ACE 5.0 and GCC 2.95 Problems

["Preston A. Elder" <prez at goth dot net>]

Hi,

I'm writing this to report some bugs I've found in both ACE 5.0 and
GCC 2.95 when compiling under both SunOS 5.6 and Linux 2.2.10 (glibc
2.1.1).  I'll list the platform, program and error message.

I can supply the source when whoever from GCC and ACE reply, but the
program we are developing is not finished yet, and not released, so
I dont like releasing the source code to mailing lists.

I'm using the following command to compile all files:
gcc -DMAGICK_HAS_EXCEPTIONS -O -I/usr/local/include -I/home/admin/prez/projects/Magick/Magick-II/include -g -c ircsocket.cpp


SunOS -- GCC:
=============
/usr/include/g++-3/stl_relops.h: In function `bool operator !=<wxExpandEnvVars(const mstring &)::Bracket>(const wxExpandEnvVars(const mstring &)::Bracket &, const wxExpandEnvVars(const mstring &)::Bracket &)':
confbase.cpp:101:   instantiated from here
/usr/include/g++-3/stl_relops.h:38: Internal compiler error in `dwarfout_file_scope_decl', at dwarfout.c:5336
Please submit a full bug report.
See <URL:http://egcs.cygnus.com/faq.html#bugreport> for instructions.
make[1]: *** [confbase.o] Error 1


SunOS and Linux -- ACE:
=======================
GCC people -- Was talking to my fellow developer.  Originally new was just
a wrapper around malloc, and malloc used to return NULL when it couldnt
allocate the memory.  Shouldnt new keep that for backward compatability?

ACE people, you might want to add in detection of GCC 2.95 or something :)
-----------------------
/usr/local/include/ace/Svc_Handler.cpp: In function `static void * ACE_Svc_Handler<ACE_SOCK_Stream,ACE_MT_SYNCH>::operator new(unsigned int)':
ircsocket.cpp:102:   instantiated from here
/usr/local/include/ace/Svc_Handler.cpp:41: operator new should throw an exception, not return NULL
make[1]: *** [ircsocket.o] Error 1


SunOS 5.6 -- GCC:
=================
/usr/include/g++-3/stl_map.h:76: Internal compiler error in `dwarfout_finish', at dwarfout.c:6097
Please submit a full bug report.
See <URL:http://egcs.cygnus.com/faq.html#bugreport> for instructions.
make[1]: *** [server.o] Error 1


SunOS -- ACE/GCC:
=================
I also got the same problem with localtime_r (8960), gmtime_r (8997) and asctime_r (9045).
On linux, I got a problem with cuserid (1180).

I fixed the problem by removing the '::' from before the function calls in question.
Dont know if GCC should be allowing ::cuserid(), or if ACE should remove the ::'s.
---------------------
/usr/local/include/ace/OS.i: In function `static char * ACE_OS::ctime_r(const time_t *, char *, int)':
/usr/local/include/ace/OS.i:8917: implicit declaration of function 	nt ctime_r(...)'
/usr/local/include/ace/OS.i:8917: return to `char *' from 	nt' lacks a cast


I could repeat ALL of the above.  We also got problems on SunOS with
our own implementation of a triplet -- it was fairly standard, but we
kept getting compiler errors on sunos -- but I dont have the actual
output available right now -- I'll go into that more when someone
replies.

Another problem I found on Linux with GCC -- when trying to compile
this program with -O, -O2 or -O3 it will go for a while, then start
taking up CPU resources like anything, and eventually, either just
use up all resources, or exit out with out of virtual memory.  I found
out that to fix this I had to add -fno-default-inline, and it worked
fine.  The particular file in question it always died on is a simply
doing (for the most part), a switch function, with if's inbetween them.
Once I saw it die, and say 'out of memory', and the line number it
gave where it died was a COMMENT.  I can also supply this file aswell.

Anyway, thanks for your time and patience, I hope to hear back from both
GCC and ACE development teams soon :)


_O_ PreZ :)                           Finger prez@srealm.net.au _O_
 |  prez@goth.net                      for full PGP Public Key   | 

PGP FingerPrint: B3 0C F3 32 DE 5A 7D 90  26 F6 FA 38 CC 0A 2D D8
Goth Code '98:   tSKeba5qaSabsaaaGbaa75KAASWGuajmsvbieqcL4BaaLb3F4
                 nId5mefqmDjmmgm#haxthgzpj4GiysNkycSRGHabiabOkauNSW
Magick Code .92: MWI S W++ N+ PCE/NA++ D A a+ C G QO+ 666 Y