Merge perlext/Compiler/... into mainline. Some files move to

ext/B/..., some to lib/B/..., O.pm and B.pm go in lib and some
move to the base perl directory (e.g. headers). Will need some
cleaning up before it builds properly, I would guess.

p4raw-id: //depot/perl@562

6 files changed, 597 insertions, 0 deletions

diff --git a/ext/B/ramblings/cc.notes b/ext/B/ramblings/cc.notes
new file mode 100644
index 0000000000..47bd65a09d
--- /dev/null
+++ b/ext/B/ramblings/cc.notes
@@ -0,0 +1,32 @@
+At entry to each basic block, the following can be assumed (and hence
+must be forced where necessary at the end of each basic block):
+
+The shadow stack @stack is empty.
+For each lexical object in @pad, VALID_IV holds for each T_INT,
+VALID_DOUBLE holds for each T_DOUBLE and VALID_SV holds otherwise.
+The C shadow variable sp holds the stack pointer (not necessarily stack_sp).
+
+write_back_stack
+    Writes the contents of the shadow stack @stack back to the real stack.
+    A write-back of each object in the stack is forced so that its
+    backing SV contains the right value and that SV is then pushed onto the
+    real stack. On return, @stack is empty.
+
+write_back_lexicals
+    Forces a write-back (i.e. achieves VALID_SV), where necessary, for each
+    lexical object in @pad. Objects with the TEMPORARY flag are skipped. If
+    write_back_lexicals is called with an (optional) argument, then it is
+    taken to be a bitmask of more flags: any lexical object with one of those
+    flags set is also skipped and not written back to its SV.
+
+invalidate_lexicals($avoid)
+    The VALID_INT and VALID_DOUBLE flags are turned off for each lexical
+    object in @pad whose flags field doesn't overlap with $avoid.
+
+reload_lexicals
+    For each necessary lexical object in @pad, makes sure that VALID_IV
+    holds for objects of type T_INT, VALID_DOUBLE holds for objects for
+    type T_DOUBLE, and VALID_SV holds for other objects. An object is
+    considered for reloading if its flags field does not overlap with the
+    (optional) argument passed to reload_lexicals.
+
diff --git a/ext/B/ramblings/curcop.runtime b/ext/B/ramblings/curcop.runtime
new file mode 100644
index 0000000000..9b8b7d52e7
--- /dev/null
+++ b/ext/B/ramblings/curcop.runtime
@@ -0,0 +1,39 @@
+PP code uses of curcop
+----------------------
+
+pp_rv2gv
+	when a new glob is created for an OPpLVAL_INTRO,
+	curcop->cop_line is stored as GvLINE() in the new GP.
+pp_bless
+	curcop->cop_stash is used as the stash in the one-arg form of bless
+
+pp_repeat
+	tests (curcop != &compiling) to warn "Can't x= to readonly value"
+
+pp_pos
+pp_substr
+pp_index
+pp_rindex
+pp_aslice
+pp_lslice
+pp_splice
+	curcop->cop_arybase
+
+pp_sort
+	curcop->cop_stash used to determine whether to gv_fetchpv $a and $b
+
+pp_caller
+	tests (curcop->cop_stash == debstash) to determine whether
+	to set DB::args
+
+pp_reset
+	resets vars in curcop->cop_stash
+
+pp_dbstate
+	sets curcop = (COP*)op
+
+doeval
+	compiles into curcop->cop_stash
+
+pp_nextstate
+	sets curcop = (COP*)op
diff --git a/ext/B/ramblings/flip-flop b/ext/B/ramblings/flip-flop
new file mode 100644
index 0000000000..183d541b98
--- /dev/null
+++ b/ext/B/ramblings/flip-flop
@@ -0,0 +1,51 @@
+PP(pp_range)
+{
+    if (GIMME == G_ARRAY)
+        return cCONDOP->op_true;
+    return SvTRUEx(PAD_SV(op->op_targ)) ? cCONDOP->op_false : cCONDOP->op_true;
+}
+
+pp_range is a CONDOP.
+In array context, it just returns op_true.
+In scalar context it checks the truth of targ and returns
+op_false if true, op_true if false.
+
+flip is an UNOP.
+It "looks after" its child which is always a pp_range CONDOP.
+In array context, it just returns the child's op_false.
+In scalar context, there are three possible outcomes:
+  (1) set child's targ to 1, our targ to 1 and return op_next.
+  (2) set child's targ to 1, our targ to 0, sp-- and return child's op_false.
+  (3) Blank targ and  TOPs and return op_next.
+Case 1 happens for a "..." with a matching lineno... or true TOPs.
+Case 2 happens for a ".." with a matching lineno... or true TOPs.
+Case 3 happens for a non-matching lineno or false TOPs.
+
+               $a = lhs..rhs;
+
+        ,------->  range
+        ^         /     \
+        |    true/       \false
+        |       /         \
+   first|     lhs        rhs
+        |      \   first   /
+        ^--- flip <----- flop
+                 \       /
+                  \     /
+                  sassign
+
+
+/* range */
+if (SvTRUE(curpad[op->op_targ]))
+    goto label(op_false);
+/* op_true */
+...
+/* flip */
+/* For "..." returns op_next. For ".." returns op_next or op_first->op_false */
+/* end of basic block */
+goto out;
+label(range op_false):
+...
+/* flop */
+out:
+...
diff --git a/ext/B/ramblings/magic b/ext/B/ramblings/magic
new file mode 100644
index 0000000000..e41930a0f0
--- /dev/null
+++ b/ext/B/ramblings/magic
@@ -0,0 +1,93 @@
+sv_magic()
+----------
+av.c
+av_store()
+	Storing a non-undef element into an SMAGICAL array, av,
+	assigns the equivalent lowercase form of magic (of the first
+	MAGIC in the chain) to the value (with obj = av, name = 0 and
+	namlen = array index).
+
+gv.c
+gv_init()
+	Initialising gv assigns '*' magic to it with obj = gv, name =
+	GvNAME and namlen = GvNAMELEN.
+gv_fetchpv()
+	@ISA gets 'I' magic with obj = gv, zero name and namlen.
+	%OVERLOAD gets 'A' magic with obj = gv, zero name and namlen.
+	$1 to $9, $&, $`, $', $+ get '\0' magic with obj = gv,
+	name = GvNAME and namlen = len ( = 1 presumably).
+Gv_AMupdate()
+	Stashes for overload magic seem to get 'c' magic with obj = 0,
+	name = &amt and namlen = sizeof(amt).
+hv_magic(hv, gv, how)
+	Gives magic how to hv with obj = gv and zero name and namlen.
+
+mg.c
+mg_copy(sv, nsv, key, klen)
+	Traverses the magic chain of sv. Upper case forms of magic
+	(only) are copied across to nsv, preserving obj but using
+	name = key and namlen = klen.
+magic_setpos()
+	LvTARG of a PVLV gets 'g' magic with obj = name = 0 and namlen = pos.
+
+op.c
+mod()
+	PVLV operators give magic to their targs with
+	obj = name = namlen = 0. OP_POS gives '.', OP_VEC gives 'v'
+	and OP_SUBSTR gives 'x'.
+
+perl.c
+magicname(sym, name, namlen)
+	Fetches/creates a GV with name sym and gives it '\0' magic
+	with obj = gv, name and namlen as passed.
+init_postdump_symbols()
+	Elements of the environment get given SVs with 'e' magic.
+	obj = sv and name and namlen point to the actual string
+	within env.
+
+pp.c
+pp_av2arylen()
+	$#foo gives '#' magic to the new SV with obj = av and
+	name = namlen = 0.
+pp_study()
+	SV gets 'g' magic with obj = name = namlen = 0.
+pp_substr()
+	PVLV gets 'x' magic with obj = name = namlen = 0.
+pp_vec()
+	PVLV gets 'x' magic with obj = name = namlen = 0.
+
+pp_hot.c
+pp_match()
+	m//g gets 'g' magic with obj = name = namlen = 0.
+
+pp_sys.c
+pp_tie()
+	sv gets magic with obj = sv and name = namlen = 0.
+	If an HV or an AV, it gets 'P' magic, otherwise 'q' magic.
+pp_dbmopen()
+	'P' magic for the HV just as with pp_tie().
+pp_sysread()
+	If tainting, the buffer SV gets 't' magic with
+	obj = name = namlen = 0.
+
+sv.c
+sv_setsv()
+	Doing sv_setsv(dstr, gv) gives '*' magic to dstr with
+	obj = dstr, name = GvNAME, namlen = GvNAMELEN.
+
+util.c
+fbm_compile()
+	The PVBM gets 'B' magic with obj = name = namlen = 0 and SvVALID
+	is set to indicate that the Boyer-Moore table is valid.
+	magic_setbm() just clears the SvVALID flag.
+
+hv_magic()
+----------
+
+gv.c
+gv_fetchfile()
+	With perldb, the HV of a gvfile gv gets 'L' magic with obj = gv.
+gv_fetchpv()
+	%SIG gets 'S' magic with obj = siggv.
+init_postdump_symbols()
+	%ENV gets 'E' magic with obj = envgv.
diff --git a/ext/B/ramblings/reg.alloc b/ext/B/ramblings/reg.alloc
new file mode 100644
index 0000000000..7fd69f2ebe
--- /dev/null
+++ b/ext/B/ramblings/reg.alloc
@@ -0,0 +1,32 @@
+while ($i--) {
+    foo();
+}
+exit
+
+    PP code			if i an int register	if i an int but not a
+				(i.e. can't be		register (i.e. can be
+				implicitly invalidated)	implicitly invalidated)
+    nextstate
+    enterloop
+
+							
+  loop:
+    gvsv  GV (0xe6078) *i	validates i		validates i
+    postdec			invalidates $i		invalidates $i
+    and if_false goto out;
+				i valid; $i invalid	i valid; $i invalid
+
+				i valid; $i invalid	i valid; $i invalid
+    nextstate
+    pushmark
+    gv  GV (0xe600c) *foo
+    entersub						validates $i; invals i
+
+    unstack
+    goto loop:
+
+				i valid; $i invalid
+  out:
+    leaveloop
+    nextstate
+    exit
diff --git a/ext/B/ramblings/runtime.porting b/ext/B/ramblings/runtime.porting
new file mode 100644
index 0000000000..4699b255cf
--- /dev/null
+++ b/ext/B/ramblings/runtime.porting
@@ -0,0 +1,350 @@
+Notes on porting the perl runtime PP engine.
+Importance: 1 = who cares?, 10 = vital
+Difficulty: 1 = trivial, 10 = very difficult. Level assumes a
+reasonable implementation of the SV and OP API already ported.
+
+OP		Import	Diff	Comments
+null		10	1
+stub		10	1
+scalar		10	1
+pushmark	10	1	PUSHMARK
+wantarray	7	3	cxstack, dopoptosub
+const		10	1	
+gvsv		10	1	save_scalar
+gv		10	1	
+gelem		3	3	
+padsv		10	2	SAVECLEARSV, provide_ref
+padav		10	2
+padhv		10	2
+padany		1	1
+pushre		7	3	pushes an op. Blech.
+rv2gv		6	5
+rv2sv		10	4
+av2arylen	7	3	sv_magic	
+rv2cv		8	5	sv_2cv
+anoncode	7	6	cv_clone
+prototype	4	4	sv_2cv
+refgen		8	3
+srefgen		8	2
+ref		8	3
+bless		7	3
+backtick	5	4
+glob		5	2	do_readline
+readline	8	2	do_readline	
+rcatline	8	2
+regcmaybe	8	1
+regcomp		8	9	pregcomp
+match		8	10
+subst		8	10
+substcont	8	7
+trans		7	4	do_trans
+sassign		10	3	mg_find, SvSETMAGIC
+aassign		10	5	
+chop		8	3	do_chop
+schop		8	3	do_chop
+chomp		8	3	do_chomp
+schomp		8	3	do_chomp
+defined		10	2
+undef		10	3
+study		4	5
+pos		8	3	PVLV, mg_find
+preinc		10	2	sv_inc, SvSETMAGIC
+i_preinc
+predec		10	2	sv_dec, SvSETMAGIC
+i_predec
+postinc		10	2	sv_dec, SvSETMAGIC
+i_postinc
+postdec		10	2	sv_dec, SvSETMAGIC
+i_postdec
+pow		10	1
+multiply	10	1
+i_multiply	10	1
+divide		10	2
+i_divide	10	1
+modulo		10	2
+i_modulo	10	1
+repeat		6	4
+add		10	1
+i_add		10	1
+subtract	10	1
+i_subtract	10	1
+concat		10	2	mg_get
+stringify	10	2	sv_setpvn
+left_shift	10	1
+right_shift	10	1
+lt		10	1
+i_lt		10	1
+gt		10	1
+i_gt		10	1
+le		10	1
+i_le		10	1
+ge		10	1
+i_ge		10	1
+eq		10	1
+i_eq		10	1
+ne		10	1
+i_ne		10	1
+ncmp		10	1
+i_ncmp		10	1
+slt		10	2
+sgt		10	2
+sle		10	2
+sge		10	2
+seq		10	2	sv_eq
+sne		10	2
+scmp		10	2
+bit_and		10	2
+bit_xor		10	2
+bit_or		10	2
+negate		10	3
+i_negate	10	1
+not		10	1
+complement	10	3
+atan2		6	1
+sin		6	1
+cos		6	1
+rand		5	2
+srand		5	2
+exp		6	1
+log		6	2
+sqrt		6	2
+int		10	2
+hex		9	2
+oct		9	2
+abs		10	1
+length		10	1
+substr		10	4	PVLV
+vec		5	4
+index		9	3
+rindex		9	3
+sprintf		9	4	do_sprintf
+formline	6	7
+ord		6	2	
+chr		6	2
+crypt		3	2
+ucfirst		6	2
+lcfirst		6	2
+uc		6	2
+lc		6	2
+quotemeta	6	3
+rv2av		10	3	save_svref, mg_get, save_ary
+aelemfast	10	2	av_fetch
+aelem		10	3
+aslice		9	4
+each		10	3	hv_iternext
+values		10	3	do_kv
+keys		10	3	do_kv
+delete		10	3
+exists		10	3
+rv2hv		10	3	save_svref, mg_get, save_ary, do_kv
+helem		10	3	save_svref, provide_ref
+hslice		9	4
+unpack		9	6	lengthy
+pack		9	6	lengthy
+split		9	9
+join		10	4	do_join
+list		10	2
+lslice		9	4
+anonlist	10	2
+anonhash	10	3
+splice		9	6
+push		10	2
+pop		10	2
+shift		10	2
+unshift		10	2
+sort		6	7
+reverse		9	4
+grepstart	6	5	modifies flow of control
+grepwhile	6	5	modifies flow of control
+mapstart	1	1
+mapwhile	6	5	modifies flow of control
+range		7	3	modifies flow of control
+flip		7	4	modifies flow of control
+flop		7	4	modifies flow of control
+and		10	3	modifies flow of control
+or		10	3	modifies flow of control
+xor
+cond_expr	10	3	modifies flow of control
+andassign	7	3	modifies flow of control
+orassign	7	3	modifies flow of control
+method		8	5
+entersub	10	7
+leavesub	10	5
+caller		2	8
+warn		9	3
+die		9	3
+reset		2	2
+lineseq		1	1
+nextstate	10	1	Update stack_sp from cxstack. FREETMPS.
+dbstate		3	7		
+unstack
+enter		10	3	cxstack, ENTER, SAVETMPS, PUSHBLOCK
+leave		10	3	cxstack, SAVETMPS, LEAVE, POPBLOCK
+scope		1	1
+enteriter	9	4	cxstack
+iter		9	3	cxstack
+enterloop	10	4
+leaveloop	10	4
+return		10	5
+last		9	6
+next		9	6
+redo		9	6
+dump		1	9	pp_goto
+goto		6	9
+exit		9	2	my_exit
+open		9	5	do_open
+close		9	3	do_close
+pipe_op		7	4
+fileno		9	2
+umask		4	2
+binmode		4	2
+tie		5	5	pp_entersub
+untie		5	2	sv_unmagic
+tied		5	2
+dbmopen		4	5
+dbmclose	4	2
+sselect		4	4
+select		7	3
+getc		7	2
+read		8	2	pp_sysread
+enterwrite	4	4	doform
+leavewrite	4	5
+prtf		4	4	do_sprintf
+print		8	6
+sysopen		8	2
+sysread		8	4
+syswrite	8	4	pp_send
+send		8	4
+recv		8	4	pp_sysread
+eof		9	2
+tell		9	3
+seek		9	2
+truncate	8	3
+fcntl		8	4	pp_ioctl
+ioctl		8	4
+flock		8	2
+socket		5	3
+sockpair	5	3
+bind		5	3
+connect		5	3
+listen		5	3
+accept		5	3
+shutdown	5	2
+gsockopt	5	3	pp_ssockopt
+ssockopt	5	3
+getsockname	5	3	pp_getpeername
+getpeername	5	3
+lstat		5	4	pp_stat
+stat		5	4	lengthy
+ftrread		5	2	cando
+ftrwrite	5	2	cando
+ftrexec		5	2	cando
+fteread		5	2	cando
+ftewrite	5	2	cando
+fteexec		5	2	cando
+ftis		5	2	cando
+fteowned	5	2	cando
+ftrowned	5	2	cando
+ftzero		5	2	cando
+ftsize		5	2	cando
+ftmtime		5	2	cando
+ftatime		5	2	cando
+ftctime		5	2	cando
+ftsock		5	2	cando
+ftchr		5	2	cando
+ftblk		5	2	cando
+ftfile		5	2	cando
+ftdir		5	2	cando
+ftpipe		5	2	cando
+ftlink		5	2	cando
+ftsuid		5	2	cando
+ftsgid		5	2	cando
+ftsvtx		5	2	cando
+fttty		5	2	cando
+fttext		5	4
+ftbinary	5	4	fttext
+chdir
+chown
+chroot
+unlink
+chmod
+utime
+rename
+link
+symlink
+readlink
+mkdir
+rmdir
+open_dir
+readdir
+telldir
+seekdir
+rewinddir
+closedir
+fork
+wait
+waitpid
+system
+exec
+kill
+getppid
+getpgrp
+setpgrp
+getpriority
+setpriority
+time
+tms
+localtime
+gmtime
+alarm
+sleep
+shmget
+shmctl
+shmread
+shmwrite
+msgget
+msgctl
+msgsnd
+msgrcv
+semget
+semctl
+semop
+require		6	9	doeval
+dofile		6	9	doeval
+entereval	6	9	doeval
+leaveeval	6	5
+entertry	7	4	modifies flow of control
+leavetry	7	3
+ghbyname
+ghbyaddr
+ghostent
+gnbyname
+gnbyaddr
+gnetent
+gpbyname
+gpbynumber
+gprotoent
+gsbyname
+gsbyport
+gservent
+shostent
+snetent
+sprotoent
+sservent
+ehostent
+enetent
+eprotoent
+eservent
+gpwnam
+gpwuid
+gpwent
+spwent
+epwent
+ggrnam
+ggrgid
+ggrent
+sgrent
+egrent
+getlogin
+syscall
+	
\ No newline at end of file