[Bug middle-end/94895] New: ICE in expand_block_tm, at trans-mem.c:2643

[Bug middle-end/94895] New: ICE in expand_block_tm, at trans-mem.c:2643

[asolokha at gmx dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94895

            Bug ID: 94895
           Summary: ICE in expand_block_tm, at trans-mem.c:2643
           Product: gcc
           Version: 10.0
            Status: UNCONFIRMED
          Keywords: ice-on-valid-code, trans-mem
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: asolokha at gmx dot com
  Target Milestone: ---

All g++ releases from g++-10.0.1-alpha20200426 snapshot
(g:29f55115583a0dab6cbac749c4f0804fd88e9536) down to 4.7 ICE when compiling
gcc/testsuite/g++.dg/tm/pr46269.C w/ -O1 -fgnu-tm --param
max-early-inliner-iterations=0:

% g++-10.0.1 -O1 -fgnu-tm --param max-early-inliner-iterations=0 -c
gcc/testsuite/g++.dg/tm/pr46269.C
during GIMPLE pass: tmmark                                                      
gcc/testsuite/g++.dg/tm/pr46269.C: In member function 'void
BuildingCompletedEvent::updateBuildingSite()':
gcc/testsuite/g++.dg/tm/pr46269.C:25:6: internal compiler error: in
expand_block_tm, at trans-mem.c:2643
   25 | void BuildingCompletedEvent::updateBuildingSite(void)
      |      ^~~~~~~~~~~~~~~~~~~~~~
0x754004 expand_block_tm
       
/var/tmp/portage/sys-devel/gcc-10.0.1_alpha20200426/work/gcc-10-20200426/gcc/trans-mem.c:2643
0x754004 execute_tm_mark
       
/var/tmp/portage/sys-devel/gcc-10.0.1_alpha20200426/work/gcc-10-20200426/gcc/trans-mem.c:3116
0x754004 execute
       
/var/tmp/portage/sys-devel/gcc-10.0.1_alpha20200426/work/gcc-10-20200426/gcc/trans-mem.c:3161

[Bug middle-end/94895] ICE in expand_block_tm, at trans-mem.c:2643

[marxin at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94895

Martin Liška <marxin at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2020-05-04
             Status|UNCONFIRMED                 |NEW
     Ever confirmed|0                           |1
                 CC|                            |marxin at gcc dot gnu.org

[Bug middle-end/94895] ICE in expand_block_tm, at trans-mem.c:2643

[xerofoify at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94895

Nicholas Krause <xerofoify at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |xerofoify at gmail dot com

--- Comment #1 from Nicholas Krause <xerofoify at gmail dot com> ---
Confirmed however seems to be latent as it seems to be hitting:
case GIMPLE_ASM:
          gcc_unreachable ();
in expand_block_tm and the commit mentioned is touching coroutines.

[Bug middle-end/94895] ICE in expand_block_tm, at trans-mem.c:2643

[asolokha at gmx dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94895

--- Comment #2 from Arseny Solokha <asolokha at gmx dot com> ---
(In reply to Nicholas Krause from comment #1)
> Confirmed however seems to be latent as it seems to be hitting:
> case GIMPLE_ASM:
>           gcc_unreachable ();

No, it's not latent. The code you quoted is exactly the immediate cause of this
ICE. There's an explicitly asserted assumption in rarely used, virtually
unmaintained, nine years old TM code that fails under some exotic artificial
circumstances no one cares about. Also, to reason about whether the issue was
latent or not you really had to bisect it, though this case is trivial and
doesn't even need bisection as the issue was in the file from its inception.

There's little value in simply pointing out that gcc_unreachable () w/o
understanding what exactly went wrong there. Anyone can do that. And anyone can
also run git blame on the file in question or click 'blame' in gitweb. If you
really want to help out, please come up w/ correct analysis which would result
in a good patch fixing the issue. Thanks.


> in expand_block_tm and the commit mentioned is touching coroutines.

How does the mentioned commit have to do w/ the issue?!