optimization/9301: mips-linux gcc 3.2.1 SEGV in try_forward_edges

optimization/9301: mips-linux gcc 3.2.1 SEGV in try_forward_edges

[cgd]

>Number:         9301
>Category:       optimization
>Synopsis:       mips-linux gcc 3.2.1 SEGV in try_forward_edges
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    unassigned
>State:          open
>Class:          ice-on-legal-code
>Submitter-Id:   net
>Arrival-Date:   Mon Jan 13 15:46:01 PST 2003
>Closed-Date:
>Last-Modified:
>Originator:     Chris Demetriou <cgd@broadcom.com>
>Release:        3.2.1
>Organization:
>Environment:
mips-linux, kernel 2.4.x.
>Description:
mips-linux (native) gcc 3.2.1 ICEs on the following test case.
gdb indicates a SEGV in try_forward_edges.

gcc 3.2.1 from ftp://ftp.gnu.org/pub/gnu/gcc/gcc-3.2.1/gcc-core-3.2.1.tar.gz
built on a mips linux system like:

	tar xvfz gcc-core-3.2.1.tar.gz
	mkdir obj inst
	cd obj
	../gcc-3.2.1/configure --prefix=`pwd`/../inst
	gmake && gmake install

test fails with -O2 or above, but not with -O1 or less.
>How-To-Repeat:
see above.

More info on where it does/doesn't fail:

tried 3.2.1 as a native compiler for x86-linux as well, no
ICE there at up to -O3.

3.2.1 also seems to have the problem on embedded mips targets
(e.g. mipsisa64sb1 8-), but I did not try that
configuration with an unmodified compiler.

a quick test of head-of-trunk with mipsisa32-elf
indicates problem is _not_ present there, but I wouldn't
call my testing comprehensive.  8-)
>Fix:
workaround is to use -O1 or less.  Fix unknown.
>Release-Note:
>Audit-Trail:
>Unformatted:
----gnatsweb-attachment----
Content-Type: application/octet-stream; name="trimmed3.c"
Content-Transfer-Encoding: base64
Content-Disposition: attachment; filename="trimmed3.c"

dm9pZCBiYXIodm9pZCk7Cgp2b2lkIGZvbyAoaW50IGEsIGludCBiLCBpbnQgYywgaW50IGQsIGlu
dCBlKQp7CgogICAgaWYgKGEpCiAgICAgIGJhcigpOwogICAgaWYgKGIgJiYgYykKICAgICAgOwog
ICAgaWYgKGQgJiYgZSkKICAgICAgOwp9Cg==

Re: optimization/9301: mips-linux gcc 3.2.1 SEGV in try_forward_edges

[cgd]

The following reply was made to PR optimization/9301; it has been noted by GNATS.

From: cgd@broadcom.com
To: gcc-gnats@gcc.gnu.org
Cc: gcc-bugs@gcc.gnu.org
Subject: Re: optimization/9301: mips-linux gcc 3.2.1 SEGV in
 try_forward_edges
Date: 21 Jan 2003 11:52:06 -0800

 At Mon, 13 Jan 2003 23:46:29 +0000 (UTC), cgd@broadcom.com wrote:
 > >Number:         9301
 
 > 3.2.1 also seems to have the problem on embedded mips targets
 > (e.g. mipsisa64sb1 8-), but I did not try that
 > configuration with an unmodified compiler.
 
 FYI, i tried the same test with some older versions of GCC, slightly
 modified from stock sources as well (same mods as in the 3.2.1
 mentioned above, which shouldn't affect this), and
 
 * "gcc version 3.0.3 with SiByte modifications" worked fine.
   (same for 3.0.4)
 
 * "gcc version 3.1.1 with SiByte modifications" also crashed.
 
 So, this seems to be a regression from previous releases (at least
 3.0.x).
 
 I'm can provide the diffs against unmodified gcc from any of those
 "with SiByte modifications" versions to the sources, if people want
 them, but really, they're not the problem.  8-)
 
 
 cgd
 
 

Re: optimization/9301: mips-linux gcc 3.2.1 SEGV in try_forward_edges

[cgd]

The following reply was made to PR optimization/9301; it has been noted by GNATS.

From: cgd@broadcom.com
To: gcc-gnats@gcc.gnu.org
Cc:  
Subject: Re: optimization/9301: mips-linux gcc 3.2.1 SEGV in
 try_forward_edges
Date: 13 Jan 2003 15:47:35 -0800

 Of course, wouldn't do to include all of the relevant info in the
 intial submission:
 
 command line to cause the problem is:
 
 	gcc -O2 -c ~/trimmed3.c
 
 which results in:
 
 /home/cgd/trimmed3.c: In function `foo':
 /home/cgd/trimmed3.c:12: internal error: Segmentation fault
 Please submit a full bug report,
 with preprocessed source if appropriate.
 See <URL:http://www.gnu.org/software/gcc/bugs.html> for instructions.
 
 
 easy enough.  8-)
 
 
 chris