[VANILLA-L:1087] Re: [VANILLA-L:1086] Problem with 2.8pl0

[Jeffrey Nowakowski <jeffno@ccs.neu.edu>]

Steve Franks <sdfranks@mit.edu> wrote:

> I recently got one of my boxes "upgraded" to Solaris 2.6.  Now there's
> a pretty major bug in the server.  Any time a player is killed, his 
> slot seems to be freed.  There is no core dump, nothing in the log
> files.  The player just disappears.

Mikael reported this problem in vanilla-l:862.  He downgraded from gcc
2.8.1 to 2.7.x to get it working.

I compiled Server 2.8pl0 on a Solaris 2.6 box using gcc 2.8.1, and
yup, had the same problem.  The ntserv process left a core file, and
the problem was at the same place Mikael reported it:

#0  0x1bb08 in main (argc=1, argv=0xefffecb8) at main.c:266

I recompiled without optimizations and that fixed the problem.

So there's probably some problem with setjmp/longjmp and
optimizations.  Yet another reason to turn off optimization when
compiling.

James, could you change system.mk.in to not put in -O by default, and
to put in -g by default?  Seems silly to submit a patch just for that.

-Jeff
+
++ Vanilla-l Mailing List ++
To unsubscribe: send "unsubscribe vanilla-l" to majordomo@real-time.com
For more information: http://archives.real-time.com

---

Follow-Ups:

• [VANILLA-L:1089] Re: Problem with 2.8pl0
• From: James Cameron <cameron@stl.dec.com>

References:

• [VANILLA-L:1086] Problem with 2.8pl0
• From: Steve Franks <sdfranks@mit.edu>

---

• Prev by Date: [VANILLA-L:1088] Re: [VANILLA-L:1084] mailing lists & dns entries
• Next by Date: [VANILLA-L:1086] Problem with 2.8pl0
• Prev by thread: [VANILLA-L:1086] Problem with 2.8pl0
• Next by thread: [VANILLA-L:1089] Re: Problem with 2.8pl0
• Index(es):
  • Main
  • Thread