[Bug c/13472] New: optimizer generates code to store data in const int, resulting in a segfault

[Bug c/13472] New: optimizer generates code to store data in const int, resulting in a segfault

[p dot van-hoof at qub dot ac dot uk]

When the attached program is compiled with the following options:

dogbert> gcc -v -pedantic -march=i686 -O2 r.c
Reading specs from /usr/local/gcc333/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/specs
Configured with: ../gcc/configure --prefix=/usr/local/gcc333
--enable-languages=c,c++,f77
Thread model: posix
gcc version 3.3.3 20031223 (prerelease)
 /usr/local/gcc333/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/cc1 -quiet -v
-D__GNUC__=3 -D__GNUC_MINOR__=3 -D__GNUC_PATCHLEVEL__=3 r.c -quiet -dumpbase r.c
-march=i686 -auxbase r -O2 -pedantic -version -o /tmp/cccX38fK.s
GNU C version 3.3.3 20031223 (prerelease) (i686-pc-linux-gnu)
        compiled by GNU C version 3.3.3 20031223 (prerelease).
GGC heuristics: --param ggc-min-expand=98 --param ggc-min-heapsize=128958
ignoring nonexistent directory "/usr/local/gcc333/i686-pc-linux-gnu/include"
#include "..." search starts here:
#include <...> search starts here:
 /usr/local/include
 /usr/local/gcc333/include
 /usr/local/gcc333/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/include
 /usr/include
End of search list.
 as -V -Qy -o /tmp/ccgjNrLm.o /tmp/cccX38fK.s
GNU assembler version 2.14.90.0.5 (i586-suse-linux) using BFD version
2.14.90.0.5 20030722 (SuSE Linux)
 /usr/local/gcc333/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/collect2 --eh-frame-hdr
-m elf_i386 -dynamic-linker /lib/ld-linux.so.2 /usr/lib/crt1.o /usr/lib/crti.o
/usr/local/gcc333/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/crtbegin.o
-L/usr/local/gcc333/lib/gcc-lib/i686-pc-linux-gnu/3.3.3
-L/usr/local/gcc333/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/../../.. /tmp/ccgjNrLm.o
-lgcc -lgcc_eh -lc -lgcc -lgcc_eh
/usr/local/gcc333/lib/gcc-lib/i686-pc-linux-gnu/3.3.3/crtend.o /usr/lib/crtn.o
dogbert> a.out
Segmentation fault

All command line parameters are essential to produce the bug (including
-pedantic; I wasn't aware that this switch affected code generation...).

The bug is triggered for all of the following platforms: i686, pentium2, athlon,
athlon-tbird.

The bug occurs with gcc 3.3.3, 3.3.2 and 3.3.1, but not with gcc 3.2.3.

Upon inspecting the assembler output it appears that the optimizer generates
code that tries to overwrite the const int q.

-- 
           Summary: optimizer generates code to store data in const int,
                    resulting in a segfault
           Product: gcc
           Version: 3.3.3
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: p dot van-hoof at qub dot ac dot uk
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug c/13472] optimizer generates code to store data in const int, resulting in a segfault

[p dot van-hoof at qub dot ac dot uk]

------- Additional Comments From p dot van-hoof at qub dot ac dot uk  2003-12-23 02:10 -------
Created an attachment (id=5357)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=5357&action=view)
code that triggers the bug


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-23 05:49 -------
pentanic changes how global const int are handled, really const and replace the variable with a 
constant which changes the code.  I do not know where the problem is.  I do know that the 
mainline does not seg fault.
Confirmed.  I am wondering if GCC loses track that q is constant.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
          Component|c                           |optimization
     Ever Confirmed|                            |1
           Keywords|                            |wrong-code
   Last reconfirmed|0000-00-00 00:00:00         |2003-12-23 05:49:29
               date|                            |
            Summary|optimizer generates code to |[3.3 Regression] optimizer
                   |store data in const int,    |generates code to store data
                   |resulting in a segfault     |in const int, resulting in a
                   |                            |segfault
   Target Milestone|---                         |3.3.3


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2003-12-23 14:19 -------
The final culprit is the reload pass: it generates the store (although the MEM
is marked as unchanging) because it thinks the memory location is a regular
spill slot.  It appears that the reload pass doesn't care about the /u flag when
recording equivalent memory locations based on the presence of REG_EQUIV notes.

But I don't know if this is really a reload problem: maybe reload implicitly
expects full equivalence between objects in pairs that carry REG_EQUIV notes.

In any cases, the dependence upon -pedantic is certainly a bug and is actually
recognized as such in the code (c-typeck.c:790):

/* Return either DECL or its known constant value (if it has one), but
   return DECL if pedantic or DECL has mode BLKmode.  This is for
   bug-compatibility with the old behavior of decl_constant_value
   (before GCC 3.0); every use of this function is a bug and it should
   be removed before GCC 3.1.  It is not appropriate to use pedantic
   in a way that affects optimization, and BLKmode is probably not the
   right test for avoiding misoptimizations either.  */

static tree
decl_constant_value_for_broken_optimization (decl)
     tree decl;
{
  if (pedantic || DECL_MODE (decl) == BLKmode)
    return decl;
  else
    return decl_constant_value (decl);
}


So I think the fix is to remove decl_constant_value_for_broken_optimization
altogether from the C front-end.  But this is probably not doable on a release
branch.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |ebotcazou at gcc dot gnu dot
                   |dot org                     |org
             Status|NEW                         |ASSIGNED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |critical


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2003-12-26 05:53 -------
Patch here: <http://gcc.gnu.org/ml/gcc-patches/2003-12/msg02005.html>

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-01-07 08:41 -------
Confirmed, thanks for this second testcase.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|2003-12-23 05:49:29         |2004-01-07 08:41:31
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-01-11 21:27 -------
newer patch here: <http://gcc.gnu.org/ml/gcc-patches/2004-01/msg00769.html>

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[dhazeghi at yahoo dot com]

------- Additional Comments From dhazeghi at yahoo dot com  2004-01-21 03:29 -------
Eric, was your patch approved? Thanks.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[gdr at integrable-solutions dot net]

------- Additional Comments From gdr at integrable-solutions dot net  2004-01-21 03:35 -------
Subject: Re:  [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

"dhazeghi at yahoo dot com" <gcc-bugzilla@gcc.gnu.org> writes:

| Eric, was your patch approved? Thanks.

Yes, I think it was.  Eric, please could you put all that in the usual
form, or even better apply it?  Thanks,

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-01-21 07:22 -------
No, I don't want to apply it until the branch has fully recovered from its
January chaos.


This patch

2004-01-16  Jan Hubicka  <jh@suse.cz>

        PR opt/11350
        * cfgcleanup.c (try_optimize_cfg): Suppress tablejump removal
	after reload.
        * cfgrtl.c (try_redirect_by_replacing_branch): Likewise.

caused a disaster on mainline on anything else than x86.


This patch

2004-01-15  J"orn Rennecke <joern.rennecke@superh.com>

	PR optimization/10392
	* reload.c (can_reload_into): New function.
	(push_reload): Use it.

should never have been put on the branch at this point.


I found this patch

2004-01-09  Richard Henderson  <rth@redhat.com>

	* recog.c (constrain_operands): Validate mem operands.

very risky too.


So don't count on me to put a patch to reload on the branch.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-01-21 07:36 -------
The original problem can be easily worked around by not passing -pedantic.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-01-21 07:37 -------
*** Bug 13634 has been marked as a duplicate of this bug. ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |rosbacke at nada dot kth dot
                   |                            |se


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[gdr at integrable-solutions dot net]

------- Additional Comments From gdr at integrable-solutions dot net  2004-01-21 07:38 -------
Subject: Re:  [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

"ebotcazou at gcc dot gnu dot org" <gcc-bugzilla@gcc.gnu.org> writes:

| No, I don't want to apply it until the branch has fully recovered from its
| January chaos.

I don't understand this part of your comment.

| This patch
| 
| 2004-01-16  Jan Hubicka  <jh@suse.cz>
| 
|         PR opt/11350
|         * cfgcleanup.c (try_optimize_cfg): Suppress tablejump removal
| 	after reload.
|         * cfgrtl.c (try_redirect_by_replacing_branch): Likewise.
| 
| caused a disaster on mainline on anything else than x86.

Do you see similar disaster on branch?

[...]

| So don't count on me to put a patch to reload on the branch.

I hear you.  Thanks,

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-01-21 08:35 -------
> I don't understand this part of your comment.

Until early January, bugfixes on the 3.3 branch had been very localized, either
in a specific back-end or front-end or optimization pass that can be disabled. 
Then, suddendly, central parts of the compiler that are run at any optimization
levels started to be touched:
- I think Richard's patch has a potential effect on every memory pattern in
every back-end,
- Joern's reload patch is... a reload patch and, judging by him asking you to
put his other reload patch (fix for PR 11864) to the branch, this latter patch
having a serious impact on the back-end (Alpha, PA, MIPS and SPARC at least),
I'm not sure that he fully evaluated the impact of this patch,
- Jan's patch deals with a very hot topic: table jump removal.  We have been
fighting this problem for a long time and I was under the impression it is more
or less under control on the 3.3 branch.

> Do you see similar disaster on branch?

No.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[gdr at integrable-solutions dot net]

------- Additional Comments From gdr at integrable-solutions dot net  2004-01-21 14:06 -------
Subject: Re:  [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

"ebotcazou at gcc dot gnu dot org" <gcc-bugzilla@gcc.gnu.org> writes:

| > I don't understand this part of your comment.
| 
| Until early January, bugfixes on the 3.3 branch had been very
| localized, either in a specific back-end or front-end or
| optimization pass that can be disabled. Then, suddendly, central
| parts of the compiler that are run at any optimization levels
| started to be touched: 

Thanks for explaining yourself.  But I would not have used the word
"chaos", though. 

-- Gaby


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-01-21 14:33 -------
I concede that the wording was a bit harsh.  But I was a bit frustrated to see
these three IMHO risky patches on the branch.  Sorry.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[gdr at gcc dot gnu dot org]

------- Additional Comments From gdr at gcc dot gnu dot org  2004-01-29 05:04 -------
Potsponed until 3.3.4

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|critical                    |normal


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[pinskia at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|3.3.3                       |3.3.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[ebotcazou at gcc dot gnu dot org]

------- Additional Comments From ebotcazou at gcc dot gnu dot org  2004-03-08 07:13 -------
The final patch is at http://gcc.gnu.org/ml/gcc-patches/2004-01/msg01296.html

But I'm still not convinced it is appropriate for the 3.3 branch, nor that it is
the real fix for the problem.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|ebotcazou at gcc dot gnu dot|unassigned at gcc dot gnu
                   |org                         |dot org
             Status|ASSIGNED                    |NEW


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2004-03-13 00:31 -------
Subject: Bug 13472

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_3-branch
Changes by:	gdr@gcc.gnu.org	2004-03-13 00:30:54

Modified files:
	gcc            : ChangeLog reload1.c 
Added files:
	gcc/testsuite/gcc.dg: const-1.c 

Log message:
	Endorse:
	2004-01-10  Eric Botcazou  <ebotcazou@libertysurf.fr>
	PR optimization/13472
	* reload.c (reload): Don't record unchanging memory locations.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.16114.2.935&r2=1.16114.2.936
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/reload1.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=1.366.2.9&r2=1.366.2.10
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/const-1.c.diff?cvsroot=gcc&only_with_tag=gcc-3_3-branch&r1=NONE&r2=1.1.2.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[gdr at gcc dot gnu dot org]

------- Additional Comments From gdr at gcc dot gnu dot org  2004-03-13 00:35 -------
http://gcc.gnu.org/ml/gcc-patches/2004-03/msg01081.html


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472

[Bug optimization/13472] [3.3 Regression] optimizer generates code to store data in const int, resulting in a segfault

[pinskia at gcc dot gnu dot org]

------- Additional Comments From pinskia at gcc dot gnu dot org  2004-04-18 21:50 -------
*** Bug 12271 has been marked as a duplicate of this bug. ***

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fargaud at yahoo dot fr


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=13472