1.3
+hashlib.h
+	- reduced the default number of buckets in a hash table to 53
 
-				   10/9
-				   ----
-lib/readline/readline.c
-	- call setlocale(LC_CTYPE, ...) after finding a legal value for
-	  one of LC_ALL, LC_CTYPE, or LANG
-
-				   10/11
-				   -----
-lib/readline/search.c
-	- make rl_history_search_internal just do previous-history or
-	  next-history as appropriate when given a null search string
-	  (rl_point == 0)
+lib/tilde/tilde.c
+	- prime the result string in tilde_expand by allocating a new string
+	  that's as long as the input string (+16 if a tilde appears in
+	  the string to be expanded).  This should reduce the number of
+	  reallocs
 
-				   10/24
-				   -----
 subst.c
-	- fixed an off-by-one error in char_is_quoted that skipped a
-	  characters after calling skip_single_quoted or skip_double_quoted
-	  [in 1.14.6]
-	- fixed an off-by-one error in string_extract_verbatim so it
-	  leaves *sindex at the separator character if the separator
-	  character is "'" [in 1.14.6]
-
-				   10/27
-				   -----
-parse.y
-	- in CHECK_FOR_RESERVED_WORD, make sure that reading a `}'
-	  decrements open_brace_awaiting_satisfaction if it is non-zero
-	  [in 1.14.6]
-
-shell.c
-	- don't run the shell startup files if the shell is running
-	  setuid
-	- don't source $ENV if the shell is running setuid
-
-variables.c
-	- new parameter to initialize_shell_variables: no_functions.  If
-	  non-zero, don't import functions from the environment
+	- broke the code that reads the output of a command substitution
+	  through the pipe to the subshell out into a separate function:
+	  read_comsub().  This does not use stdio, but rather reads
+	  directly from the pipe into a local 128-character buffer
 
-variables.h
-	- change to function prototype for initialize_shell_variables
-
-lib/readline/complete.c
-	- fix for the code that decides whether or not a char is quoted for
-	  applications that don't supply a value for rl_char_is_quoted_p
-	- fixed insert_match to not remove a user-supplied opening quote
-	  character if make_quoted_replacement does not return a string
-	  beginning with that quote character
-
-				   11/2
-				   ----
-general.c
-	- new function ungetc_with_restart that handles the local buffering
-	  [in 1.14.6]
-
-parse.y
-	- changed yy_stream_unget to call ungetc_with_restart if the OS
-	  does not have restartable syscalls [in 1.14.6]
-
-				   11/3
-				   ----
-bashline.c
-	- rewrote _ignore_completion_names to actually free and remove
-	  names from the array if more than one names in the array
-	  passed is acceptable, instead of just bailing [in 1.14.6]
-
-				   11/7
+				   4/11
 				   ----
-bashline.c
-	- if no matches are acceptable to _ignore_completion_names,
-	  free the entries in NAMES, set NAMES[0] == 0, and let the
-	  caller clean up [in 1.14.6]
-	- fixed a problem with backslash-quoted characters in
-	  bash_dequote_filename that caused the character after the
-	  backslash to be interpreted
-	- bash_dequote_filename now takes a quote_char parameter that,
-	  if non-zero, gives the quote character (`'' or `"') that
-	  delimits the filename.  Used to initialize the quoting state
-
-lib/readline/complete.c
-	- if the completion ignore function returns with MATCHES == 0
-	  or MATCHES[0] == 0, ring the bell and quit the completion
-	  attempt [in 1.14.6]
-	- pass quote_char to gen_completion_matches so it can pass it
-	  along to the app-specific filename dequoting function
-	- have gen_completion_matches pass quote_char to whatever function
-	  is pointed to by rl_filename_dequoting_function
+execute_cmd.c
+	- some systems need both <sys/time.h> and <time.h>, so include both
+	  if it's possible, otherwise include <sys/time.h> (if present)
 
-pathexp.c
-	- rewrote ignore_globbed_names to be more like _ignore_completion_names
+lib/readline/rl{tty,defs}.h
+	- moved includes of <sys/stream.h>, <sys/ptem.h>, etc. to rltty.h
 
-lib/readline/doc/rltech.texinfo
-	- updated documentation for rl_filename_dequoting_function
+lib/readline/terminal.c
+	- include rltty.h after rldefs.h
 
-				   11/10
-				   -----
-subst.c
-	- fixed remove_quoted_nulls so that it is now a full function that
-	  removes unquoted CTLNUL chars from the string it is passed
-	  [in 1.14.6]
-	- fixed expand_word_internal to avoid generating some unneeded
-	  quoted nulls (if the string is partially quoted, note that we
-	  have seen a quoted null and add one if the rest of the string
-	  doesn't expand to anything)
-
-				   11/13
-				   -----
 variables.c
-	- bind HOSTTYPE, OSTYPE, and HOSTNAME unconditionally
-
-pathexp.c
-	- new function quote_globbing_chars, adds `\' before globbing
-	  chars in its string argument, returns new string
-
-pathexp.h
-	- extern declaration for quote_globbing_chars
+	- changes to make environment creation faster and use less memory
+	  (fewer malloc/free calls, too):
+
+		o two new variables: export_env_index (how many environment
+		  strings are in export_env) and export_env_size (the
+		  number of slots for strings allocated in export_env)
+		o added new function add_to_export_env, since adding the
+		  exported shell variables and shell functions does not
+		  need to search the export_env for a definition to supersede
+		  (we just cleared it out!)
+		o renamed add_or_supersede to add_or_supersede_exported_var,
+		  since it always works on export_env, and changed the second
+		  argument to a flag saying whether or not to allocate a new
+		  copy of the string placed into the environment
+		o changed calls to add_or_supersede to the new
+		  add_or_supersede_exported_var with the appropriate flags
+		o don't free and reallocate export_env in maybe_make_export_env,
+		  just free the strings and start anew
+		o prime the size of export_env from the total number of shell
+		  variables and shell functions -- this will always be enough
+		  for the exported shell functions and variables, and big
+		  enough most of the time for the entire environment
 
-bashhist.c
-	- if the previous line contains globbing chars, run it through
-	  quote_globbing_chars before trying to match.  This affects
-	  only HISTIGNORE patterns containing `&'
+builtins/cd.def
+	- efficiency hack in bindpwd():  if PWD is exported, we will have to
+	  rebuild the entire exported environment after every time we change
+	  directories.  What we do is see if array_needs_making changes value
+	  from 0 to 1 after bind_variable ("PWD", dirname) is called, and
+	  that PWD is exported.  If that happens, we just replace the value
+	  of PWD in the exported environment with a call to
+	  add_or_supersede_exported_var
+
+bashline.c, parse.y
+	- check calls to pre_process_line to make a fresh copy of the line
+	  if pre_process_line returns what it was passed, in preparation
+	  for future changes
 
-				   11/14
-				   -----
 bashhist.c
-	- if the history line to be added contains globbing chars, quote
-	  them with backslashes by calling quote_globbing_chars in
-	  history_should_ignore before comparing them against the
-	  patterns in HISTIGNORE
-
-bashline.c
-	- make sure that we erase the current readline line after running
-	  fc on it and executing the resultant commands in
-	  vi_edit_and_execute_command, so the original readline line
-	  doesn't get returned [in 1.14.6]
-
-jobs.h
-	- added a new job listing format: JLIST_NONINTERACTIVE -- like
-	  JLIST_LONG, but does not print the job number
-
-jobs.c
-	- added code to pretty_print_job to handle JLIST_NONINTERACTIVE
-	- call notify_and_cleanup() from wait_for even if the shell is
-	  running a script (interactive_shell == 0)
-	- changed notify_and_cleanup to call notify_of_job_status if
-	  interactive or interactive_shell == 0, so scripts report
-	  about jobs they run
-	- changed notify_of_job_status to call pretty_print_job with a
-	  JLIST_NONINTERACTIVE format if interactive_shell is 0 and a
-	  job is marked JDEAD, after printing the script name and line
-	  number.  This message is printed only if the job dies due to
-	  a fatal signal
-
-support/mkversion.c
-	- added support for a `-status status' argument to set the
-	  `release status' of the shell (alpha, beta, or release).  It
-	  defines `RELSTATUS' in version.h and changes the definition
-	  of SCCSVERSION, if present
-
-Makefile.in
-	- set a RELSTATUS variable that is included when printing the build
-	  message and passed to mkversion with the -status option
-	- make RELSTATUS one of the variables sed sets when it creates
-	  bashbug from support/bashbug.sh
-
-version.c
-	- added a `release_status' variable that's set to RELSTATUS if
-	  it's defined
-
-support/bashbug.sh
-	- added the `RELSTATUS' variable to the report with heading
-	  `Release Status'
-	- set the bug address to chet@po.cwru.edu if the release status
-	  is `alpha' or `beta'
-
-				   11/15
-				   -----
-
-shell.c, execute_cmd.c
-	- new variable expand_aliases to control alias expansion.  For now,
-	  this is set to the same value as interactive_shell when that is
-	  set
-
-parse.y
-	- perform alias expansion if expand_aliases is non-zero rather than
-	  checking the value of interactive_shell
-
-				   11/16
-				   -----
-builtins/shopt.def
-	- new option `expand_aliases' to control the value of expand_aliases
-
-doc/bash.1, doc/bash.html
-	- updated the description of `shopt' with the `expand_aliases' option
-
-				   11/28
-				   -----
-bashline.c
-	- if there is only one completion in _ignore_completion_names, see
-	  if it is acceptable and return right away [in 1.14.6]
+	- pre_process_line now returns its argument if it did not make
+	  any changes to it
 
-configure.in
-	- change to define WAITPID_BROKEN on SCO 3.2v5
+alias.c
+	- free the bucket entry holding the alias to be removed in
+	  remove_alias, as well as the data
 
-				   12/6
+				   4/14
 				   ----
-parse.y
-	- call prompt_again in read_token_word if a newline is read in an
-	  interactive shell and bash_input.type is either st_stdin or
-	  st_stream
-	- remove superfluous call to reset_readline_prompt in yy_readline_get
+unwind_prot.c
+	- if an unwind-protect frame is being discarded, and its cleanup
+	  function is `restore_variable', the `arg' member points to a
+	  SAVED_VAR that must be freed.  This change is made in
+	  remove_unwind_protect_internal and unwind_frame_discard_internal
 
-				   12/7
-				   ----
 parse.y
-	- combine delimiters, delimiter_depth, and delimiter_space into a
-	  single structure of type `struct dstack'
-	- replace all references to delimiter* with dstack.delimiter*
-	- include parser.h for struct dstack
-	- new define pop_delimiter(), analogous to push_delimiter
-	- current_delimiter, push_delimiter, and pop_delimiter defines now
-	  include the delimiter stack struct as the first parameter
-
-parser.h
-	- now includes definition of struct dstack
-
-bashline.c
-	- include parser.h for struct dstack
-	- refer to dstack.delimiter_depth instead of delimiter_depth
-
-				   12/12
-				   -----
-execute_cmd.c
-	- before doing a longjmp(subshell_top_level,...) when executing a
-	  shell script without a leading #!, set history_lines_this_session
-	  to 0 to forget about the history and not save it on an exec
-	  (we don't free the memory with clear_history(), though -- that
-	  would slow bash down)
-
-				   12/14
-				   -----
-jobs.c, nojobs.c
-	- cause the sigwinch handling code to be compiled into the shell
-	  even if READLINE is defined
-	- new functions: set_sigwinch_handler and unset_sigwith_handler
-	  to enable and disable catching of SIGWINCH and adjusting $LINES
-	  and $COLUMNS
-	- changed initialize_job_signals to install a signal handler for
-	  SIGWINCH with set_sigwinch_handler
-
-jobs.h
-	- extern declarations for set_sigwinch_handler and
-	  unset_sigwinch_handler
-
-				   12/20
-				   -----
-doc/bash.{1,html}, doc/readline.3, lib/readline/doc/rluser.texinfo
-	- documented the readline `visible-stats' variable
+	- need to free memory allocated by parse_arith_cmd if it is an
+	  arithmetic command, after using it to make a new word
 
-				   12/21
-				   -----
-trap.c
-	- new global variable `running_trap' incremented and decremented
-	  around running a trap command in _run_trap_internal
-
-execute_cmd.c
-	- new variable currently_executing_command, set to the COMMAND *
-	  currently being processed by execute_command_internal, unless a
-	  trap command is being run (running_trap != 0)
-	- new function executing_line_number, which returns the line number
-	  of the currently executing command (which may not be the same as
-	  line_number)
-	- don't run the debug trap if it was not set before the current
-	  simple command was executed, since we don't want to run it after
-	  the trap command that sets the DEBUG trap
-
-variables.c
-	- new function assign_lineno, to assign a value to line_number
-	  when LINENO is set
-	- changed get_lineno to call executing_line_number() rather than
-	  returning line_number
-
-parse.y
-	- new argument for push_stream telling it whether or not to reset
-	  line_number to 0
+subst.c
+	- fixed some memory leaks caused by not freeing the argument to
+	  make_bare_word, which duplicates its string argument
+	- need to dispose list generated by list_rest_of_args in
+	  paramter_list_remove_pattern
+	- make sure the return value from getpattern() is freed
+	- make sure array_value_internal always returns newly-allocated
+	  memory
+	- get_var_and_type returns a new type: VT_ARRAYMEMBER if the
+	  string passed is of the form var[index]
+	- make sure parameter_brace_substring frees the memory allocated
+	  by get_var_and_type if verify_substring_values returns 0
 
-builtins/common.c
-	- change to parse_and_execute to call push_stream with an argument
+hashlib.c, hashlib.h
+	- new function, dispose_hash_table (table), which frees the
+	  table's bucket array and the table itself
 
-				   12/29
-				   -----
-subst.c
-	- set subshell_environment in command_substitute and process_substitute
-	  for the child process
+alias.c
+	- call dispose_hash_table from delete_all_aliases instead of just
+	  freeing the table
 
-				    1/2
-				    ---
-trap.c
-	- made decode_signal recognize signal names case insensitively
+pathexp.c
+	- make sure to free `newnames' (but *not* its contents) before
+	  returning from ignore_globbed_names
 
-shell.c
-	- make the shell exit on a longjmp (DISCARD, ...) if
-	  subshell_environment is non-zero
+builtins/exec.def
+	- make sure the argv created to pass to shell_execve is freed if
+	  the execve fails and we're not exiting on failed execs
 
-				   1/16
-				   ----
-lib/readline/histexpand.c
-	- let the `!' in ${!xxx} pass through without error
+expr.c
+	- broke evalexp into two functions: evalexp, which sets up the
+	  jmp_buf that errors jump to, and subexpr, which does the
+	  evaluation and pushing and popping of contexts
+	- readtok now calls subexpr to evaluate subexpressions in
+	  parentheses
+	- evalexp now takes an additional paramter, a pointer to an int.
+	  If the expression contains errors, the location to which this
+	  points gets 0, otherwise it gets 1 to show that the value
+	  returned may be used.  This plugs up memory leaks that were
+	  the result of evalexp() longjmping back to top_level
+	- fixed a memory leak: expr_stack[0] was not being freed
 
-shell.c
-	- new --help long option
-	- new function show_shell_usage() for use by --help
+externs.h
+	- changed extern declaration for evalexp
 
-				   1/19
-				   ----
-parse.y
-	- changes to shell_getc to make sure that lines consisting of only a
-	  newline get added to the history correctly when they are part of
-	  a quoted string
+variables.c, subst.c, builtins/let.def
+	- changed calls to evalexp appropriately.  They either cause a
+	  longjmp (top_level, DISCARD) (which is what the old stuff in
+	  expr.c did) or are handled by returning an appropriate error
+	  value (e.g., &expand_word_error in subst.c)
 
-				   1/24
+				   4/16
 				   ----
-aclocal.m4, configure.in
-	- added a check for -lsocket (and -lnsl) to fix solaris problems
-	  with isnetconn()
-
 shell.c
-	- rearranged the code in isnetconn to check for a socket using
-	  getpeername() first, before any SVR4 or SVR4.2-specific checks
-	- added checks for ttys (isatty) and FIFOs (S_ISFIFO) to the
-	  SVR4/SVR4.2 case of isnetconn()
-
-general.h
-	- new macro RESIZE_MALLOCED_BUFFER to check and see whether there
-	  is enough room in a string to add a given number of characters
-	  and to resize it if there is not
+	- make sure to free dollar_vars[0] before assigning it the first
+	  argument following `-c command'
 
-				   1/25
-				   ----
-parse.y, general.c, subst.c, bashhist.c, alias.c, array.c, variables.c
-	- use RESIZE_MALLOCED_BUFFER where appropriate
+variables.c
+	- if unsetting a local variable with a previous context, make sure	
+	  to free the hash table element used to store the local variable
 
-				   1/26
-				   ----
-support/config.{guess,sub}
-	- merged in latest changes from GNU master copies
+lib/readline/terminal.c
+	- rearrange the includes so <sys/ioctl.h> is included before rltty.h,
+	  as it is in rltty.c
 
-				   1/30
+				   4/17
 				   ----
+flags.c
+	- new function: reset_shell_flags, which resets all of the flags
+	  back to their initial values
 
-jobs.c
-	- make sure to freeze the jobs list when calling a trap handler
-	  for SIGINT
-	- added code to waitchld() so that a SIGINT trap handler is called
-	  if the shell is running a shell script and a SIGINT is received
-	  while waiting for a foreground job, even if that job does not
-	  die from the SIGINT
+flags.h
+	- extern declaration for reset_shell_flags
 
-jobs.c, nojobs.c
-	- make get_tty_state reset $LINES and $COLUMNS after each process
-	  exits if the variable `check_window_size' is non-zero
+builtins/set.def
+	- new function: reset_shell_options, which resets all of the -o
+	  options that are not also shell flags back to their initial values
 
 builtins/shopt.def
-	- new variable `checkwinsize', which controls the value of
-	  check_window_size
-
-doc/{bash.{1,html},features.texi}
-	- updated description of `shopt' to include `checkwinsize'
-
-execute_cmd.c, lib/readline,complete.c
-	- some changes from the GNU WIN32 project for the bash port to
-	  Windows NT and Windows 95
-
-bashwait.h
-	- new file, with `union wait' defines from jobs.h
-
-jobs.h
-	- include `bashwait.h' if <sys/wait.h> is not present and
-	  _POSIX_VERSION is not defined
-	- removed define of pid_t for non-Posix systems; now provided by
-	  autoconf in config.h
-
-				   1/31
-				   ----
-parse.y
-	- new temporary delimiter stack, used when decoding prompt strings.
-	  This is needed so command substitutions in the prompt strings
-	  (especially PS2) don't screw up the parser's quoting state
+	- new function: reset_shopt_options, which resets all of the shopt
+	  options that are not also shell flags or -o options back to their
+	  initial values
 
-lib/readline/complete.c
-	- new variable for readline library users:
-	  rl_completion_append_character.  The value of this variable is
-	  the character appended to a completion when it occurs at the
-	  end of a line.  Setting it to '\0' causes nothing to be
-	  appended.
-
-lib/readline/readline.h
-	- declaration for rl_completion_append_character
-
-lib/readline/doc/rltech.texinfo
-	- documented rl_completion_append_character as int variable available
-	  to library users
-
-subst.c
-	- new code for maintaining a string array saying which words in the
-	  output of expand_word_list_internal are the result of globbing
-
-variables.c
-	- new function put_gnu_argv_flags_into_env (pid, flags_string)
-	  to put Roland's GNU getopt helper variable into the export_env
+builtins/common.h
+	- extern declarations for reset_shell_options and reset_shopt_options
 
 execute_cmd.c
-	- in execute_disk_command, after forking the child, put the GNU
-	  getopt helper environment variable into the child's export_env
-
-unwind_prot.c
-	- changed unwind_protect_var and restore_variable to do the
-	  bcopy of the variable's value if it's shorter than the size of
-	  an int, as well as if it's longer.  This keeps stray data
-	  from being copied if a short is being unwind-protected
-
-unwind_prot.h
-	- new define, unwind_protect_short, to protect variables smaller
-	  than an int
+	- broke the code that reinitializes things when an executable script
+	  without a leading `#!' is found out into a new function:
+	  initialize_subshell
+	- initialize_subshell now calls the reset_* functions that reset the
+	  shell flags and options
 
-jobs.c
-	- in the code that handles SIGCHLD traps, call unwind_protect_short
-	  if that is the size of a pid_t (for last_made_pid)
-
-				    2/5
-				    ---
-Makefile.in
-	- only try make distclean in HIST_LIBDIR if Makefile exists -- could
-	  have already been removed if HIST_LIBDIR is the same as RL_LIBDIR
-
-general.c
-	- make canonicalize_pathname be more careful about what it checks
-	  for being a directory name when processing a full pathname
-
-Makefile.in, configure.in, doc/Makefile.in
-	- small changes to get bash to build better in a directory not the
-	  source directory
-
-				   2/12
-				   ----
-Makefile.in
-	- added `install-strip' target
+general.c, general.h
+	- move_to_high_fd now takes a third argument: the highest fd at which
+	  to start looking.  If that's less than 20, the maximum number of
+	  open files as returned by getdtablesize() is used (which is what
+	  it did before this)
 
-				   2/14
-				   ----
-doc/bash.{1,html}, doc/features.texi
-	- documented the `--verbose' startup option
+jobs.c, shell.c, subst.c
+	- changed calls to move_to_high_fd appropriately
 
-lib/readline/complete.c
-	- fix to gen_completion_matches to fix a memory leak
-	- fix to rl_complete_internal to fix a memory leak
+[bash-2.01-beta1 frozen]
 
-				   2/15
+				   4/18
 				   ----
-bashwait.h
-	- changed to use WORDS_BIGENDIAN instead of LITTLE_ENDIAN or
-	  BIG_ENDIAN
-
-configure.in, config.h.in
-	- call AC_C_BIGENDIAN, define WORDS_BIGENDIAN
-
-Makefile.in
-	- remove all references to mkendian.c, mkendian, and bash_endian.h
-	- added a `symlinks' target that just runs support/fixlinks
-
-MANIFEST
-	- mkendian.c is no longer in the distribution
+general.c
+	- itos now uses a local char buffer to do its conversion, but still
+	  returns newly-allocated memory
 
-				   2/16
+				   4/21
 				   ----
-execute_cmd.c
-	- include <sys/times.h> if HAVE_SYS_TIMES_H and HAVE_TIMES are
-	  defined
-
-bashline.c
-	- fix to bash_directory_completion_hook to compensate for
-	  canonicalize_pathname returning NULL
-
 variables.c
-	- fix to initialize_shell_variables to compensate for
-	  canonicalize_pathname returning NULL
-
-				   2/22
-				   ----
-tests/test-tests, tests/test.right
-	- changes to avoid writing in the source directory -- all temp files
-	  are created in /tmp
-
-[First alpha release at Thu Feb 22 15:59:51 EST 1996]
+	- be a little more careful when checking for backwards-compatibility
+	  with exported function definitions
 
-				   2/23
+				   4/22
 				   ----
-lib/readline/rldefs.h
-	- work around SVR4.2 bug including <sys/ptem.h> and <termios.h>
-
-lib/readline/chardefs.h
-	- fix to CTRL_CHAR macro for chars > 128 on systems with signed
-	  characters
-
 builtins/ulimit.def
-	- protect more of the RLIMIT_* defines with checks
-	- new macro RETINVALID() to set errno and return the correct
-	  value for an invalid request
-
-doc/Makefile.in
-	- add a definition for INSTALL, set by autoconf
-
-				   2/26
-				   ----
-support/bashbug.sh
-	- if USER is unset, assign it the value of $LOGNAME or `whoami`
-
-Makefile.in,{lib/*,doc,builtins}/Makefile.in
-	- use `test' instead of `[' to conform to GNU coding standards
-	- `incdir' -> `includedir' as per latest GNU coding standards
-
-lib/readline/Makefile.in, lib/glob/Makefile.in
-	- use $(srcdir)/ instead of $(srcdir) in the CSOURCES variable
-
-general.h
-	- new define, FS_NODIRS, to not find directory names when searching
-	  $PATH
-
-execute_cmd.c
-	- fix to find_in_path_element so it does not return directories
-	- executable_file() no longer returns directories as executable
+	- translate RLIM_INFINITY to limit.rlim_max if the current limit is
+	  less than or equal to the hard limit, not just strictly less than
+	  (the change of 3/28 was too drastic)
 
-				   2/27
-				   ----
-jobs.h
-	- new flag value: J_NOHUP
-
-jobs.c
-	- new function: nohup_job(job).  Sets J_NOHUP flag for specified
-	  job
-	- change hangup_all_jobs so that jobs marked J_NOHUP are not
-	  sent SIGHUP.  If stopped, the job still gets SIGCONT.
-	- changed calls to report_error to call internal_error, which will
-	  not exit the shell
-	- changed FIND_CHILD define to call internal_error, restore the
-	  SIGINT handler, set termination_state to 127 and return, rather
-	  than aborting the shell
-
-builtins/jobs.def
-	- new option for disown: `-h'.  Marks the specified jobs J_NOHUP.
-
-doc/{bash.{1,html},features.texi}
-	- added description of `disown -h'
-
-jobs.c, nojobs.c
-	- include some files needed for struct winsize by SCO
-
-lib/readline/input.c
-	- added some #ifdefs to avoid including <sys/time.h> on systems
-	  with select but without <sys/select.h>
-
-configure.in
-	- check for <stdarg.h>, define HAVE_STDARG_H in config.h if found
-
-config.h.bot
-	- define USE_VARARGS and either PREFER_STDARG or PREFER_VARARGS
-	  if one of <stdarg.h> or <varargs.h> is present
-
-aclocal.m4
-	- moved default mail directory check here from configure.in, macro
-	  name is BASH_DEFAULT_MAIL_DIR
-	- rewrote BASH_CHECK_DEV_FD to cache the value
-	- minor fixes from Bruno Haible
-
-shell.c
-	- no longer includes <varargs.h>
-
-builtins/Makefile.in
-	- add -I$(topdir)/builtins to list of includes
-
-execute_cmd.c, lib/glob/glob.c
-	- include memalloc.h for correct alloca definitions
-
-error.[ch], print_cmd.c, builtins/common.[ch]
-	- changes to include new ANSI-C stdargs code if PREFER_STDARG is
-	  defined
-
-				   2/28
-				   ----
-aclocal.m4
-	- more minor fixes from Andreas Schwab
-
-doc/Makefile.in
-	- change TEXINPUTS makefile variables to TEXINPUTDIR to avoid
-	  conflict with shell variable of the same name
-
-builtins/pushd.def
-	- fix to avoid a bad call to free after a call to
-	  polite_directory_format does not change its argument string
-
-lib/readline/bind.c
-	- fixes to _rl_macro_dumper_internal so that it prints whatever
-	  prefix it's passed, if any
-
-				   2/29
-				   ----
-Makefile.in
-	- slight change to the rule for `stamp-h': it should be created by
-	  running `config.status', not explicitly by the makefile rule
-
-builtins/Makefile.in
-	- replace `..' in the dependencies with `$(topdir)'
-	- replace `.' in the dependencies with `$(srcdir)'
-
-				    3/1
-				    ---
-Makefile.in
-	- add a rule to build builtins/builtext.h for the benefit of
-	  deficient makes like the SunOS one
-
-variables.c
-	- fix to assign_in_env so that values in the environment are
-	  properly null-terminated
-
-builtins/Makefile.in
-	- added dependencies for object files made from .c files in this
-	  directory: common.o, getopt.o, bashgetopt.o.  SunOS /bin/make
-	  doesn't seem to be able to handle anything else
-
-support/mkclone
-	- new script to replace clone-bash that works from MANIFEST to link
-	  only those files contained in a distribution
-
-support/mkversion.c
-	- change so that it doesn't try to get `.build' from the source
-	  directory when that's different from the build directory
-
-				    3/4
-				    ---
-bashjmp.h
-	- #undef setjmp and longjmp before redefining them as sigsetjmp and
-	  siglongjmp, respectively
-
-bashhist.c
-	- fixed an uninitialized variable problem in expand_histignore_pattern
-
-builtins/set.def
-	- used `on_or_off' where `value' was needed in minus_o_option_commands
-
-builtins/common.h
-	- added extern declaration for set_var_attribute
-
-print_cmd.c
-	- include `bashansi.h' instead of just string.h or strings.h
-
-builtins/*.def, builtins/common.c
-	- include `../bashansi.h' where appropriate
-
-parse.y
-	- fixed a parenthesization problem in alias_expand_token
-
-general.h
-	- added extern declaration for `ungetc_with_restart'
-
-lib/readline/readline.c
-	- renamed LibraryVersion to rl_library_version, made it extern,
-	  assigned `2.1' to it
-
-lib/readline/readline.h
-	- extern declaration for `rl_library_version'
-
-lib/readline/doc/rltech.texinfo
-	- added description of `rl_library_version'
-
-lib/glob/glob.c
-	- changed call to sprintf in glob_dir_to_array to a couple of
-	  calls to strcpy, since we keep the length of the first string
-	  we copy
-
-				    3/7
-				    ---
-aclocal.m4
-	- added new macro `BASH_FUNC_LSTAT' to check for lstat on Linux,
-	  which defines it as an inline function in <sys/stat.h>
-
-configure.in
-	- call BASH_FUNC_LSTAT if $ac_cv_func_lstat has value `no'
-
-				    3/8
-				    ---
-parse.y
-	- changed the occurrences of `list' in the if command productions
-	  to use `compound_list' instead
-
-				   3/11
+				   4/23
 				   ----
-parse.y
-	- changed the occurrences of `list' in the while and until command
-	  productions to use `compound_list'
+oslib.c
+	- fixed definition of to_lower on machines without strcasecmp
 
-lib/readline/complete.c
-	- fix to filename_completion_function -- off-by-one error when
-	  expand-tilde is enabled and a filename to be completed begins
-	  with `~/'
+trap.c
+	- don't free the trap command in change_signal if the SIG_INPROGRESS
+	  is set in the signal's flags -- it will cause memory to be freed
+	  twice if a trap command resets the signal handler more than once,
+	  and _run_trap_internal keeps a pointer to the trap command so it
+	  can free it, so there will be no leaks
 
-				   3/12
+				   4/24
 				   ----
-builtins/cd.def
-	- made the POSIX.2 behavior of PWD not containing symlinks after
-	  using $CDPATH part of `posix mode', not default shell behavior
-
-lib/readline/display.c
-	- fix to update_line to handle update problems when using
-	  horizontal scroll mode.  This is a dumb update solution -- it
-	  should use a better one
+aclocal.m4,configure.in
+	- removed BASH_CC_WORKS, since AC_PROG_CC now has the functionality
 
-				   3/14
-				   ----
-examples/functions/csh-compat
-	- replaced the defintion for `alias' with a better one posted to
-	  usenet by Mohit Aron <aron@cs.rice.edu>
+shell.c, externs.h
+	- get_current_user_info is now a void function
 
-				   3/15
-				   ----
-jobs.c
-	- fix to wait_for_background_pids to keep `wait' from hanging
+bashline.c
+	- alias_expand_line_internal was removed
+	- new function, alias_expand_line, performs alias expansion on
+	  rl_line_buffer and either replaces rl_line_buffer or signals
+	  an error
+	- new bindable commands: alias-expand-line and
+	  history-and-alias-expand-line, available if ALIAS is defined
 
-				   3/19
+				   4/25
 				   ----
-lib/readline/input.c
-	- new function `_rl_input_available()' returns > 0 if there is
-	  input available on the readline input file descriptor.  Only
-	  works if select(2) or FIONREAD are available
+Makefile.in, lib/malloc/malloc.c
+	- changed the define that turns on malloc range checking from
+	  `rcheck' to `RCHECK'
 
 lib/readline/isearch.c
-	- slight change to the isearch termination behavior -- ESC still
-	  terminates the search, but if there is pending input or if input
-	  arrives within 0.1 seconds (on systems with select(2)) it is
-	  used as a prefix character with rl_execute_next
-
-shell.c
-	- the GNU coding standards say to write the output generated by
-	  the --help command line option to stdout, not stderr
-	- show_shell_version now takes an `extended' option that displays
-	  copyright information if non-zero
-	- show_shell_version now prints the value of `MACHTYPE' by default
-	- the `--version' option now causes the shell to exit successfully
-	  after printing the extended version information
-
-externs.h
-	- changed prototype for show_shell_version
-
-shell.c, bashline.c, builtins/help.def
-	- changed calls to show_shell_version to add appropriate argument
-
-Makefile.in
-	- pass a `MACHTYPE' define to the compiler
-
-				   3/22
-				   ----
-general.c
-	- changed print_timeval() and print_time_in_hz() to output three
-	  fractional digits after the decimal point
-
-examples/loadables/sleep.c
-	- changed to an implementation that will sleep fractional portions
-	  of seconds if select() is available
-
-				   3/25
-				   ----
-builtins/shopt.def
-	- fixes to shopt -o from Andreas Schwab.  Use FLAG_ON/FLAG_OFF
-	  instead of SETOPT/UNSETOPT
-
-				   3/26
-				   ----
-Makefile.in, builtins/Makefile.in
-	- use `@includedir@' instead of `@incdir' for autoconf 2.9
+	- fixed a couple of places where rl_search_history would try to
+	  free a NULL pointer
 
-Makefile.in
-	- the `info', `dvi', and `ps' targets do not depend on `texindex'
-	- add a `dist' target that just prints a message describing how
-	  distributions are constructed
-
-support/mkdirs
-	- replace uses of [...] with `test'
-
-				   3/28
+				   4/29
 				   ----
-parse.y
-	- fix for a problem with \@ prompt expansion from Tim Mooney
-
-jobs.c
-	- broke the code that prints a pipeline out into a separate
-	  function: print_pipeline().  This gets called by
-	  pretty_print_job and can be used for debugging
-	- two new functions to save and restore the_pipeline around calls
-	  to make_child that you don't want to disturb the current pipeline,
-	  for example in process substitution
-
-subst.c
-	- changed process_substitute to call save_pipeline and
-	  restore_pipeline in the appropriate places
+unwind_prot.c
+	- fixed a problem with saving a variable that is a null pointer
+	  in unwind_protect_var.  It happens only on machines where the
+	  size of a pointer is not the size of an int.  The old FASTCOPY
+	  code would copy the eight bytes at memory location zero, which
+	  did not necessarily make a null pointer
 
-				   3/29
+				   4/30
 				   ----
-general.c
-	- ansicstr now takes an additional parameter, the length of the
-	  string to be translated.  It's the second paramter.
-
-builtins/echo.def
-	- changed call to ansicstr to pass strlen(list->word->word)
-
-parse.y
-	- redid the $'...' expansion so it works like it's supposed to:
-	  the quoted strings may appear anywhere in a token, and multiple
-	  ansi-c quoted strings may appear in a token
-	- redid the $"..." expansion so it works like it's supposed to
-
-tests/nquote.{tests,right}, tests/run-nquote
-	- tests for the $'...' and $"..." quoting stuff -- simple-minded
-
-subst.c
-	- on systems without /dev/fd, open the named pipe for a `reading
-	  in child' process substitution (>(...)) with O_NONBLOCK
-
-lib/posixheaders/filecntl.h
-	- add code to make sure the O_NONBLOCK is defined to O_NDELAY if
-	  it is present and O_NONBLOCK is not defined by <fcntl.h>
-
-general.c
-	- don't bother handling both O_NONBLOCK and O_NDELAY in
-	  unset_nodelay_mode, since filecntl.h defines O_NONBLOCK as
-	  O_NDELAY for non-Posix systems
+shell.c
+	- run_startup_files should turn off job control, since the startup
+	  files should be run without job control enabled -- this makes
+	  SIGINT interrupt startup file execution again
+	- if we get a SIGINT or other longjmp to top_level while executing
+	  the startup files, re-enable job control for interactive shells
+	  before setting locally_skip_execution
 
-				    4/1
+				    5/2
 				    ---
-lib/readline/funmap.c
-	- made `vi-fetch-history' a bindable command name
-
-doc/readline.3
-	- many cleanups, updated the list of default bindings 
+lib/readline/nls.c
+	- if we have setlocale(3), don't bother with checking the
+	  environment variables for valid values; just use setlocale()
+	  to set the locale categories from the environment variables
+	  directly and go into eight-bit mode if the current locale is
+	  not C or POSIX
 
-				    4/4
+				    5/5
 				    ---
-doc/bash.1, doc/readline.3
-	- fixed up use of \-; now it is not used unless the text is being
-	  printed in bold or italic
+sig.c
+	- make sure that the handler for SIGPROF is not changed if it has
+	  been set to something other than SIG_IGN or SIG_DFL -- this makes
+	  profiling work after the terminating signals have been initialized
 
-configure.in, config.h.in
-	- add a configuration option, --enable-usg-echo-default, to turn
-	  on DEFAULT_ECHO_TO_USG and make `echo' expand backslash-escaped
-	  characters by default
+bashline.c
+	- if a filename containing `!' is completed, and the user has started
+	  the string with a `"', change the completion style to backslash-	
+	  quoting, since there's no way to use `!' with double quotes (this
+	  requires more changes to readline to really work right)
 
-				    4/8
+				    5/6
 				    ---
-parse.y
-	- another small change to localeexpand to handle backslash-escaped
-	  double quotes in the double-quoted string
+lib/readline/complete.c
+	- changes to make_quoted_replacement, insert_all_matches, and
+	  insert_match and their callers to allow the application-specific
+	  filename quoting function to change the quote character (e.g., for
+	  bash to change a filename containing a `!' and started with a
+	  double quote by the user into a filename with the `!' quoted by
+	  a backslas and no double quote)
 
-				    4/9
+				    5/8
 				    ---
-shell.c
-	- add the value of MACHTYPE to the text output by `--help'
-
-				   4/11
-				   ----
-parse.y
-	- more changes to the $"..." and $'...' code to move it into
-	  read_token
-	- don't try to check token[token_index - 1] unless token_index
-	  is greater than 0
-
-				   4/12
-				   ----
-trap.c
-	- new function run_trap_cleanup to clean up after _run_trap_internal
-	  in the event that parse_and_execute does not return normally
-	  (e.g., if a `return' is executed in the trap command)
-	- set running_trap to the number of the signal whose trap is being
-	  run plus one in _run_trap_internal
-
-trap.h
-	- extern declaration for run_trap_cleanup
-
 jobs.c
-	- new function `unfreeze_jobs_list' to set freeze_jobs_list back to 0.
-	  called from parse_and_execute_cleanup so a `return' while running
-	  an interrupt trap does not leave the jobs list frozen
+	- new function: nohup_all_jobs(), calls nohup_job for each entry in
+	  the jobs list
+	- delete_all_jobs is now global
 
 jobs.h
-	- new extern declaration for unfreeze_jobs_list
-
-builtins/common.c
-	- if running_trap is non-zero in parse_and_execute_cleanup, indicating
-	  that parse_and_execute was running a trap command when it got a
-	  `return', call run_trap_cleanup (running_trap - 1)
-	- parse_and_execute_cleanup now calls unfreeze_jobs_list
-
-				   4/16
-				   ----
-Makefile.in
-	- don't try to make `doc' as a dependency of `install'
-
-doc/Makefile.in
-	- make `info' a dependency of `install' to make sure the info
-	  document is present and up to date before installing it
-	- make the `install' target install the documents with a $(srcdir)/
-	  prefix in case we're building in another directory
-
-lib/glob/glob.c
-	- fixed an off-by-one error in glob_dir_to_array
-
-shell.c
-	- removed an extra increment of arg_index when setting up the
-	  arguments for -c command
-
-				   4/17
-				   ----
-pathexp.c
-	- made quote_globbing_chars backslash-quote backslashes as well as
-	  `?*[]'; changed it to be a little faster
+	- new extern declarations for delete_all_jobs() and nohup_all_jobs()
 
-bashhist.c
-	- call quote_globbing_chars to quote backslashes in the previous
-	  history line even if no other globbing chars are present.  This
-	  is done only if we're matching against a HISTIGNORE pattern of `&'
-	- don't bother quoting globbing characters in the current history
-	  line in history_should_ignore -- fnmatch ignores special chars in
-	  its `string' argument
-
-[bash-2.0-alpha2 frozen]
+builtins/jobs.def
+	- `disown' without any jobspec arguments means the current job.  Fix
+	  a core dump printing the error message when there is no current job
 
-				   4/18
+				   5/12
 				   ----
-command.h
-	- new flag value for a word: W_NOSPLIT.  A word with this bit set
-	  in its flags will not have word splitting performed
-
-parse.y
-	- turn on the W_NOSPLIT flags for assignment statements appearing
-	  where an assignment statement is acceptable (words that would
-	  return ASSIGNMENT_WORD rather than WORD).  This means that, for
-	  the time being, assignment statement arguments to builtins like
-	  `declare' or `alias' will be split unless they are quoted
-
 subst.c
-	- don't split a word in expand_word_internal if the W_NOSPLIT flag
-	  is set, rather than checking W_ASSIGNMENT; do the same thing in
-	  expand_word_list_internal
-
-builtins.h
-	- a new flag, ASSIGNMENT_BUILTIN, indicating that this builtin takes
-	  assignment statements as arguments
-	- rearranged the values of the builtin flags, so the BUILTIN_* flags
-	  come first, then the *_BUILTIN flags
+	- process an expansion like $((foo); bar) as a command substitution,
+	  not as an arithmetic expansion.  An arithmetic expansion must have
+	  a closing `))'
 
-builtins/makebuiltins.c
-	- added code for an array of `assignment builtins' -- builtins that
-	  take assignment statements as arguments -- and to add the
-	  ASSIGNMENT_BUILTIN flag for those builtins
-
-execute_cmd.c
-	- added a hack function `fix_assignment_words', which checks the
-	  first word of a builtin command to see if it is a builtin that
-	  has the ASSIGNMENT_BUILTIN flag set, and adds W_NOSPLIT to the
-	  flags for all words with the W_ASSIGNMENT bit set.  This means
-	  that word splitting is not done for any of the assignment
-	  statements in commands like `declare z=$a'
-
-				   4/19
+				   5/14
 				   ----
-execute_cmd.c
-	- fixed a memory-freed-twice error in find_in_path_element
-
-lib/readline/rltty.c
-	- include <sys/ioctl.h> if GWINSZ_IN_SYS_IOCTL is defined and
-	  SHELL is not defined
-
-lib/readline/input.c
-	- changed a stray HAVE_FIONREAD to FIONREAD
-	- include <sys/ioctl.h> if FIONREAD_IN_SYS_IOCTL is defined
-
-aclocal.m4
-	- new macro, BASH_HAVE_FIONREAD, to check for a #define of
-	  FIONREAD in <sys/ioctl.h> or one of the files it includes;
-	  defines FIONREAD_IN_SYS_IOCTL if present
-
-config.h.in
-	- new line for FIONREAD_IN_SYS_IOCTL
-
-configure.in
-	- call BASH_HAVE_FIONREAD
-
-				   4/22
-				   ----
-builtins/cd.def
-	- fixed a memory-freed-twice error in mkpath()
-
-jobs.c
-	- don't print a job termination message for a non-interactive shell
-	  with startup_state == 2 (those shells are started to run commands
-	  when bash is invoked with `-c command')
-
-				   4/23
-				   ----
-general.h
-	- definition for a `generic pointer' type PTR_T -- `void *' on ANSI
-	  C systems, `char *' otherwise
-
-jobs.h
-	- new members of job struct: `j_cleanup', a function to call when the
-	  job is marked JDEAD, and `cleanarg', argument to pass to j_cleanup
-
-jobs.c
-	- initialize j_cleanup and cleanarg to NULL in stop_pipeline
-	- call j_cleanup from waitchld() when a job is marked JDEAD
-
-test.c
-	- fixed binop() so it recognizes `<' and `>' as binary operators
-	- fixed a bug in binary_operator so `<' works correctly
-
-tests/test-tests
-	- fixed the tests so they no longer rely on the modes of files in
-	  the file system to test -u, -g, -r, -x, -w, and so on, except
-	  for a few `standard' files like /dev/tty and /dev/null
-	- added tests for string < string and string > string
-
-xmalloc.c
-	- include <unistd.h> if HAVE_UNISTD_H is defined for a prototype
-	  for sbrk
-	- add an extern declaration for sbrk if SBRK_DECLARED is not defined
-	- use PTR_T to cast the return value of sbrk() when finding out where
-	  the break is and how many bytes of memory have been allocated
-
-lib/malloc/malloc.c
-	- added code to write 0xcf into memory as it's freed, to uncover
-	  callers that try to refer to freed memory, and writes 0xdf into
-	  newly-allocated memory, to uncover callers that assume something
-	  about new allocations (e.g., that newmem[0] == 0)
-
-lib/malloc/gmalloc.c
-	- latest version from GNU sources
-
-				   4/25
-				   ----
-doc/bash.1
-	- changed the description of `unset' to include a description of
-	  the `-v' option
-
-parse.y
-	- fixed a problem with conversion to 12-hour time in
-	  decode_prompt_string that made 12:00pm show up as 00:00pm
-
-				   4/26
-				   ----
-builtins/common.c, subst.c
-	- a couple of fixes from Andreas Schwab
-
-configure.in
-	- set up the minimal configuration after checking for
-	  --enable-minimal-config with AC_ARG_ENABLE, before checking any
-	  of the other options, instead of after checking all options.
-	  This makes `--enable-minimal-config --enable-prompt-string-decoding'
-	  work as documented
-	- changed AC_PREREQ to require autoconf version 2.8 or higher
-	- added some code to disable gnu malloc by default on the systems
-	  listed in NOTES
-	- set MALLOC_SRC to have the $(ALLOC_LIBSRC)/ prefix to avoid having
-	  a directory in the dependencies when making without the gnu
-	  malloc
-
-lib/malloc/Makefile.in
-	- make gmalloc.o depend on $(BUILD_DIR)/config.h
-
-				   4/30
-				   ----
-COMPAT
-	- new file listing user-visible incompatibilites between bash-1.14
-	  and bash-2.0
-
-parse.y
-	- new prompt expansions, \v (version) and \V (version + patchlevel)
-
-config.h.top
-	- the default value of PS1 is now '\s-\v\$ '
-
-doc/{bash.{1,html},features.texi}
-	- added descriptions of \v and \V prompt expansions
-	- changed default value of PS1
-
-Makefile.in
-	- added dependencies on config.h.top for files that use definitions
-	  included there
-
-builtins/exec.def
-	- removed reference to `no_exit_on_failed_exec' from the help text
-
-				    5/3
-				    ---
-builtins/cd.def
-	- fixed a typo in fix from 4/22
-
-				    5/7
-				    ---
-builtins/common.c
-	- new function `no_options (WORD_LIST *)' to be called by builtins
-	  that do not take options to check for options and -?.  It returns
-	  0 on success, non-zero if an unwanted option is supplied
+builtins/evalstring.c
+	- the third argument to parse_and_execute() is now a flags word.
+	  The caller can control the value of `interactive' and whether
+	  or not history is disabled while parse_and_execute() runs
 
 builtins/common.h
-	- extern declaration for no_options
-
-builtins/return.def
-	- corrected the error message to include returning from a sourced
-	  script
-
-builtins/{getopts,eval,wait,source,fg_bg,help}.def
-	- changed the following builtins to either call no_options or test
-	  explicitly for options and return EX_USAGE if any are found:
-
-		getopts eval wait source . fg bg help
-
-builtins/help.def
-	- error messages are now printed using builtin_error
-
-				    5/8
-				    ---
-builtins/jobs.def
-	- fixed an unitialized variable problem
-
-tests/more-exp.tests
-	- added additional tests for IFS problem uncovered by `iffe'
-
-				    5/9
-				    ---
-lib/malloc/Makefile.in
-	- use $(ALLOCA_SOURCE) instead of $< in rule that builds alloca.o.
-	  Some makes don't expand the $< in non-suffix rules
-
-lib/readline/rldefs.h
-	- fix workaround for SVR4.2 bug
-
-shell.c
-	- even though SVR4.2 has getpeername(), isnetconn() should not use it
-
-maxpath.h
-	- rework to make sure PATH_MAX and NAME_MAX are defined, and remove
-	  the BUILDING_MAKEFILE code
-
-general.c, parse.y, oslib.c, jobs.c, builtins/common.c, builtins/cd.def
-	- change uses of MAXPATHLEN to PATH_MAX
-
-examples/loadables/{{log,base,dir}name,tty,pathchk,tee}.c
-	- new loadable builtins:
-
-		logname basename dirname tty pathchk tee
-
-configuure.in, config.h.in
-	- look for tzset(3), define HAVE_TZSET if found
-
-subst.c
-	- new special variable function, sv_tz, which calls tzset when
-	  TZ is changed, if tzset(3) exists and the shell is compiled
-	  to do prompt string decoding
-
-subst.h
-	- extern declaration for sv_tz
-
-				   5/10
-				   ----
-builtins/echo.def
-	- only call printf if the string to print is non-null
-	- add an fflush(stdout) after the printf call to work around a
-	  bug in SunOS 5.5
-
-				   5/16
-				   ----
-shell.c
-	- moved some variable declarations out of this file to more
-	  logical places
-
-support/config.guess
-	- small fixes from rfg@monkeys.com
-
-version.c
-	- moved functions for getting, setting, and displaying shell version
-	  information here from shell.c
-
-general.c
-	- moved set_lines_and_columns to variables.c
-	- moved getc_with_restart and ungetc_with_restart to input.c
-	- new function argv_to_word_list(), converts an array of strings
-	  into a WORD_LIST
-	- renamed find_name_in_list to find_name_in_array to match rest
-	  of functions that operate on arrays of strings
+	- new #defines for the flag values for parse_and_execute()
 
-configure.in, Makefile.in, jobs.c, nojobs.c
-	- changes so that jobs.c no longer includes nojobs.c, and the correct
-	  object file (jobs.o or nojobs.o) is selected by configure
-	  and substituted into the Makefile
+{bashline,jobs,shell,subst,trap,variables}.c, parse.y, builtins/evalfile.c,
+builtins/{eval,fc}.def
+	- changed calls to parse_and_execute appropriately
 
-list.c
-	- new file, list manipulation functions from general.c
-
-externs.h, general.h
-	- moved extern declarations for functions defined in oslib.c and
-	  list.c to externs.h from general.h
-
-				   5/17
-				   ----
-locale.c
-	- new file, with locale code from parse.y and shell.c
-
-shell.c
-	- don't turn off job control if act_like_sh is set
-	- if an unknown option is supplied at startup or if -c is
-	  supplied without an argument, exit with EX_USAGE
-	- call posix_initialize after parsing all the options, to
-	  catch `bash -o posix'
-	- new functions: init_interactive(), init_noninteractive()
-	- exit with EX_NOTFOUND if a script argument is not found
-	- exit with EX_NOINPUT if a script file cannot be opened for
-	  some reason
-
-shell.h
-	- new define for EX_NOINPUT exit status (126)
-
-				   5/18
-				   ----
-lib/readline/bind.c
-	- fixed _rl_get_keyname() so that it properly handles C-\ and outputs
-	  it as \C-\\ and C-" as \C-\".  This fixes the improper binding
-	  commands written by `bind -p' that caused weird things to happen
-	  when people used the output of `bind -p' as a start for their
-	  own inputrc files
-	- new function _rl_init_file_error to print error messages encountered
-	  while parsing the inputrc file
-	- print out an error message if parsing an inputrc line and no closing
-	  double quote is found for a key binding
-
-lib/readline/complete.c
-	- in rl_complete_internal, if the completion function results in
-	  multiple matches, but none match up to even the first character, 
-	  use what the user typed in (which is presumably a glob pattern
-	  that expanded into multiple files) as matches[0].  This makes
-	  things like show-all-if-ambiguous work right, too
-
-oslib.c
-	- if we are providing our own version of getenv(), provide a function
-	  _getenv() as well, which just calls getenv()
-
-builtins/ulimit.def
-	- #define _KERNEL before including <sys/resource.h> if HPUX is
-	  defined.  This makes the full set of limits available on hpux
-	  version 8 and above
+builtins/evalfile.c
+	- if _evalfile() is passed FEVAL_HISTORY as part of the flags arg,
+	  don't pass SEVAL_NOHIST to parse_and_execute
+	- new function: fc_execute_file, which sets FEVAL_HISTORY in the
+	  flags argument to _evalfile()
 
 bashline.c
-	- new function, enable_hostname_completion, sets up readline to
-	  perform or not perform hostname completion.  Hostname completion
-	  is on by default.
-	- don't attempt hostname completion in attempt_shell_completion()
-	  unless perform_hostname_completion is set
-
-builtins/shopt.def
-	- add a `set_func' member to the shopt options structure, to provide
-	  a hook for those variables that require a little more than just
-	  toggling a variable on or off
-	- new shopt variable `hostcomplete', turns hostname completion on and
-	  off
-
-doc/{bash.{1,html},features.texi}
-	- added description of new shopt `hostcomplete' variable
-
-				   5/21
-				   ----
-sig.c
-	- don't try to save the shell history in termination_unwind_protect
-	  if we're dying due to SIGABRT
+	- call bash_add_history instead of add_history from
+	  vi_edit_and_execute_command so the bash state variables get
+	  updated properly.  This keeps the `v' command from operating
+	  on an empty command when the history list is stifled
 
 bashhist.c
-	- new functions: last_history_entry (static), last_history_line
+	- bash_add_history is now global
 
 bashhist.h
-	- extern declarations for new functions in bashhist.c
-
-error.c
-	- have programming_error() report the last command in the history
-	  before aborting
+	- extern declaration for bash_add_history
 
 builtins/fc.def
-	- if the editor returns a non-zero exit status when using `fc -e',
-	  return immediately without trying to execute the commands, as
-	  per Posix.2, 5.12.2
-	- when using `fc -s', echo the command to be executed to stderr,
-	  not stdout
-
-lib/readline/terminal.c
-	- new file, with all code related to termcap/terminfo
-
-lib/readline/{readline,rltty,display,util}.c
-	- moved functions dealing with termcap to terminal.c
-
-lib/readline/readline.c
-	- readline_initialize_everything now calls _rl_enable_meta_key()
-	  to turn on the meta key, based on the value of _rl_enable_meta
-	  (on by default)
-
-lib/readline/rltty.c
-	- do not enable and disable the meta key each time readline is
-	  called; do it once at initialization
-
-				   5/23
-				   ----
-bashhist.c
-	- when `hist_verify' is set, call re_edit with the expanded line,
-	  not the original one, and do not print the results of the
-	  expansion before re-editing
-
-				   5/24
-				   ----
-support/config.guess
-	- recognize linux/sparc
-
-configure.in
-	- don't use GNU malloc on linux/sparc
-
-variables.c
-	- initialize a new $MACHTYPE variable to the value of MACHTYPE as
-	  set by autoconf
-
-doc/{bash.{1,html},features.texi}
-	- documented $MACHTYPE
-
-				   5/30
-				   ----
-builtins/ulimit.def
-	- allow the max vm size to be set via setrlimit() if RLIMIT_VMEM
-	  is defined
-
-				   5/31
-				   ----
-shell.h
-	- changed EX_USAGE to 258, EX_BADUSAGE is now 2
-	- added a number of execution failure statuses, so that builtins
-	  can indicate various failures.  All are greater than 256, so
-	  they cannot be returned by other utilities
-
-execute_cmd.c
-	- new function, builtin_status(), to translate the new return
-	  status codes to something the shell can export
-	- changed execute_simple_command to call builtin_status after
-	  execute_builtin_or_function to translate error codes if a
-	  builtin is invoked.  If a function or regular builtin is
-	  invoked, all return values greater than EX_SHERRBASE get
-	  translated to EXECUTION_FAILURE, with the exception that
-	  EX_USAGE gets translated to EX_BADUSAGE.  If a special
-	  builtin failed with a status > EX_SHERRBASE, special_builtin_failed
-	  is set to cause the shell to exit in POSIX mode
-	- changed execute_builtin_or_function to return EX_REDIRFAIL if
-	  redirections fail
-	- changed execute_subshell_builtin_or_function to translate
-	  EX_USAGE to EX_BADUSAGE if a builtin is executed
-	- set this_command_name to NULL in execute_for_command before binding
-	  each member of the word list to the loop variable, to avoid
-	  garbled error messages if the variable has the integer attribute
-
-builtins/{set,unset}.def
-	- return EX_USAGE instead of EXECUTION_FAILURE if a bad option is
-	  supplied
-
-builtins/source.def
-	- return EX_USAGE instead of EXECUTION_FAILURE if the required
-	  filename argument is missing
-
-builtins/setattr.def
-	- return EX_BADASSIGN if any assignment statements given as arguments
-	  to readonly, export, etc. return failures (assignment to
-	  non-identifier, assignment to readonly variable)
-
-builtins/declare.def
-	- don't allow `declare var=value' to assign value to a readonly
-	  variable
-
-lib/readline/rldefs.h
-	- include <stdarg.h> if we're going to be using stdarg instead of
-	  varargs
-
-lib/readline/display.c
-	- added `stdarg' version of rl_message
-
-lib/readline/readline.h
-	- changed extern declaration for rl_message to be in ANSI-C format
-	  if __STDC__ is defined and we're using stdarg
-
-variables.c
-	- make sure this_command_name is set to null before calling
-	  make_variable_value in assign_array_var_from_string so any
-	  error messages are not garbled
-
-				    6/2
-				    ---
-builtins/common.c
-	- don't execute the command in parse_and_execute if -n has been
-	  enabled and the shell is not interactive
-	- include flags.h for read_but_dont_execute
-
-[bash-2.0-alpha3 released]
-
-				    6/5
-				    ---
-subst.c
-	- array_length_reference and parameter_brace_expand_length now
-	  return -1 on errors
-	- parameter_brace_expand now returns &expand_param_error if
-	  paramter_brace_expand_length returns something < 0
-
-variables.c
-	- assign_in_env now checks for assignments to readonly shell
-	  variables and disallows them
-
-				    6/6
-				    ---
-xmalloc.c
-	- only do pointer arithmetic on pointers cast to (char *), since
-	  it's not required that compilers support arithmetic on void *
-
-shell.c
-	- make sure <sys/socket.h> is included with the same conditions
-	  used to call getpeername(2) in isnetconn()
+	- call fc_execute_file instead of maybe_execute_file in the
+	  edit-and-re-execute case (fc -e ...)
+	- don't manually insert the commands from the file created by `fc -e'
+	  into the history list, just set remember_on_history and let
+	  fc_execute_file take care of telling parse_and_execute to do the
+	  right thing.  This makes compound commands and the `cmdhist'
+	  and `lithist' settings work better.  This supersedes the fix of
+	  1/27.  This was reported again by rchen@fractal.eng.yale.edu.
 
 parse.y
-	- added a new function, paren_match, to parse the contents of $(...)
-	  constructs.  This function correctly handles embedded quoted
-	  strings, embedded command substitutions, embedded command
-	  substitutions with embedded quoted strings, etc.  The implementation
-	  is much closer to that described in POSIX.2, section 3.6.3.
-	- changed read_token_word to call paren_match where appropriate
-
-subst.c
-	- rewrote extract_delimited_string so that it correctly understands
-	  quoted strings inside command substitutions, a la paren_match
-	  in parse.y.  Corrected the description of the function's output
-	  in the code.  extract_delimited_string is now very recursive, and
-	  does not try to do everything in one pass without recursion
-
-				    6/7
-				    ---
-test.c
-	- fixed three_arguments() to test the second argument for a valid
-	  binary operator before checking whether or not the first argument
-	  is `!'.  This is what POSIX.2 says to do.
-
-lib/malloc/malloc.c
-	- don't compile in the memory scrambling code (#define MEMSCRAMBLE)
-	  if NO_MEMSCRAMBLE is defined.
+	- the body of a `for' command (the commands between do...done or
+	  {...}) should be a `compound_list' instead of a `list'.  Problem
+	  reported by cpg@research.bell-labs.com
 
-configure.in
-	- define NO_MEMSCRAMBLE in LOCAL_CFLAGS for SCO 3.2v[45]
-
-				    6/9
-				    ---
-parse.y
-	- renamed paren_match to parse_matched_pair
-	- call parse_matched_pair to parse <(...) and >(...) constructs,
-	  since those should be treated the same as $(...) for the purposes
-	  of matching parens and skipping embedded quoted strings
-
-				   6/13
+				   5/19
 				   ----
-parse.y
-	- augmented parse_matched_pair
-		o fixed up backslash quoting and CTLESC handling after
-		  backslashes
-		o match pairs of ${...} and $[...] inside double-quoted
-		  strings or `` strings
-		o keep track of line number quoted string starts on for
-		  better error messages
-		o push and pop delimiters when parsing quoted strings
-		  inside $(...), ${...}, and $[...] for use by the
-		  history code
-		o set EOF_Reached to 1 if we get an EOF before finding the
-		  match closer
-	- rewrote read_token_word from scratch to use parse_matched_pair:
-		o shell quote characters use parse_matched_pair to find
-		  the matching quote
-		o fixed up backslash parsing code
-		o compound array assignments use parse_matched_pair
-		o corrected all_digits and dollar_present computations, so,
-		  for instance, 2''>/dev/null is no longer accepted as
-		  identical to 2>/dev/null
-		o converted ${...} parsing to use parse_matched_pair
-		o removed printing of error messages about unmatched quotes;
-		  now handled by parse_matched_pair
-		o converted $'...' and $"..." to use parse_matched_pair
-		o removed all of the delimited_xxx and dollar_xxx variables
-		  and embedded_quoted_string
-	- modified report_syntax_error to regularize error messages:
-		o non-interactive shells and interactive shells running
-		  scripts with `.' now always report line numbers
-		o unexpected EOF is always reported when EOF_Reached is true
-		o non-interactive shells no longer duplicate the script name
-		  in error messages (e.g., `./z3: ./z3: line 3: syntax error')
-
-bashline.c
-	- make sure `newnames' is freed in _ignore_completion_names
-
 lib/readline/complete.c
-	- don't insert the first match on TAB completion in
-	  rl_complete_internal() if matches[0] is the empty string --
-	  that will overwrite whatever the user has typed, if dequoting
-	  what the user typed results in the empty string
+	- in filename_completion_function, if we find that the directory
+	  pointer (return value from opendir(3)), is not null when state
+	  is 0 (indicating that this is the first time the completion
+	  function has been called for the current completion), call
+	  closedir on it, assuming that it was left open by a previous
+	  (interrupted) completion
 
-execute_cmd.c
-	- include <limits.h> if HAVE_LIMITS_H is defined for ARG_MAX
-	- only put the special environment variable for GNU getopt into
-	  the environment if ARG_MAX exceeds 10240
+[bash-2.01-beta2 frozen]
 
-subst.c
-	- rewrote extract_dollar_brace_string so that it fully obeys
-	  the POSIX.2 rules for finding the closing `}' and uses
-	  skip_{single,double}_quoted rather than trying to do it with
-	  inline delimiters
-	- removed the INC_NEST and DEC_NEST defines
-
-				   6/14
-				   ----
-jobs.c
-	- don't print status messages for background jobs that complete
-	  successfully when running shell scripts
-
-builtins/shopt.def
-	- when using shopt -o as a synonym for set -o, make sure $SHELLOPTS
-	  is updated
-	- make sure to update $SHELLOPTS when using shopt to set or unset
-	  interactive comments
-
-execute_cmd.c
-	- fixed find_in_path_element to not return non-executable files
-	  when FS_EXEC_PREFERRED is one of the flags, after saving such
-	  a file as file_to_lose_on (fix from william@nscs.fast.net)
-
-				   6/17
-				   ----
-lib/readline/complete.c
-	- after printing possible completions, all lines of a multi-line
-	  prompt are redisplayed using rl_forced_update_display()
-
-subst.c
-	- in expand_word_internal, only do tilde expansion on words where
-	  (flags & (W_ASSIGNMENT|W_QUOTED)) == W_ASSIGNMENT (unquoted
-	  assignment statements) and contain unquoted `=~' or `:~', and
-	  only if posixly_correct is off.  POSIX.2 says that only the
-	  assignments preceding the command name should be tilde-expanded.
-
-				   6/18
-				   ----
-support/config.guess
-	- recognize the new Pyramid DC-OSx as `mips-pyramid-sysv4'
-
-configure.in
-	- set LOCAL_CFLAGS to -Xa for mips-pyramid-sysv4 as per
-	  Peter Chubb
-
-subst.c
-	- variable assignment errors when there is not a command name
-	  after the assignments now causes a non-interactive shell to 
-	  exit in POSIX.2 mode
-
-execute_cmd.c
-	- if the iteration variable in a for statement is a read-only
-	  variable, a variable assignment error occurs.  This causes
-	  non-interactive shells to exit in posix mode, and a failure
-	  status to be returned for other non-interactive shells and
-	  interactive shells
-	- if the selection variable in a select statement is a read-only
-	  variable, a variable assignment error occurs, with the same
-	  consequences as a for statement variable assignment error
-
-doc/bashref.*
-	- renamed features.* to bashref.*, since that more accurately
-	  reflects the current contents
-
-				   6/19
-				   ----
-test.c
-	- changed three_arguments() to do one-argument tests on $1 and
-	  $3 and return the appropriate values if -a or -o is given as
-	  the second argument
-	- changed three_arguments() to perform a one-argument test on
-	  $2 if $1 == '(' and $3 == ')' and return that result
-	- make sure test_stat translates /dev/fd/xx to DEV_FD_PREFIX/xx
-	  so /dev/fd/0 always means the same thing, even on linux,
-	  which uses /proc/self/fd
-
-aclocal.m4
-	- added a check for libncurses in BASH_CHECK_LIB_TERMCAP
-
-configure.in
-	- added an initial message saying we're configuring for bash-2.0
-
-lib/glob/fnmatch.c
-	- made a small change to fnmatch() to fix a bug matching patterns
-	  with multiple consecutive `*'s
-
-				   6/20
-				   ----
-parse.y
-	- inhibit history expansion when calling pre_process_line() from
-	  shell_getc() if the current delimiter is a single quote, even
-	  if we're on a different line than the opening quote
-	- make sure read_a_line prints a prompt if we're interactive and
-	  not using readline, and make sure we call clearerr() if we
-	  get EOF when interactive and not using readline
-
-bashhist.c
-	- new function history_expansion_p(char *) returns 1 if the string
-	  passed contains the history expansion or history substitution
-	  characters
-	- pre_process_line() now calls history_expansion_p and calls
-	  the history expansion code only if that returns 1
-
-				   6/26
-				   ----
-lib/readline/isearch.c
-	- swap the behavior of ^J and ^M when i-searching.  ^J now terminates
-	  the search without accepting the line.  ^M terminates the search
-	  and accepts the line, executing the command
-
-examples/loadables/sprintf.c
-	- new `sprintf' builtin: sprintf var format [args...]
-
-lib/readline/readline.c
-	- added a line in _rl_dispatch that sets _rl_suppress_redisplay if
-	  the function is rl_insert and there is input available
-
-				   6/27
-				   ----
-builtins/getopts.def
-	- new function, getopts_bind_variable(), used when assigning a value
-	  to the user-supplied variable passed to getopts().  It makes sure
-	  that that variable name is legal and prints an error message if it
-	  is not
-
-builtins/getopt.c
-	- don't increment sh_optind until after we've checked whether or not
-	  the current option character appears in the list of options passed
-	  to getopts by the user
-
-				   6/28
-				   ----
-variables.c
-	- added a new parameter to print_var_value: QUOTE.  If it's non-zero
-	  and the variable's value contains shell metacharacters, quote the
-	  value so it can be read back in
-
-variables.h
-	- changed extern declaration for print_var_value
-
-siglist.h
-	- if SYS_SIGLIST_DECLARED is not defined, but HAVE_UNDER_SYS_SIGLIST
-	  is, only declare sys_siglist as extern char *sys_siglist[] if
-	  sys_siglist is not `#define'd
-
-parse.y
-	- make sure a reserved word is acceptable before returning `}' as
-	  a token from special_case_tokens, even if the token read so far
-	  is "}"
-	- make reset_parser set allow_open_brace to 0
-
-				    7/1
-				    ---
-test.c
-	- fixed binop to not try to read s[3] if s[2] == '\0'.  There are
-	  no one-character binary operators that begin with `-'
-
-variables.c
-	- fixed kill_all_local_variables to do nothing if variable_context
-	  is >= local_variable_stack_size (indicating that we do have some
-	  local variables, but not at this level of function nesting, and
-	  we are beyond the end of the array allocated the last time we
-	  had local variables)
-
-				    7/2
-				    ---
-execute_cmd.c
-	- non-interactive shells in posix mode should exit if the variable
-	  in a for command is not a valid identifier.  This should really
-	  be caught by the parser
-	- non-interactive shells in posix mode should exit if a function
-	  name is not a valid identifier.  This should also be caught by
-	  the parser
-
-				    7/3
-				    ---
-jobs.c
-	- non-interactive shells should ignore stopped children, so
-	  waitchld() should set waitpid_flags to WUNTRACED only if the
-	  shell is interactive and it's not a subshell environment
-
-lib/readline/signals.c
-	- if not being compiled as part of the shell, readline should
-	  catch SIGTERM and clean up
-
-shell.c
-	- change to isnetconn: on Solaris 2.5, getpeername() can return
-	  EINVAL rather than ENOTSOCK if the fd passed as the first
-	  argument is not a socket
-
-				    7/5
-				    ---
-lib/readline/complete.c
-	- new variable, available to library users, to inhibit completion
-	  and cause the completion character to be inserted into the
-	  line with self-insert: rl_inhibit_completion.  Set to 0 by default.
-
-lib/readline/readline.h
-	- extern declaration for rl_inhibit_completion
-
-lib/readline/doc/rltech.texinfo
-	- documented rl_inhibit_completion
-
-lib/readline/bind.c
-	- new readline user variable: disable-completion.  This provides
-	  users a way to toggle the value of rl_inhibit_completion
-
-doc/{bash.{1,html},readline.3}, lib/readline/doc/rluser.texinfo
-	- documented disable-completion variable
-
-lib/readline/doc/Makefile
-	- added `html' target to create html from rlman.texinfo and
-	  hist.texinfo
-
-				    7/8
-				    ---
-builtins/cd.def
-	- only try cd spelling correction if the shell is currently
-	  interactive, regardless of value of `cdspell' option
-
-lib/posixheaders/alloca.h
-	- AIX 4.2 needs `#pragma alloca' if gcc is not being used
-
-				    7/9
-				    ---
-builtins/getopts.def
-	- return EXECUTION_FAILURE from getopts if it's trying to assign
-	  to a read-only variable
-	- reset sh_badopt to zero in getopts_reset()
-
-builtins/getopt.c
-	- new variable, sh_badopt, set to 1 if we find an illegal option
-	- do the increment of sh_optind if sh_badopt is non-zero and we're
-	  at the end of the option string at the beginning of sh_getopt().
-	  The increment is deferred until the next call so $OPTIND is
-	  correct
-
-builtins/getopt.h
-	- extern declaration of sh_badopt
-
-				   7/10
+				   5/27
 				   ----
 Makefile.in
-	- don't pass ALLOCA_SOURCE or ALLOCA_OBJECT to the make in lib/malloc
-	- don't make libmalloc.a depend on ALLOCA_DEP -- it causes some
-	  compilers to build alloca.o in the top directory, and if ALLOCA
-	  is not defined, makes the malloc library depend on a directory
-
-aclocal.m4
-	- some small changes to make the output look better
+	- make sure that `make distclean' (and other clean targets) remove
+	  the `printenv' executable
 
-memalloc.h
-	- made some small changes suggested by the autoconf documentation
+tests/execscript, tests/redir.tests
+	- make sure to set LANG=C and LC_ALL=C so the messages show up in
+	  English
 
-				   7/11
-				   ----
-parse.y
-	- swapped meanings of \h and \H in prompt_string_decode for
-	  bash-1.14.6 compatibility
-
-doc/{bash.{1,html},bashref.texi}
-	- updated prompt string decoding section with swap of \h and \H
-
-general.c
-	- print_timeval and print_time_in_hz now accept a FILE * as the
-	  first argument, so `times' can print to stdout while `time'
-	  prints to stderr
-
-execute_cmd.c
-	- changed time_command to print the timing statistics to stderr
-
-builtins/times.def
-	- added the FILE * argument to the calls to print_timeval and
-	  print_time_in_hz
-
-support/texi2html
-	- upgraded to version 1.50
-
-error.c
-	- include bashhist.h if HISTORY is defined for declarations of
-	  last_history_line() and remember_on_history
-
-lib/glob/glob.c
-	- add a `#pragma alloca' as the first thing in the file if we're
-	  not using gcc and _AIX is defined
-
-lib/posixheaders/memalloc.h
-	- remove the `#pragma alloca' stuff; it is done in the C source
-	  files now, to avoid problems with other AIX header file
-	  declarations
-
-				   7/12
-				   ----
-lib/readline/rltty.c
-	- some changes to work around AIX 4.2 bugs (sometimes OPOST is unset
-	  in termios, sometimes FLUSHO is inexplicably set in termios, etc.)
-
-				   7/15
-				   ----
-lib/readline/rltty.c
-	- make SETATTR use TCSADRAIN on POSIX machines.  This setting is
-	  supposed to be used for changes that affect output
-
-[bash-2.0-alpha4 released]
+tests/run-func
+	- add a warning about exported functions in the environment
 
-				   7/16
+				   5/29
 				   ----
-trap.c
-	- changed reset_or_restore_signal_handlers to honor traps before
-	  special signals, so that `trap "" 2' works right and causes
-	  children to ignore SIGINT
-
-locale.c
-	- added lc_all variable to track the value of LC_ALL
-	- added new functions: set_default_locale_vars, set_locale_var,
-	  set_lang, get_locale_var, removed get_current_messages_locale
-
-subst.h
-	- added extern declaration for sv_locale
-
-subst.c
-	- LC_ALL, LC_CTYPE, LC_MESSAGES, and LANG are now special variables;
-	  sv_locale() is called when they are assigned a value
-
-shell.c
-	- call set_default_locale_vars after initializing the shell variables
-	  in main(), so LC_CTYPE and LC_MESSAGES (if present) have values
-	- shell.c does not need to include <locale.h>
-
-general.c
-	- changed print_timeval and print_time_in_hz to format their output
-	  as Posix.2 specifies for `time -p' if the new POSIX_TIME argument
-	  is non-zero
-
-parse.y
-	- new grammar production: timespec, to handle `time' and `time -p'
-	- change to special_case_tokens so that -p after `time' returns
-	  TIMEOPT
-
-execute_cmd.c
-	- set posix_time in time_command if the command's flags include
-	  CMD_TIME_POSIX
-	- call print_timeval and print_time_in_hz with posix_time as
-	  the new third argument
-	- use a space to separate the word (`real') and the time if we're
-	  in posix mode.  This is what POSIX.2 defines for `time -p'
-
-builtins/times.def
-	- call print_timeval and print_time_in_hz with 0 as the new third
-	  argument
-
-command.h
-	- new flag: CMD_TIME_POSIX, used to handle `time -p'
-
-doc/{bash.{1,html},bashref.texi}
-	- documented `time -p'
-
-				   7/17
-				   ----
-support/mkversion.c
-	- include `config.h' before any other include files
-
-parse.y
-	- initialize was_dollar to 0 in parse_matched_pair
-
-configure.in
-	- removed duplicate check for <sys/file.h> from call to
-	  AC_HAVE_HEADERS
-
-sig.c
-	- include siglist.h so we can find out if it defines HAVE_SYS_SIGLIST
-
-siglist.h
-	- define HAVE_SYS_SIGLIST if we end up defining sys_siglist as
-	  _sys_siglist (Solaris 2.[45])
-
-siglist.c
-	- include "siglist.h" before testing whether or not HAVE_SYS_SIGLIST
-	  is defined; it may be defined in there (Solaris 2.[45])
-
-				   7/19
-				   ----
-examples/misc/alias-conv.sh
-	- updated, now uses code from examples/functions/csh-compat to do
-	  the alias conversion, uses `command' builtin in functions it
-	  creates, does some substitution of csh special variables to
-	  bash equivalents
-
-examples/misc/cshtobash
-	- a more ambitious script that attempts to convert csh aliases,
-	  environment variables, and local variables to bash equivalents
-
-				   7/22
-				   ----
-bashline.c
-	- added `:' to the set of completion word break characters so that
-	  individual directories in $PATH assignments can be completed
-
-aclocal.m4
-	- added a new macro, BASH_CC_WORKS, that aborts configuration if
-	  ${CC} can't compile a simple program successfully
-
-configure.in
-	- call BASH_CC_WORKS before doing anything else
-
-				   7/23
-				   ----
-execute_cmd.c
-	- fixed select_query to print the prompt ($PS3) to stderr
-
-support/bashbug.sh
-	- workaround for bug in SunOS 5.x /bin/sh that causes it to
-	  not ignore interrupts while waiting for a foreground process
-	  to exit
-
-				   7/24
-				   ----
-general.c
-	- added new functions to decompose `timeval's and `clock_t's into
-	  seconds and thousandths of seconds:  timeval_to_secs and
-	  clock_t_to_secs
-	- changed print_timeval and print_time_in_hz to call timeval_to_secs
-	  and clock_t_to_secs, respectively
-	- removed no-longer-used third argument for print_timeval and
-	  print_time_in_hz -- `time' output is now done in execute_cmd.c
-
-general.h
-	- new extern declarations for timeval_to_secs and clock_t_to_secs
-
-builtins/times.def
-	- changed calls to print_timeval and print_time_in_hz, removing the
-	  no-longer-used third argument
+builtins/hash.def
+	- if one of the arguments passed to `hash' is an absolute pathname,
+	  just continue the loop, don't do list=list->next first.  This
+	  fixes the `hash a/b' -> core dump bug
 
-execute_cmd.c
-	- changed the way time_command prints its output:  it now interprets
-	  a format string, replacing several escape sequences prefixed with
-	  a `%' with the real, user, and system times.  There are options in
-	  the format string to ask for the `long' format that `times' uses
-	  as well as to specify the number of places after the decimal point,
-	  and whether to output fractional seconds at all
-	- time_command now uses the value of the `TIMEFORMAT' variable, if
-	  present, to format the timing output.  This is flexible enough to
-	  encompass the bash default, posix `time -p', BSD, and SV time
-	  formats
-
-				   7/25
+				   5/30
 				   ----
-Makefile.in
-	- `make clean' should delete the files listed in $(CREATED_SUPPORT);
-	  `make mostlyclean' should not
-
-doc/{bash.{1,html},bashref.texi}
-	- fixed an error in the description of the effect that setting and
-	  unsetting GLOBIGNORE has on the setting of the `dotglob' option
-
-doc/bashref.texi
-	- updated the section listing the major differences between bash
-	  and the SVR4.2 shell
-
-lib/readline/readline.c
-	- removed the setting of _rl_suppress_redisplay, since it doesn't
-	  really do anything yet
-
-oslib.c
-	- new function get_clk_tck (void), returns the value of _SC_CLK_TCK
-	  if sysconf(3) is available, otherwise returns the value of
-	  CLOCKS_PER_SEC (default 60)
-
 general.c
-	- moved CLOCKS_PER_SEC defines to oslib.c
-	- changed clock_t_to_secs to call get_clk_tck once to get the
-	  right value to use
+	- change canonicalize_pathname to leave a leading `/' alone, as
+	  POSIX requires
 
-externs.h
-	- extern declaration for get_clk_tck
-
-execute_cmd.c
-	- changed print_formatted_time so that a `%' at the end of the string
-	  is output literally
-	- changed print_formatted_time to output a newline after the
-	  translated format string, so it does not need to be included in
-	  the format
-	- changed BASH_TIMEFORMAT (the default time format) to have a leading
-	  newline to match ksh93
-
-configure.in
-	- use `shlicc2' on BSD/OS machines
-
-				   7/26
-				   ----
-doc/bash.html
-	- major overhaul, cleaned up text, corrected some minor HTML errors
-
-variables.c
-	- call sv_optind and sv_opterr to initialize the getopts stuff in
-	  initialize_shell_variables
-
-				   7/29
-				   ----
-shell.c
-	- don't let maybe_execute_file try to read files that are not
-	  `regular files'
-
-				   7/30
-				   ----
-parse.y
-	- cast `string' in yy_string_get() to unsigned char * to avoid
-	  sign extension bugs.  For example, bash -c $'ls\377who'
-
-
-builtins/reserved.def
-	- add help text for the `time' reserved word
-	- deleted description of `HISTCONTROL'
-	- added descriptions for HISTIGNORE, PWD, HOSTNAME, GLOBIGNORE,
-	  MACHTYPE
-
-				   7/31
-				   ----
-builtins/common.c
-	- parse_and_execute saves and restores the line number around
-	  calls, and resets it to 0 when pushing the string onto the
-	  input stack
-
-parse.y
-	- added code that parses `` within a double-quoted string as a
-	  single word
-
-				    8/6
-				    ---
-configure.in, config.h.in
-	- look for `textdomain' as well as `gettext' for internationalization
-
-locale.c
-	- when LC_MESSAGES is set, call textdomain as well as setlocale
-	- new variable `default_domain', used to track the value of
-	  $TEXTDOMAIN
-
-subst.c	
-	- call sv_locale when TEXTDOMAIN changes
-
-				    8/7
-				    ---
-error.c
-	- new function, sys_error(), which prints a formatted string followed
-	  by a colon and strerror(errno) to stderr
-
-error.h
-	- extern declaration for sys_error
-
-jobs.c, nojobs.c, input.c, execute_cmd.c, subst.c
-	- changes to use sys_error()
-
-[many files]
-	- changes to many of the literal strings in the code for consistency
-	  and easier potential translation
-
-				    8/8
+				    6/2
 				    ---
-parse.y
-	- new function debug_parser(int) to toggle the value of yydebug
-	  if YYDEBUG != 0
+support/xenix-link.sh
+	- shell script for linking bash under Xenix
 
-				    8/9
+				    6/3
 				    ---
-oslib.c
-	- don't use CLOCKS_PER_SEC, just try to get CLK_TCK (60 by default)
-
 bashline.c
-	- set saved_history_line_to_use to -1 when it's not going to be used
-	  rather than 0.  A value of 0 can be confused with the first
-	  history line
-
-builtins/evalstring.c
-	- moved parse_and_execute and auxiliary functions here from
-	  builtins/common.c
-
-builtins/evalfile.c
-	- moved maybe_execute_file here from shell.c
-
-Makefile.in, builtins/Makefile.in
-	- changes necessitated by the new files
-
-				   8/12
-				   ----
-builtins/common.c
-	- fixed an off-by-one bug in single_quote that could make bash
-	  run off the end of a string if it consisted entirely of single
-	  quotes
-
-				   8/13
-				   ----
-shell.c
-	- moved code from main() to a new function: open_shell_script
-	- don't try to fclose(default_input) if BUFFERED_INPUT is defined
-	- moved argument binding code out of two places in main() and
-	  open_shell_script() to a new function: bind_args().  This takes
-	  an additional argument telling whether to bind starting at $0
-	  or $1
-	- moved the code that calls the with_input_from_* functions into
-	  a new function: set_bash_input
-	- moved the normal shell command-line option parsing code out of
-	  main() into a new function: parse_shell_options
-	- moved the long option parsing code out of main() into a new
-	  function: parse_long_options()
-
-general.c
-	- new function: move_to_high_fd(fd), which tries to move FD to a
-	  file descriptor close to the allowed maximum, returning the new
-	  fd and closing the old one (or returning the old one if something
-	  goes wrong)
-
-general.h
-	- new extern declaration for move_to_high_fd
-
-variables.c
-	- moved indirection_level_string() here from shell.c
-
-jobs.c
-	- change initialize_jobs to use move_to_high_fd() instead of the
-	  inline code to do the same thing
-
-				   8/14
-				   ----
-lib/readline/readline.c
-	- add `en_US.ISO8859-1' to the list of legal LANG values that turns
-	  on readline's 8-bit mode
-
-				   8/15
-				   ----
-builtins/evalfile.c
-	- combined source_file and maybe_execute_file into a new function
-	  _evalfile(fname, flags), where the flags select the appropriate
-	  behavior
-
-eval.c
-	- moved the functions that perform the read-eval loop here from
-	  shell.c
-
-execute_cmd.c
-	- changed CPU calculation for printing timing statistics using
-	  ideas and code from Deven Corzine (deven@ties.org)
-
-eval.c, builtins/evalstring.c
-	- make sure we dispose of global_command if set -n is on and the
-	  shell is not interactive
-
-				   8/16
-				   ----
-variables.c
-	- make a new array variable, BASH_VERSINFO, with version information:
-
-		BASH_VERSINFO[0] = release	(2)
-		BASH_VERSINFO[1] = version	(00)
-		BASH_VERSINFO[2] = patch level	(0)
-		BASH_VERSINFO[3] = build version (1058)
-		BASH_VERSINFO[4] = release status (beta1)
-		BASH_VERSINFO[5] = $MACHTYPE	(sparc-sun-sunos4.1.4)
-
-doc/{bash.{1,html},bashref.texi}
-	- Documented BASH_VERSINFO
-
-configure.in, config.h.in
-	- look for bindtextdomain(), set HAVE_BINDTEXTDOMAIN if found
-
-subst.c
-	- call sv_locale if TEXTDOMAINDIR is set or changes
-
-locale.c
-	- handle TEXTDOMAINDIR in set_locale_var; call bindtextdomain if
-	  it is present
-
-tests/array.tests
-	- added calls to egrep to filter out BASH_VERSINFO and PIPESTATUS,
-	  which can vary from system to system
-
-				   8/19
-				   ----
-shell.c
-	- moved line_buffer_stream from here to oslib.c, made it into a
-	  replacement for setlinebuf() if that is not available
-
-externs.h
-	- extern declaration for setlinebuf()
-
-lib/readline/input.c
-	- in rl_gather_tyi, punt if chars_avail < 0 after checking input fd
-
-				   8/20
-				   ----
-builtins/suspend.def
-	- send SIGSTOP instead of SIGTSTP
-
-				   8/21
-				   ----
-builtins/colon.def
-	- added true and false as builtins (undocumented)
-
-CWRU/POSIX.NOTES, doc/bashref.texi
-	- note that process substitution is not available in posix mode
-
-bashintl.h
-	- new header file for internationalization, included by locale.c
-
-				   8/22
-				   ----
-subst.c
-	- in command_substitute, set istring to NULL before trying to
-	  make the pipe, so if that fails, the code at error_exit:
-	  doesn't try to free it
-
-execute_cmd.c
-	- when executing a shell script without a leading `#!' internally,
-	  and the shell is not interactive, close the fd to the script
-	  we're reading.  If this is called by the exec builtin, the shell
-	  will not have forked, and the fd will not have been closed, even
-	  though it's marked close-on-exec
-	- if subshell_argv is non-null, free its members (except 0) and
-	  subshell_argv in shell_execve before reassigning it
-	- dispose of currently_executing_command before jumping back to
-	  subshell_top_level
-
-builtins/hash.def
-	- use a sentinel variable so we only initialize the hash table
-	  once in initialize_filename_hashing
-
-shell.c
-	- don't flush the filename hash table in shell_reinitialize
-	- free dollar_vars[0] in set_shell_name before assigning it
-	- don't fetch the current hostname more than once
-	- don't fetch the current user information more than once unless
-	  the uid changes
-
-parse.y
-	- initialize_bash_input should free bash_input.name if it is
-	  non-null before zeroing it
+	- fixed a memory leak in command_word_completion_function, courtesy
+	  of a.pfaller@pop.gun.de
 
-				   8/23
-				   ----
-aclocal.m4
-	- changed BASH_CHECK_SOCKLIB:  if -lsocket is present, check for
-	  and cache the existence of -lnsl.  Define LIBS appropriately
-	  based on the values of the cache variables rather than relying
-	  on the autoconf tests to do it
-
-				   8/26
-				   ----
-parse.y
-	- fix up yy_readline_get and yy_stream_get to fix the sign
-	  extension problem
-
-shell.c
-	- move the setjmp(top_level) that catches early SIGINTs to the
-	  first statement in main()
-
-Makefile.in
-	- added a `strip' target to strip the binary
-
-[bash-2.0-beta1 frozen]
-
-				   8/27
-				   ----
-parse.y
-	- replaced several static state variables with a single flags word:
-	  parser_state
-
-				   8/28
-				   ----
-parse.y
-	- in read_token_word, combined the ${...} case with the other
-	  expansion cases -- the code was identical
-	- added code to handle ksh-style ((...)) -- equivalent to
-	  `let "..."'.  In fact, the above construct is translated internally
-	  into `let "..."', so error messages output by the evaluator
-	  will contain `let'
-
-expr.c
-	- changed evalerror to call internal_error so the name of the
-	  shell script is prefixed to the message, if necessary
-
-Makefile.in
-	- changed RELSTATUS to `beta2'
-
-tests/{arith,new-exp}.{tests,right}
-	- changed the expected error message because of the change in format
-	  of the arithmetic evaluation error messages in expr.c
-
-builtins/pushd.def
-	- changed pushd_builtin so `pushd -' is equivalent to `pushd $OLDPWD'
-
-externs.h
-	- changed the extern declaration for setlinebuf() to avoid having
-	  to include stdio.h before externs.h everywhere
-
-				   8/29
-				   ----
-error.c
-	- new function, parser_error, for use by parts of the parser that
-	  don't want to call report_syntax_error
-
-error.h
-	- extern declaration for parser_error
-
-parse.y
-	- changed some parts of the parser to use parser_error
-	- made the ((...)) code #ifdef'd on DPAREN_ARITHMETIC
-
-shell.c
-	- added a line to the --help output saying to use `bashbug' to
-	  report bugs
-
-configure.in
-	- added new option: --enable-dparen-arithmetic; it controls
-	  the DPAREN_ARITHMETIC define in config.h
-	- changed $host_os check for SCO to only enable -DWAITPID_BROKEN
-	  for sco3.2v5; all other SCO versions enable -DMUST_UNBLOCK_CHILD
-
-config.h.in
-	- added line for DPAREN_ARITHMETIC, modified by configure
-
-doc/{bash.{1,html},bashref.texi}
-	- documented new ((...)) command and new configuration option
-
-execute_cmd.c
-	- shell_execve should not try to close default_buffered_input unless
-	  it is >= 0.  Other code that closes it and deallocates the buffer
-	  should set default_buffered_input to -1
-
-				   8/30
-				   ----
-configure.in
-	- more changes to LOCAL_CFLAGS for sco machines
-
-error.c
-	- changed parser_error to handle all cases of the shell being
-	  interactive, not interactive, reading from a script, and reading
-	  from some other file
-
-parse.y
-	- changed report_syntax_error to call parser_error for consistent
-	  messages
-
-dispose_cmd.c
-	- change dispose_command to use programming_error instead of
-	  report_error
-
-execute_cmd.c, input.c, general.c
-	- changed calls to report_error to internal_error, since we don't
-	  want to possibly exit in any of those cases
-
-				    9/3
-				    ---
-input.c
-	- changed the `localbuf' type to `unsigned char'
-
-parse.y
-	- changed yy_stream_get to use feof to test for end of file when
-	  not using getc_with_restart
-
-				    9/6
-				    ---
-config.h.top
-	- added a commented-out define for SYS_BASH_LOGOUT, which is a
-	  system-wide .bash_logout file, run when a login shell exits
-	- added /sbin to STANDARD_UTILS_PATH
-
-builtins/exit.def
-	- added support for SYS_BASH_LOGOUT
-
-				   9/10
-				   ----
-lib/readline/{{chardefs,histlib}.h,{bind,histexpand,readline,util,vi_mode}.c}
-	- changed to_upper to _rl_to_upper and to_lower to _rl_to_lower
-	- changed digit_value to _rl_digit_value
-	- changed digit_p to _rl_digit_p
-	- changed uppercase_p and lowercase_p to _rl_uppercase_p and
-	  _rl_lowercase_p respectively
-	- changed pure_alphabetic to _rl_pure_alphabetic
-
-support/zecho.c
-	- new file, implements a bare-bones echo (no -n, no \-escapes),
-	  to be used by tests
-
-tests/rhs-exp.tests
-	- changed calls to echo to use recho to avoid errors when bash is
-	  configured with --enable-usg-echo-default
-
-tests/more-exp.tests
-	- changed some calls to echo to use zecho to avoid errors when bash
-	  is configured with --enable-usg-echo-default
-
-				   9/11
-				   ----
-lib/readline/histexpand.c
-	- fixed a possible off-by-one error in history_expand_internal when
-	  deciding whether or not to reallocate `result'
-
-expr.c
-	- fixed expcond() so that an assignment may be supplied as the
-	  token after the `?'
-	- illegal bases now cause an evaluation error
-	- multiple base specifications now cause an evaluation error
-
-tests/arith.tests
-	- added more tests for conditional operator and associativity
-
-tests/{run-heredoc,heredoc.{tests,right}}
-	- new set of tests for here-documents
-
-builtins/common.c
-	- print out the name returned by get_name_for_error in
-	  builtin_error()
-
-subst.c
-	- set this_command_name to NULL before processing command-line
-	  assignment statements so arithmetic evaluation errors don't
-	  print garbage
-	- according to POSIX.2, echo ${foo?$x} should print the expanded
-	  value of $x, even when it doesn't expand into anything
-
-				   9/12
-				   ----
-shell.c
-	- changed the execution of startup files so that $ENV is never
-	  run by a non-interactive shell and an interactive shell begun
-	  as `sh' reads and executes $ENV, as posix specifies
-
-doc/{bash.{1,html},bashref.texi}
-	- updated description of startup file behavior
-
-doc/Makefile.in
-	- make sure references to bashref.texi are preceded by $(srcdir)/
-	  so we can build the documentation in another directory just by
-	  typing `make'; e.g. after giving configure the --srcdir option
-
-subst.c
-	- change to array_remove_pattern to avoid evaluating the `param'
-	  part of the ${param#word} spec twice when doing things like
-	  echo ${A[x=(i+=1)]#${A[y=(i+=2)]}}
-
-				   9/13
-				   ----
-builtins/bashgetopt.c
-	- fixed a problem with `;' option specifier (optional argument)
-
-support/config.{sub,guess}
-	- added cases to recognize CRAY C90 and T90
-
-builtins/ulimit.def
-	- rewritten almost from scratch to use internal_getopt and to not
-	  mask the return value from getrlimit/setrlimit, and to make it
-	  easier to maintain
-	- added `-l' option for systems with RLIMIT_MEMLOCK (4.4 BSD)
-
-doc/{bash.{1,html},bashref.texi}
-	- documented new `-l' option to `ulimit'
-
-				   9/24
-				   ----
-parse.y
-	- added `in' to the list of tokens that should not be followed by
-	  a semicolon when combining history lines
-	- removed the special-case code in history_delimiting_chars that
-	  tried to be smart about `()'
-
-				   9/25
-				   ----
-print_cmd.c
-	- fixes to print_group_command from Andreas Schwab
-	- rewrote indent() to use a single malloced string for the spaces
-	  to print, rather than calling cprintf multiple times
-
-shell.c
-	- make sure to set the strings in current_user to NULL after freeing
-	  their contents in uidget().
-
-parse.y
-	- make sure decode_prompt_string quotes the values inserted by the
-	  \w and \W escape sequences using backslashes if the string will be
-	  expanded later because promptvars is set
-
-support/mkdirs
-	- removed leading `^' from regular expressions to be matched by
-	  `expr' on the advice of meyering@appaloosa.asic.sc.ti.com
-
-				   9/26
-				   ----
-parse.y
-	- added a few more possible flag values for parser_state
-	- changed history_delimiting_chars to not add a semicolon after a
-	  right paren if we're parsing a case statement
-
-				   9/27
-				   ----
-builtins/ulimit.def
-	- a few minor cleanups
-
-				   10/1
-				   ----
-builtins/trap.def
-	- make sure first_arg is non-empty before calling signal_object_p()
-
-[bash-2.0-beta2 frozen]
-
-				   10/2
-				   ----
-bashintl.h
-	- if we have setlocale(), but don't have a definition for LC_ALL,
-	  undefine HAVE_SETLOCALE
-
-configure.in
-	- set LOCAL_LDFLAGS on AIX 4.2 to allow dynamic loading of builtins
-
-builtins/enable.def
-	- use correct flags for AIX 4.2 in dlopen call if _AIX is defined
-
-examples/loadables/Makefile
-	- add sample commands for AIX 4.2
-
-				   10/3
-				   ----
-variables.c
-	- new functions: merge_temporary_env(), merge_builtin_env(), to
-	  take temporary_env and builtin_env, respectively, and create
-	  shell variables from them
-
-variables.h
-	- new extern declarations for merge_temporary_env() and
-	  merge_builtin_env()
-
-execute_cmd.c
-	- if variable assignments preceds a posix special builtin, and the
-	  shell is in posix mode, those assignments should persist in the
-	  shell environment after the builtin completes
-
-doc/bashref.texi
-	- modified description of what's changed by posix mode to include
-	  the persistance of assignment statements preceding special builtins
-
-doc/mkposix
-	- new script to create CWRU/POSIX.NOTES directly from the `Bash
-	  POSIX Mode' node of bashref.texi, like INSTALL is created
-
-				   10/4
-				   ----
-lib/{malloc,glob,readline}/Makefile.in, builtins/Makefile.in
-	- don't have autoconf substitute LOCAL_CFLAGS into the CFLAGS
-	  assignment so people on SCO 3.2v[45] can still build with
-	  NO_MEMSCRAMBLE even when they run `make CFLAGS=...'
-
-test.c
-	- renamed eaccess() to test_eaccess() to avoid conflict with SCO
-
-config.h.bot
-	- #undef HAVE_GETCWD if GETCWD_BROKEN is defined, so we use the
-	  replacement in getcwd.c
-
-general.c
-	- fixed a small memory leak in full_pathname that occurs if
-	  getcwd() returns NULL
-
-oslib.c
-	- new constant string variable: bash_getcwd_errstr, to use when
-	  getcwd() returns NULL
-
-{general,jobs}.c, parse.y, builtins/{cd.def,common.c}
-	- change calls to getwd() to use getcwd() instead
-
-general.c, builtins/common.c, builtins/cd.def
-	- changed error behavior when getcwd() returns NULL to use
-	  bash_getcwd_errstr, sys_error(), and strerror(errno) where
-	  appropriate
-
-sig.c
-	- change a couple of calls to signal to set_signal_handler
-
-aclocal.m4
-	- added definitions of _popen and _pclose to the BASH_FUNC_GETCWD
-	  macro
-
-configure.in
-	- added -DPATH_MAX=1024 to LOCAL_CFLAGS for sco3.2v[45] on the
-	  advice of <grog@lemis.de>
-	- removed check for getwd(3)
-
-externs.h
-	- removed extern declaration for getwd(), added extern declaration
-	  for getcwd() if HAVE_GETCWD is not defined
-
-oslib.c
-	- removed definition of getwd(); it is no longer used
-
-lib/readline/rltty.c
-	- changed SETATTR macro for SCO POSIX systems to call tcsetattr
-	  with TCSANOW, on advice of <wbader@nacs.fast.net>
-
-				   10/7
-				   ----
-support/mkdirs
-	- slight change to work around a bug in the HP-UX `expr'
-
-				   10/8
-				   ----
-lib/readline/vi_mode.c
-	- changes so that the text inserted with an `i' command is available
-	  to be reinserted with `.'
-
-examples/functions/autoload.v2
-	- new implementation: uses arrays, keeps a list of autoloaded
-	  functions, allows autoloaded functions to be removed, allows
-	  autoloaded functions to be listed
-
-execute_cmd.c
-	- when freeing the members of subshell_argv in shell_execve(), free
-	  only the members up to subshell_argc
-
-				   10/9
-				   ----
-configure.in
-	- replaced calls to AC_HAVE_FUNCS with AC_CHECK_FUNCS
-	- replaced calls to AC_FUNC_CHECK with AC_CHECK_FUNC
-	- replaced calls to AC_HAVE_HEADERS with AC_CHECK_HEADERS
-	- added additional code to check for libintl.{a,so} for the
-	  internationalization functions if bindtextdomain is not found
-	  in libc
-	- define HAVE_WAIT3 if AC_FUNC_CHECK(wait3,...) succeeds, rather
-	  than HAVE_FUNC_WAIT3, which is not used in the source
-	- added a call to AC_FUNC_STRCOLL to check for the presence of a
-	  working strcoll(3)
-
-config.h.in
-	- removed HAVE_FUNC_WAIT3
-	- added HAVE_STRCOLL
-
-variables.c, stringlib.c
-	- move sort_char_array from variables.c to stringlib.c
-
-variables.h, externs.h
-	- move extern declaration of sort_char_array from variables.h to
-	  externs.h
-
-variables.c
-	- don't sort the exported environment, even when we're not in
-	  posix mode
-
-stringlib.c
-	- changed qsort_string_compare to use strcoll(3) if it is available
-	  (HAVE_STRCOLL is defined).  This means that the results of glob
-	  expansions are now sorted in the per-locale collation order
-
-locale.c
-	- track the value of $LC_COLLATE and call setlocale(LC_COLLATE,...)
-	  when it changes.  Also, set its default value
-
-subst.c
-	- LC_COLLATE is now a special variable; call sv_locale when it is
-	  assigned to or unset
-
-lib/readline/util.c
-	- move compare_strings here from complete.c; renamed it
-	  _rl_qsort_string_compare; made it use strcoll(3) if it's
-	  available; made it global so other library files can use it
-
-lib/readline/complete.c
-	- use _rl_qsort_string_compare instead of static compare_strings.
-	  This means that completions are now sorted in a locale-specific
-	  collation order
-
-lib/readline/funmap.c
-	- removed static declaration of qsort_string_compare; qsort now
-	  uses _rl_qsort_string_compare instead
-
-oslib.c
-	- rewrote stricmp, strnicmp; renamed them strcasecmp and
-	  strncasecmp, since they're only compiled in if HAVE_STRCASECMP
-	  is not defined
-
-externs.h
-	- removed extern declarations of strnicmp, stricmp, replacing them
-	  with declarations of strcasecmp and strncasecmp if HAVE_STRCASECMP
-	  is not defined; removed #defines of stricmp and strnicmp to
-	  strcasecmp and strncasecmp
-
-trap.c
-	- changed calls to stricmp to direct calls to strcasecmp
-
-Makefile.in
-	- bumped RELSTATUS to `beta3'
-
-doc/bash.{1,html}
-	- added mention of LC_COLLATE
-
-doc/bashref.texi
-	- removed item about sorting environment from the section on posix
-	  mode
-
-				   10/16
-				   -----
-subst.c
-	- fixed array_length_reference to return 0 for an unbound variable
-	  and 1 for a variable that is not an array
-
-examples/scripts.v2/cdhist.bash
-	- package with functions to replace cd and maintain a cd directory
-	  stack
+hashcmd.c
+	- fixed find_hashed_filename to always return a newly-allocated
+	  string
 
-examples/scripts.v2/pmtop
-	- a `poor man's top' for SunOS 4.x and BSD
-
-examples/scripts.v2/bashrand
-	- a script to return a random number within a specified range, with
-	  an optional seed
-
-examples/scripts/zprintf
-	- printf(1) replacement that uses gawk to do the output
-
-doc/{bash.{1,html},bashref.texi}, lib/readline/doc/hsuser.texinfo
-	- several small corrections
-
-				   10/17
-				   -----
-variables.c
-	- changed the calls to list_string_with_quotes and then
-	  expand_words_no_vars in assign_array_var_from_string to a single
-	  call to expand_string.  This behaves more rationally, and closer
-	  to ksh93
-
-subst.c
-	- commented out list_string_with_quotes -- it is no longer used
-
-lib/readline/terminal.c
-	- new function, _rl_set_screen_size(int rows, int cols), to set
-	  readline's idea of the screen size
-
-jobs.c, nojobs.c
-	- make sure that window size changes are propagated back down to
-	  the readline code, by calling _rl_set_screen_size with the
-	  new sizes.  This happens when bash gets a SIGWINCH or if
-	  `checkwinsize' is set with `shopt'
-
-				   10/18
-				   -----
-support/config.guess
-	- added code to recognize the various NetBSD ports
-
-lib/readline/signals.c
-	- don't install a handler for SIGALRM if the containing application
-	  has installed a handler with SA_RESTART as part of the flags.
-	  This is in effect, obviously, only for systems with POSIX signals.
-
-subst.c
-	- make parameter_brace_expand_length expand variables in a double-
-	  quoted environment so it takes spaces in the variable value into
-	  account
-
-				   10/21
-				   -----
-command.h
-	- flags for possible values of subshell_environment telling what
-	  caused the subshell
-
-builtins/command.def
-	- the `command' builtin should not be setting CMD_NO_FORK at all;
-	  it doesn't know enough about the execution context to do so
-
-execute_cmd.c, subst.c
-	- set subshell_environment to one of the values in command.h rather
-	  than to `1', so we know what kind of subshell it is (just for
-	  informational purposes; nothing actually uses this yet)
-
-
-				   10/22
-				   -----
 execute_cmd.c
-	- just return EXECUTION_SUCCESS immediately in execute_command_internal
-	  if read_but_dont_execute is set
-
-				   10/23
-				   -----
-general.c
-	- fixed check_dev_tty to return immediately if ttyname() returns NULL
-
-builtins/suspend.def
-	- SIGSTOP cannot be caught, so there's no reason to try to do so
-
-examples/loadables/{finfo,pathchk,print}.c
-	- minor fixes
-
-pathexp.c
-	- fix to off-by-one error in setup_ignore_patterns
-
-jobs.c
-	- try to set the tty pgrp to our pgrp if we change our pgrp to our
-	  pid and become a process group leader.  If it doesn't work,
-	  reset our pgrp to what it was originally and disable job control
-
-aclocal.m4
-	- new test, BASH_STRUCT_TERMIO_LDISC, to test for a c_line member
-	  of struct termio
-
-config.h.in
-	- add a new preprocessor variable, TERMIO_LDISC, undefined by default
-
-configure.in
-	- call BASH_STRUCT_TERMIO_LDISC after BASH_STRUCT_TERMIOS_LDISC
-
-				   10/25
-				   -----
-lib/readline/terminal.c
-	- increased the size of the termcap buffers to 4096
-
-				   10/29
-				   -----
-parse.y
-	- include memalloc.h for uses of alloca that the code in the
-	  generated bison parser does not catch (e.g., HP-UX 10.10)
-
-aclocal.m4
-	- new test, BASH_KERNEL_RLIMIT_CHECK, to check whether a given
-	  HP-UX machine needs _KERNEL defined for the RLIMIT_* defines in
-	  <sys/resource.h> (9.05, 10.01 yes, 10.10,10.20 no)
-
-configure.in
-	- call BASH_KERNEL_RLIMIT_CHECK if $host_os begins with `hpux'
-
-config.h.in
-	- new #undef for RLIMIT_NEEDS_KERNEL
-
-builtins/ulimit.def
-	- only define _KERNEL before including <sys/resource.h> if HPUX
-	  and RLIMIT_NEEDS_KERNEL are both defined
-
-[bash-2.0-beta3 frozen]
-
-				   11/1
-				   ----
-locale.c
-	- if setting LC_ALL to a null value (as is done when `unset LC_ALL'
-	  is run), call setlocale() to set LC_ALL back to default_locale
-
-lib/readline/readline.c
-	- moved i18n code from here to nls.c
-
-lib/readline/Makefile.in, Makefile.in
-	- new readline library member, nls.c
-
-lib/readline/nls.c
-	- new file, with internationalization code from readline.c (such as
-	  it is)
-	- updated legal_lang_values to just contain a list of codesets
-	- added code modified from a diff sent by Ulrich Drepper
-	  <drepper@cygnus.com> to parse a locale specification and isolate
-	  and normalize the codeset part, for checking against the values
-	  in legal_lang_values
-
-				   11/4
-				   ----
-builtins/setattr.def, locale.c, aclocal.m4, oslib.c
-	- minor cleanups
-
-				   11/5
-				   ----
-lib/readline/complete.c, test.c, trap.c
-	- more minor cleanups
-
-				   11/6
-				   ----
-parse.y
-	- added a `flags' parameter to parse_matched_pair, currently unused
-
-				   11/7
-				   ----
-configure.in, config.h.in
-	- add a check for time_t, defaulting to `long' if it's not
-	  defined in <sys/types.h>
-
-builtins/fc.def
-	- use `time_t *' instead of `long *' in call to time()
-
-				   11/8
-				   ----
-configure.in
-	- removed check for getpwent in libsun.a; it's not necessary for
-	  recent versions of irix (irix[56].?)
-
-				   11/11
-				   -----
-test.c
-	- used `&&' where `||' was needed when testing if HAVE_LSTAT and
-	  S_IFLNK are both defined
-
-[bash-2.0-beta3 net release version frozen]
-
-				   11/12
-				   -----
-builtins/getopt.c
-	- when sh_getopt detects that a required argument is not present,
-	  set sh_optarg to "" so that getopts does the right thing
-
-				   11/15
-				   -----
-builtins/ulimit.c
-	- made the default (without -H or -S) report the soft limit and set
-	  both soft and hard limits (like sh and ksh)
-
-expr.c
-	- changed the maximum base in base#num constants to 64, implemented
-	  just like ksh93
-
-doc/{bash.{1,html},bashref.texi}
-	- updated description of ulimit builtin to new behavior when
-	  neither -H nor -S is supplied
-	- updated arithmetic evaluation section with new maximum base and
-	  explanation of how numbers between 10 and 63 are represented
-
-lib/readline/vi_mode.c
-	- in rl_vi_done_inserting, don't try to save the text entered while
-	  in insert mode unless rl_undo_list is non-null (indicating that
-	  there was actually some text to be saved)
-
-				   11/19
-				   -----
-lib/readline/tcap.h
-	- new file, for termcap library definitions.  Uses <termcap.h> if
-	  it is available
-
-lib/readline/rldefs.h
-	- remove inclusion of termcap.h; now included in tcap.h
-
-lib/readline/{terminal,display,readline}.c
-	- include tcap.h
-
-lib/readline/Makefile.in
-	- update dependencies to include tcap.h
-
-				   11/21
-				   -----
-lib/readline/complete.c
-	- broke the code that computes the lcd of the list of completion
-	  matches out into a separate function, compute_lcd_of_matches()
-
-lib/readline/complete.c
-	- call compute_lcd_of_matches after calling the application-specific
-	  completion ignore function, since it may have eliminated some
-	  names from the match list
-
-builtins/cd.def
-	- replaced instances of MAXNAMLEN in the spelling correction code
-	  with PATH_MAX
-
-lib/readline/terminal.c
-	- new termcap capabilities: kh (sent by Home key) and kH (sent by
-	  End key) automatically bound to beginning-of-line and end-of-line
-	  respectively in emacs_standard and vi_movement maps
-
-mailcheck.c
-	- fix to remember_mail_dates so any message following the
-	  filename is saved correctly
-
-				   11/22
-				   -----
-lib/readline/bind.c
-	- fixed extern declaration of _rl_parsing_coditionalized_out
-	  to match declaration in readline.c
-
-				   11/25
-				   -----
-doc/{bash.{1,html},bashref.texi}
-	- fixed a gross documentation error in the description of
-	  ${parameter:offset:length} (offset and length were transposed)
-	- slight changes to the description of output redirection
-
-subst.c
-	- new function, quoted_substring, to return substrings of variables
-	  when the expression is within double quotes and the value has
-	  CTLESC characters embedded within it.  Called by
-	  parameter_brace_substring
-	- get_var_and_type should return VT_POSPARMS if `*' is passed as
-	  the variable name
-	- added a new first parameter for pos_params -- the variable name
-	- changed pos_params to obey the rules for "$*" wrt using the
-	  first character of IFS as the separator in the result
-	- if an unquoted $* or $@ is expanded, call quote_escapes on the
-	  result to protect any special characters from inadvertently
-	  causing quoting
-
-				   12/3
-				   ----
-builtins/ulimit.def
-	- corrected a misspelling of `unlimited'
-	- added support for linux RLIMIT_AS resource limit for -v option
-	  and changed the block multiplier to 1024 if we have RLIMIT_VMEM
-	  or RLIMIT_AS
-
-				   12/5
-				   ----
-lib/readline/readline.c
-	- fix rl_insert to handle a count of exactly 1024
-
-shell.c
-	- new function: unbind_args(), which removes all the positional
-	  parameters.  Just calls remember_args(NULL, 1)
-
-execute_cmd.c
-	- call unbind_args when executing a subshell, after setting up
-	  subshell_argX
-
-bashintl.h
-	- include <libintl.h> if HAVE_LIBINTL_H defined
-
-config.h.in
-	- add HAVE_LIBINTL_H define
-
-configure.in
-	- add call to AC_CHECK_HEADERS for libintl.h
-
-doc/Makefile.in
-	- added infodir and various mandir defines for completeness
-
-Makefile.in
-	- changed some of the *dir defines to use the autoconf values
-	  (e.g., bindir = @bindir@)
-	- changed manroot to mandir, removed old mandir and manext variables
-	- added `manpfx' variable; should be either `man' or `cat'
-
-				   12/6
-				   ----
-lib/readline/readline.c
-	- fixed rl_digit_loop so it once again displays the argument count
-	  as it's being entered (added calls to _rl_save_prompt and
-	  _rl_restore_prompt)
-
-				   12/7
-				   ----
-support/config.guess
-	- recognize netbsd/vax
-	- recognize openbsd on all supported platforms
-	- recognize tenon's machten running on powerpc and macintosh
-
-				   12/9
-				   ----
-support/config.guess
-	- added all sorts of systems from config.guess distributed with
-	  autoconf-2.12
-
-				   12/10
-				   -----
-subst.c
-	- new function quoted_strchr(); does the same thing as strchr but
-	  honors shell backslash and CTLESC quoting conventions
-	- parameter_brace_patsub now calls quoted_strchr
-
-				   12/11
-				   -----
-shell.c
-	- if the shell is named `-su', run the normal login shell startup
-	  files, even if the shell is not interactive.  This handles the
-	  case of `su - username -c command' compatibly
-
-lib/readline/readline.c
-	- minor change to rl_digit_loop so that M-- and M--1 are equivalent,
-	  as the prompt for digit arguments has always implied
-
-lib/malloc/malloc.c
-	- added definitions of calloc() and cfree(), compiled in unless
-	  NO_CALLOC is defined to cpp
-
-
-				   12/16
-				   -----
-parse.y
-	- protect uses of history_expansion_inhibited with #ifdef
-	  BANG_HISTORY
-
-[bash-2.0-release candidate frozen]
-
-builtins/common.c
-	- changed backslash_quote to quote a tilde if it's the first
-	  character in the string
-
-				   12/17
-				   -----
-aclocal.m4
-	- changed BASH_CC_WORKS to provide a cross-compiling default for
-	  better error messages
-
-configure.in
-	- generated configure using autoconf-2.12
-	- changed the default value of CFLAGS to `-g -O2'
-
-				   12/18
-				   -----
-aclocal.m4
-	- look for /var/spool/mail before /usr/spool/mail when checking
-	  for the default mail directory
-
-jobs.c, builtins/shopt.def
-	- fixed a couple of `macro replacement within a string literal'
-	  problems
-
-Makefile.in
-	- removed `realclean' target
-
-shell.c
-	- changed the first line ouput by bash --version to conform to
-	  the latest GNU coding standards (replace space between version
-	  and machtype with a hyphen)
-
-				   12/19
-				   -----
-execute_cmd.c
-	- when looking at the debug trap in execute_function and
-	  execute_command_internal, make sure it's trapped and not
-	  ignored before trying to save the value of the trap string
-
-				   12/20
-				   -----
-trap.c
-	- changed set_signal and ignore_signal to handle EXIT_TRAP and
-	  DEBUG_TRAP specially, so we don't try to fetch signal handlers
-	  or do other unnecessary things
-
-				   12/22
-				   -----
-builtins/ulimit.def
-	- if the current euid is not 0, and we're trying to set the hard
-	  limit of a resource to RLIM_INFINITY, set it to the current hard
-	  limit to avoid error messages
+	- since find_hashed_filename returns newly-allocated memory, don't
+	  call savestring() on it, and free it if the data is stale (in
+	  search_for_command()).  Another memory leak fixed courtesy of
+	  a.pfaller@pop.gun.de
 
-builtins/read.def
-	- fix for bug that caused core dump if line read had leading white
-	  space and IFS was set to ''
+builtins/type.def
+	- free the value returned by find_hashed_filename
 
-[bash-2.0-release frozen]
+[bash-2.01-release frozen]