Recent changes to patches

Free DCE forking bug

Anonymous — Tue, 29 Jul 2003 20:45:03 -0000

We have discovered a bug in the current release of
Free DCE which causes an unhandled exception and
crash when an application using Free DCE forks.

The exception (pthread_badparam_e) is thrown by
pthread_detach in cases where it is called immediately
after a call to pthread_join on the same thread. It is
thrown because, in newer versions of pthreads, the
pthread_join call automatically detaches the thread.
The exception is harmless, it indicates an attempt to
detach an already detached thread.

The fix we have implemented is to simply catch and
ignore the exception.

These modules are affected:

/ncklib/transport/dg/dgclive.c
/ncklib/transport/dg/dgslive.c
/ncklib/transport/dg/dglsn.c
/ncklib/com/comnlst.c
/ncklib/rpc/rpctimer.c

We can provide modified versions of these modules on
request.

Bill Sudbrink
Senior Software Engineer
Docucorp International
bsudbrink@docucorp.com

yacc file fix for RH 9.0

Anonymous — Wed, 16 Jul 2003 09:49:16 -0000

Compilation error while compiling DCE-RPC on RH 9.0.
There is a missing semi-colons (line 309 and 944) in the
nidl_y.y file. I fixed it and now source can compile on RH
8.0 and RH 9.0.

Clobber fix for RPC server

Mark Brown — Tue, 01 Jan 2002 19:07:53 -0000

This fixes another PowerPC compile failure, though depending
on the whim of GCC I guess it could show up anywhere.
The failure is:

| gcc -DPACKAGE=\&quot;freedce\&quot; -DVERSION=\&quot;1.1.0.7\&quot; -DHAVE_DLFCN_H=1 -DYYTEXT_POINTER=1
-DHAVE_LIBPTHREAD=1 -DHAVE_FEATURES_H=1 -DHAVE_SYS_CDEFS_H=1 -D_REENTRANT=1
-D_GNU_SOURCE=1 -DDEBUG=1 -DENABLE_DCOM=1 -DHAVE_GETTIMEOFDAY=1
-DHAVE_SOCKET=1 -DPROT_NCACN=1 -DPROT_NCADG=1 -DNAF_IP=1 -DUSE_LOOPBACK=1 -I.
-I. -I../include -g -Wall -W -O -pipe -Werror -DNCK -D_POSIX_SOURCE=1 -I../ncklib/include
-I../ncklib/include/linux-gnu -c rpcdsliv.c
| cc1: warnings being treated as errors
| rpcdsliv.c: In function `sliv_task2':
| rpcdsliv.c:223: warning: variable `have_db_lock' might be clobbered by
| `longjmp' or `vfork'

which is down to the setjmp()/longjmp() exception handling
implementation. The patch fixes this by marking have_db_lock
as DO_NOT_CLOBBER.

Thinko fix for PowerPC

Mark Brown — Tue, 01 Jan 2002 19:02:00 -0000

On PowerPC systems freedce fails to build due to a thinko
in the definition of NDR_LOCAL_INT_REP. This patch fixes
that (assuming I have my big and little endians the right
way round).