[Bug regression/62102] New: [4.10 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[Bug regression/62102] New: [4.10 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[hp at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

            Bug ID: 62102
           Summary: [4.10 Regression]: gcc.dg/torture/pr48953.c -O2 -flto
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code
          Severity: normal
          Priority: P3
         Component: regression
          Assignee: unassigned at gcc dot gnu.org
          Reporter: hp at gcc dot gnu.org
                CC: hubicka at ucw dot cz
              Host: x86_64-unknown-linux-gnu
            Target: cris-axis-elf

These tests previously passed, now they fail.
A patch in the revision range (last_known_working:first_known_failing)
212108:212117 exposed or caused these regressions.  Yes, I know this is more
than a month ago, sorry for the delay.  Since then they fail as follows:

Running /tmp/hpautotest-gcc1/gcc/gcc/testsuite/gcc.dg/torture/dg-torture.exp
...
FAIL: gcc.dg/torture/pr48953.c  -O2 -flto -fno-use-linker-plugin
-flto-partition=none  (internal compiler error)
FAIL: gcc.dg/torture/pr48953.c  -O2 -flto -fno-use-linker-plugin
-flto-partition=none  (test for excess errors)
FAIL: gcc.dg/torture/pr48953.c  -O2 -flto -fuse-linker-plugin
-fno-fat-lto-objects  (internal compiler error)
FAIL: gcc.dg/torture/pr48953.c  -O2 -flto -fuse-linker-plugin
-fno-fat-lto-objects  (test for excess errors)
WARNING: gcc.dg/torture/pr48953.c  -O2 -flto -fuse-linker-plugin
-fno-fat-lto-objects  compilation failed to produce executable

Messages in gcc.log (from r213837):

Executing on host: /tmp/hpautotest-gcc1/cris-elf/gccobj/gcc/xgcc
-B/tmp/hpautotest-gcc1/cris-elf/gccobj/gcc/
/tmp/hpautotest-gcc1/gcc/gcc/testsuite/gcc.dg/torture/pr48953.c 
-fno-diagnostics-show-caret -fdiagnostics-color=never    -O2 -flto
-fno-use-linker-plugin -flto-partition=none  -fno-tree-dce   -isystem
/tmp/hpautotest-gcc1/cris-elf/gccobj/cris-elf/./newlib/targ-include -isystem
/tmp/hpautotest-gcc1/gcc/newlib/libc/include
-B/tmp/hpautotest-gcc1/cris-elf/gccobj/cris-elf/./libgloss/cris/
-L/tmp/hpautotest-gcc1/cris-elf/gccobj/cris-elf/./libgloss/cris
-L/tmp/hpautotest-gcc1/gcc/libgloss/cris 
-B/tmp/hpautotest-gcc1/cris-elf/gccobj/cris-elf/./newlib/
-L/tmp/hpautotest-gcc1/cris-elf/gccobj/cris-elf/./newlib -sim3  -lm    -o
./pr48953.exe    (timeout = 300)
/tmp/hpautotest-gcc1/gcc/gcc/testsuite/gcc.dg/torture/pr48953.c: In function
'main':
/tmp/hpautotest-gcc1/gcc/gcc/testsuite/gcc.dg/torture/pr48953.c:14:5: error:
type mismatch in array reference
struct S

struct S

# .MEM_5 = VDEF <.MEM_4>
MEM[(struct S[2] *)&D.1358][0]{lb: 0 sz: 12}.value = 0;
/tmp/hpautotest-gcc1/gcc/gcc/testsuite/gcc.dg/torture/pr48953.c:14:5: internal
compiler error: verify_gimple failed
0x8a872c verify_gimple_in_cfg(function*, bool)
        /tmp/hpautotest-gcc1/gcc/gcc/tree-cfg.c:4994
0x7dc2c1 execute_function_todo
        /tmp/hpautotest-gcc1/gcc/gcc/passes.c:1749
0x7dd783 execute_todo
        /tmp/hpautotest-gcc1/gcc/gcc/passes.c:1806
Please submit a full bug report,
with preprocessed source if appropriate.
Please include the complete backtrace with any bug report.
See <http://gcc.gnu.org/bugs.html> for instructions.
lto-wrapper: fatal error: /tmp/hpautotest-gcc1/cris-elf/gccobj/gcc/xgcc
returned 1 exit status
compilation terminated.
collect2: fatal error: lto-wrapper returned 1 exit status
compilation terminated.
compiler exited with status 1

Author of suspect patches in revision range CC:ed.
The test passes on the 4.9 branch.
Unfortunately, the failure is appearently LTO-specific and only exposed on this
target, from a quick browse of gcc-testresults archives.

[Bug regression/62102] [4.10 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

Richard Biener <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.10.0

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Likely

+2014-06-28  Jan Hubicka  <hubicka@ucw.cz>
+
+       * tree-inline.c (remap_type_1): Do not duplicate fields
+       that are shared in between type and its main variant.

maybe you can verify that reverting fixes the issue?

[Bug regression/62102] [5 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

Jan Hubicka <hubicka at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2014-08-13
                 CC|                            |hubicka at gcc dot gnu.org
           Assignee|unassigned at gcc dot gnu.org      |hubicka at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #2 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
looking into it...

[Bug regression/62102] [5 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

--- Comment #3 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
-fno-early-inlining triggers segfaults on non-cris targets.

[Bug regression/62102] [5 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[hp at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

--- Comment #4 from Hans-Peter Nilsson <hp at gcc dot gnu.org> ---
(In reply to Richard Biener from comment #1)
> +2014-06-28  Jan Hubicka  <hubicka@ucw.cz>
> +
> +       * tree-inline.c (remap_type_1): Do not duplicate fields
> +       that are shared in between type and its main variant.
> 
> maybe you can verify that reverting fixes the issue?

That is r212111; I confirmed that reverting that indeed fixes the issue.

[Bug regression/62102] [5 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

--- Comment #5 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
I think that would be rather symptomatic fix and I plan to push out the type
verifier soon that will ICE without the patch on all targets. I failed to
reproduce this on x86 (I thought originally I have reproducer but it was
unrelated problem).

I will try to find time to build binutils so I can reproduce it for
x86_64-unknown-linux-gnu next week. Sorry for the delays - the reproducibility
of LTO bugs is somewhat annoying.

[Bug regression/62102] [5 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[hubicka at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

--- Comment #6 from Jan Hubicka <hubicka at gcc dot gnu.org> ---
Since this testcase also involves VLA, can you, please, test if the patch for
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62127
(now in mainline) fixes the problem?

[Bug regression/62102] [5 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[hp at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

Hans-Peter Nilsson <hp at gcc dot gnu.org> changed:

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

--- Comment #7 from Hans-Peter Nilsson <hp at gcc dot gnu.org> ---
(In reply to Jan Hubicka from comment #6)
> Since this testcase also involves VLA, can you, please, test if the patch
> for
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62127
(now in mainline)
> fixes the problem?

My autotester picked up that commit, and this regression is gone, thanks!
I'm closing this PR.

Specifically, this regression is gone with a commit in the range
(216146:216158], matching the commit in question.  Incidentally, at r216158
there's only gcc.dg/tls/alias-1.c i.e. PR61548 (since I started tracking
regressions for cris-elf in 2007).

[Bug regression/62102] [5 Regression]: gcc.dg/torture/pr48953.c -O2 -flto

[hubicka at ucw dot cz]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62102

--- Comment #8 from Jan Hubicka <hubicka at ucw dot cz> ---
> 
> My autotester picked up that commit, and this regression is gone, thanks!
> I'm closing this PR.

Great!  I was starring in that patch and did not notice this quite obvious
omision
for at least 20 times :(  Good it is gone.
The consequences of not remapping the type were particularly confusing to deal
with...

Honza
> 
> Specifically, this regression is gone with a commit in the range
> (216146:216158], matching the commit in question.  Incidentally, at r216158
> there's only gcc.dg/tls/alias-1.c i.e. PR61548 (since I started tracking
> regressions for cris-elf in 2007).
> 
> -- 
> You are receiving this mail because:
> You are on the CC list for the bug.