Re: Threads with AIX (was RE: gtk+ on IBM AIX )

[Valdis Kletnieks vt edu]

On Thu, 08 Jun 2000 15:21:51 +1000, ALLA Raphael said:
> AC_CHECK_HEADER(pthread.h,have_threads=posix) does not work on AIX because
> the pthread.h header file makes gcc complaining that there is no whitespace
> where it should:
> 
> In file included from configure:5395:
> /usr/include/pthread.h:113: warning: missing white space after `#define
> PTHREAD_MUTEX_INITIALIZER'
> /usr/include/pthread.h:128: warning: missing white space after `#define
> PTHREAD_COND_INITIALIZER'
> /usr/include/pthread.h:147: warning: missing white space after `#define
> PTHREAD_RWLOCK_INITIALIZER'
> /usr/include/pthread.h:157: warning: missing white space after `#define
> PTHREAD_ONCE_INIT'
> 
> A solution is to replace line 638 of configure.in by:
>   AC_TRY_COMPILE([#include <pthread.h>],[], have_threads=posix)

(Just so everybody is on the same page, this is actually all in the
Glib configure, not in Gtk+ itself).

Umm.. Hrum... The line in question in config.in currently reads
	AC_CHECK_HEADER(pthread.h, have_threads=posix)
in both my 1.2.7 and 1.2.8 source trees, and detect the Posix
pthreads just fine using the IBM 'ibmcxx' compiler.

At least in my source tree, line 638 of configure.in ends up
around line 4583 of configure, checking for the presense of pthread.h
I know that autoconf is strange and wonderful, but 800 lines
off? ;)

I'm wondering if this isn't a gcc problem with the GCC cpp
and/or fixincludes.   I don't *SEE* an error message there,
warnings only.  Could you go back and check the config.log
and paste in ALL the relevant output?  At the very least, could
you show the actual 'configure' code (from the 'cat > conftest.$ac_ext'
through the if/then/else where it tries it out and then possibly
cats onto the end of confdefs.h, so we can identify *which* test
it's dying on?

				Valdis Kletnieks
				Operating Systems Analyst
				Virginia Tech