From xemacs-m  Wed Sep 24 13:00:01 1997
Received: from fronsac.ensg.u-nancy.fr (fronsac.ensg.u-nancy.fr [192.93.48.17])
	by xemacs.org (8.8.5/8.8.5) with ESMTP id MAA02787
	for <xemacs-beta@xemacs.org>; Wed, 24 Sep 1997 12:59:59 -0500 (CDT)
Message-Id: <199709241759.MAA02787@xemacs.org>
Received: by fronsac.ensg.u-nancy.fr
	(8.7.1/16.2) id NAA19496; Wed, 24 Sep 1997 13:59:59 -0400 (EDT)
Date: Wed, 24 Sep 1997 13:59:59 -0400 (EDT)
From: Richard Cognot <cognot@ensg.u-nancy.fr>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
To: xemacs-beta@xemacs.org
Subject: [PATCH?] Re: More info 19.16 crash on Irix 6.X
In-Reply-To: <199709241136.GAA27246@xemacs.org>
References: <199709240952.EAA26808@xemacs.org>
	<199709241136.GAA27246@xemacs.org>
X-Mailer: VM 6.34 under 19.16 "Bronx" XEmacs Lucid
Reply-to: cognot@ensg.u-nancy.fr
X-Face:  .|{6#t`YCBNfg_E.8;@IFK9kd'Ol7>~2S7U!o3+g)+\`hV5&I]k,UwC%g%Y\,-KV+[eEgZm
 i(NgMB@L_n/A!jk;}@!?$<t5Aw`B$R=xJSv[F$2&sz*cwxF!|B3)MH,6YBDNU!$9;91N_p*>}</%ZO
 +EYRSc{a_#KXB)sJlPpxQq,/:}\*noO!;(5PY~MTnspl;&XA$JccP;N~;v5E>Yh.


Following up to myself here:

>>>>> "Me" == Richard Cognot <cognot@ensg.u-nancy.fr> writes:

    Me> I've run XEmacs 19.16 under purify. Although that prevents it
    Me> from crashing in the same circumstances, I get these errors
    Me> repeatedly:

    Me>       UMR: Uninitialized memory read (260 times) This is
    Me> occurring while in: _setcontext [syscalls.c:9354] _siglongjmp
    Me> [siglongjmp.c:43] unwind_to_catch [eval.c:1406]
[...]
    Me> And others, all going through LONGJMP.

    Me> and HAVE_SIGSETJMP is set to 1.

    Me> 2 questions: 1) should I try to undef HAVE_SIGSETJMP?

I've just tried this, and sure enough, both the messages from purify
and the core goes away...

Now, the question is, should I add this to src/s/irix6-0.h? If yes,
here is the patch. Note that it is done only for the N32 ABI. 

*** irix6-0.h.orig      Wed Sep 24 18:55:04 1997
--- irix6-0.h   Wed Sep 24 18:55:17 1997
***************
*** 26,28 ****
--- 26,37 ----
   */
  #define HAVE_UINTPTR_T 1
  
+ /* R. Cognot 09/24/97
+  * This may be needed for other ABIs, but at least I'm sure it is
+  * is needed on n32, as purify reports UMRs in siglongjmp and 
+  * xemacs dumps core every once in a while...
+  */
+ #if (_MIPS_SIM==_MIPS_SIM_ABIN32)
+ #undef HAVE_SIGSETJMP
+ #endif
+ 

-- 
|-------------------------------------------------------------|
| Richard Cognot           | Proceed, with fingers crossed... |
|                          |           /\^^/\                 |
| <cognot@ensg.u-nancy.fr> |             `'                   |
|-------------------------------------------------------------|
| http://www.ensg.u-nancy.fr/~cognot                          |
|-------------------------------------------------------------|

