diff options
Diffstat (limited to 'ChangeLog')
| -rw-r--r-- | ChangeLog | 70 |
1 files changed, 70 insertions, 0 deletions
diff --git a/ChangeLog b/ChangeLog index 75f4c4a5217..051bff538e3 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,73 @@ +Sat Jun 2 13:05:20 2001 Douglas C. Schmidt <schmidt@tango.doc.wustl.edu> + + * ace/Timer_Wheel_T.cpp: Since we removed the call to + is_empty() from earliest_time() we need to modify the code in + reschedule() to test for is_empty. + + * tests/Cached_Accept_Conn_Test.cpp (accept_svc_handler): Use the + ACE_Accept_Strategy::acceptor() accessor method rather than + accessing the acceptor_ directly. + + * TAO/rules.tao.GNU (TAO_IDL_DEP), + * include/makeinclude/rules.local.GNU: Added support for the new EXEEXT + Makefile macro. Thanks to Cristian Ferretti + <cristian_ferretti@yahoo.com> for pointing this out. + + * include/makeinclude/platform_gnuwin32_common.GNU: Added a number of + fixes for the mingw platform. Thanks to Cristian Ferretti + <cristian_ferretti@yahoo.com> for pointing this out. + + * ace/Configuration.{h,cpp}: "Un-inlined" the == and != operators for + the ACE_Configuration_Win32Registry class. + + * ace/Configuration.cpp (get_binary_value): Changed the type of "data" + from "unsigned char" to "BYTE". Thanks to Cristian Ferretti + <cristian_ferretti@yahoo.com> for pointing this out. + + * ace/config-win32-mingw.h: Removed the #define ACE_AS_STATIC_LIBS + macro. Thanks to Cristian Ferretti + <cristian_ferretti@yahoo.com> for pointing this out. + + * ace/config-win32-common.h: Protect certain DLL export macros with + #if !defined (__MINGW32__). Thanks to Cristian Ferretti + <cristian_ferretti@yahoo.com> for pointing this out. + + * ace/ACE.cpp (handle_timed_complete): Added an ACE_UNUSED_ARG + for the is_tli flag for ACE_WIN32 to keep the compiler happy. + Thanks to Cristian Ferretti <cristian_ferretti@yahoo.com> for + pointing this out. + + * ace/Acceptor.cpp: Fixed the comment and logic in + accept_svc_handler() so that it now cleans up the svc handler + whenever accept() returns -1. Thanks to Boris Temkin + <borist@allcharge.com> for suggesting this and to Irfan for + helping clarify what needed to be done. + + * ace/Strategies_T.cpp: When the number of clients + connecting to a server exceeds the process handle limit the + server would busy loop. The ACE_Accept_Strategy + accept_svc_handler() method is called which (1) constructs a + handler and (2) attempts to accept the connection, which fails + (because there is no available handle). The handler is + destructed, but the connect request is still queued and so the + acceptor gets called back again.... and so on until one of the + client which is already connected disconnects and frees up a + descriptor. Changed the accept_svc_handler() method in + ACE_Accept_Strategy to close and reopen the connection when + accept() fails. A check is done to make sure address reuse is + set. So this fix will only work in case where it is set. Thanks + to Dominic Hughes <dominic@aersoft.ie> for this fix and to Irfan + for recognizing that it ought to be added to ACE_Accpt_Strategy. + + * ace/Strategies_T.cpp (open): Fixed the strategy acceptor so that it + uses non-blocking semantics to handle certain failure cases + correctly. This is a safe-guard against the race condition that + can otherwise occur between the time when <select> indicates + that a passive-mode socket handle is "ready" and when we call + <accept>. During this interval, the client can shutdown the + connection, in which case, the <accept> call can hang! Thanks + to Irfan for pointing this out. + Fri Jun 1 13:04:31 2001 Douglas C. Schmidt <schmidt@tango.doc.wustl.edu> * ace/Makefile: |