Re: M4 1.4.14 on AIX 7.1BETA

[Rainer Tammer]

 Hello,

On 29.07.2010 13:56, Eric Blake wrote:
> [Please post to public mailing lists, not to random developers,
> especially since you have no guarantee whether a random developer's
> address will still be valid - redirecting to bug-gnulib]
>
> On 07/29/2010 05:03 AM, Rainer Tammer wrote:
>> Hello,
>> I now have tested M4 1.4.14 on AIX 7.1BETA. I have replaced
>> build-aux/config.guess with the new AIX 7.1 aware version.
> If you post config.guess patches to the correct upstream location
> (config-patches AT gnu DOT org), then the next release of m4 will
> automatically include it. 
I just did that 10 minutes ago :-))
>  Meanwhile, your remaining bugs both relate to
> code that m4 borrows from gnulib, so fixing it in gnulib will aid in
> porting other GNU projects to AIX.
>
OK,
I already CCed Bruno Haible, we worked together on the AIX POSIX spawn
problem in the first place.
>> I also have
>> changed m4/posix_spawn.m4 so that
>> "gl_cv_func_posix_spawn_works="guessing yes";;" is set. According to my
>> tests the old POSIX spawn problem (which was present in early AIX
>> 5.3/6.1 versions) is gone.
> That would actually be a problem to report to bug-gnulib, but why is
> configure rejecting the implementation if you don't seed the cache?
>
That`s a good question. The configure log says:

  gl_cv_func_posix_spawn_works=yes


>> The make run shown one problem:
>>
>>           source='execute.c' object='execute.o' libtool=no  DEPDIR=.deps
>> depmode=aix /bin/sh ../build-aux/
>> depcomp  cc -qlanglvl=extc89 -qlanglvl=extc99  -I.      -g -c execute.c
>> "./spawn.h", line 141.10: 1506-236 (W) Macro name POSIX_SPAWN_RESETIDS
>> has been redefined.
> Hmm - even though you requested the posix_spawn override, you either did
> not request enough other overrides, so your incomplete cache seeding is
> causing interference with the rest of things gnulib is trying to
> guarantee, or there's still a bug in AIX posix_spawn that needs working
> around, but in a way that gnulib is not yet prepared for.
>
This might be possible, I'll need to check this again.
>> src/spawn.h
>>
>> # define POSIX_SPAWN_RESETIDS           0x01
>> # define POSIX_SPAWN_SETPGROUP          0x02
>> # define POSIX_SPAWN_SETSIGDEF          0x04
>> # define POSIX_SPAWN_SETSIGMASK         0x08
>> # define POSIX_SPAWN_SETSCHEDPARAM      0x10
>> # define POSIX_SPAWN_SETSCHEDULER       0x20
>>
>> AIX 7.1 defines spawn.h:
>>
>> #define POSIX_SPAWN_SETPGROUP       0x1
>> #define POSIX_SPAWN_SETSIGMASK      0x2
>> #define POSIX_SPAWN_SETSIGDEF       0x4
>> #define POSIX_SPAWN_SETSCHEDULER    0x8
>> #define POSIX_SPAWN_SETSCHEDPARAM   0x10
>> #define POSIX_SPAWN_RESETIDS        0x20
>> #define POSIX_SPAWN_FORK_HANDLERS   0x1000
>>
>>
>>
>> The make check run yields one problem:
>>
>>   /bin/sh: 5505092 IOT/Abort trap(coredump)
>>   FAIL: test-strtod
>>
>> Is this a know problem ?
> Did configure try to replace AIX' strtod?
According to config.h:

/* Define to 1 if strtod is declared even after undefining macros. */
#define HAVE_RAW_DECL_STRTOD 1

>   Which line is it failing on?
FAIL: test-strtod (exit: 134)
=============================

test-strtod.c:234: assertion failed
>  Chances are there's a bug in AIX' strtod with C99 compliance that is
> not yet being detected at configure time, but which we can work around
> if we know what the bug is.
>
That is possible.

> Can you also post the relevant config.log bits for the posix_spawn- and
> strtod-related probes?
>

I had attached all log files to my last mail. I'll attache the logs
again: m4-1.4.14.tar.gz

Bye
  Rainer

m4-1.4.14.tar.gz
Description: GNU Zip compressed data