[Bug c++/113347] New: ICE during gimplification building TVM

[Bug c++/113347] New: ICE during gimplification building TVM

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

            Bug ID: 113347
           Summary: ICE during gimplification building TVM
           Product: gcc
           Version: 13.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c++
          Assignee: unassigned at gcc dot gnu.org
          Reporter: rguenth at gcc dot gnu.org
  Target Milestone: ---

I see

> ./cc1plus -fpreprocessed cross_thread_reduction.cc.ii -quiet -dumpbase-ext .cc -mtune=generic -march=x86-64 -g -g -O2 -O2 -O2 -Wall -Werror=return-type -std=c++17 -version -faligned-new=1 -fPIC -fstack-protector-strong -funwind-tables -fasynchronous-unwind-tables -fstack-clash-protection -o cross_thread_reduction.cc.s
GNU C++17 (GCC) version 13.2.1 20240112 (x86_64-pc-linux-gnu)
        compiled by GNU C version 7.5.0, GMP version 6.1.2, MPFR version
4.0.2-p6, MPC version 1.1.0, isl version isl-0.18-GMP

GGC heuristics: --param ggc-min-expand=30 --param ggc-min-heapsize=4096
Compiler executable checksum: b21c83316346ea37120ce6d56dd5bdc5
In member function 'tvm::tir::ExprRV
tvm::meta_schedule::CrossThreadReductionNode::GetThreadIdxExtentFromTrace(const
tvm::tir::Trace&)':
cc1plus: internal compiler error: Segmentation fault
0x19cbf8f crash_signal
        /home/rguenther/src/gcc-13-branch/gcc/toplev.cc:314
0xcc7260 contains_struct_check(tree_node*, tree_node_structure_enum, char
const*, int, char const*)
        /home/rguenther/src/gcc-13-branch/gcc/tree.h:3653
0xd9e44f cp_gimplify_expr(tree_node**, gimple**, gimple**)
        /home/rguenther/src/gcc-13-branch/gcc/cp/cp-gimplify.cc:552
0x15b96fd gimplify_expr(tree_node**, gimple**, gimple**, bool (*)(tree_node*),
int)
        /home/rguenther/src/gcc-13-branch/gcc/gimplify.cc:16291
0x158d62a gimplify_stmt(tree_node**, gimple**)
        /home/rguenther/src/gcc-13-branch/gcc/gimplify.cc:7226
0x158a696 gimplify_compound_expr
        /home/rguenther/src/gcc-13-branch/gcc/gimplify.cc:6412
...

on the GCC 13 branch head.  The ICE doesn't occur on trunk.

[Bug c++/113347] ICE during gimplification building TVM

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Created attachment 57047
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57047&action=edit
preprocessed source

[Bug c++/113347] ICE during gimplification building TVM

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |ice-on-valid-code,
                   |                            |needs-bisection

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
Reducing it, bisection to the duplicate bug (if any) would be nice.

[Bug c++/113347] [13 Regression] ICE during gimplification building TVM

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|ICE during gimplification   |[13 Regression] ICE during
                   |building TVM                |gimplification building TVM
   Target Milestone|---                         |13.3
      Known to work|                            |13.2.0
           Priority|P3                          |P1

--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
Hmm, and 13.2.0 worked for me.

[Bug c++/113347] [13 Regression] ICE during gimplification building TVM

[fabian@ritter-vogt.de]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

Fabian Vogt <fabian@ritter-vogt.de> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |fabian@ritter-vogt.de

--- Comment #4 from Fabian Vogt <fabian@ritter-vogt.de> ---
Created attachment 57048
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57048&action=edit
reduced source

I ran cvise twice, it spat out this minimal reproducer.

[Bug c++/113347] [13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
   Last reconfirmed|                            |2024-01-12
             Status|UNCONFIRMED                 |NEW
            Summary|[13 Regression] ICE during  |[13 Regression] ICE during
                   |gimplification building TVM |gimplification building TVM
                   |                            |since
                   |                            |r13-8079-gd237e7b291ff52
                 CC|                            |jason at gcc dot gnu.org

--- Comment #5 from Richard Biener <rguenth at gcc dot gnu.org> ---
(gdb) p debug_tree (*expr_p)
 <modify_expr 0x7ffff6bd35c8
    type <boolean_type 0x7ffff6a2eb28 bool public unsigned type_6 QI
        size <integer_cst 0x7ffff6a300c0 constant 8>
        unit-size <integer_cst 0x7ffff6a300d8 constant 1>
        align:8 warn_if_not_align:0 symtab:0 alias-set -1 canonical-type
0x7ffff6a2eb28 precision:1 min <integer_cst 0x7ffff6a30318 0> max <integer_cst
0x7ffff6a30348 1>>
    side-effects readonly
    arg:1 <integer_cst 0x7ffff6a30318 type <boolean_type 0x7ffff6a2eb28 bool>
constant 0>>

so the LHS of the MODIFY_EXPR is NULL.  This is built by finalize_nrv_r:

      /* If a cleanup might throw, we need to clear current_retval_sentinel on
         the exception path so an outer cleanup added by
         maybe_splice_retval_cleanup doesn't run.  */
      if (cp_function_chain->throwing_cleanup) 
        {
          tree clear = build2 (MODIFY_EXPR, boolean_type_node,
                               current_retval_sentinel,
                               boolean_false_node);

          /* We're already only on the EH path, just prepend it.  */
          tree &exp = CLEANUP_EXPR (*tp);
          exp = build2 (COMPOUND_EXPR, void_type_node, clear, exp);

and current_retval_sentinel is NULL here.

Caused by the backport of PR112301, r13-8079-gd237e7b291ff52.

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |12.3.1, 13.2.1
      Known to work|                            |12.3.0
            Summary|[13 Regression] ICE during  |[12/13 Regression] ICE
                   |gimplification building TVM |during gimplification
                   |since                       |building TVM since
                   |r13-8079-gd237e7b291ff52    |r13-8079-gd237e7b291ff52
   Target Milestone|13.3                        |12.4

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |needs-bisection

--- Comment #6 from Richard Biener <rguenth at gcc dot gnu.org> ---
I'd like to know what fixed it on trunk still so we can decide whether
reverting or backporting sth additional.

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |mpolacek at gcc dot gnu.org

--- Comment #7 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
I cannot figure out what fixed it on trunk because r14-5086 doesn't ICE for me.

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[mpolacek at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

Marek Polacek <mpolacek at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |boris at kolpackov dot net

--- Comment #8 from Marek Polacek <mpolacek at gcc dot gnu.org> ---
*** Bug 113400 has been marked as a duplicate of this bug. ***

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |csfore at posteo dot net

--- Comment #9 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
*** Bug 113547 has been marked as a duplicate of this bug. ***

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

Jason Merrill <jason at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |jason at gcc dot gnu.org

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

--- Comment #10 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-13 branch has been updated by Jason Merrill
<jason@gcc.gnu.org>:

https://gcc.gnu.org/g:713fbaff8b17a01d3b72110f89112851ed43a90a

commit r13-8249-g713fbaff8b17a01d3b72110f89112851ed43a90a
Author: Jason Merrill <jason@redhat.com>
Date:   Tue Jan 23 15:41:09 2024 -0500

    c++: throwing cleanup after return [PR113347]

    Here we were assuming that current_retval_sentinel would be set if we have
    seen a throwing cleanup, but that's not the case if the cleanup is after
all
    returns.

    This change isn't needed on trunk, where current_retval_sentinel is set for
    all NRV functions.

            PR c++/113347

    gcc/cp/ChangeLog:

            * semantics.cc (finalize_nrv_r): Handle null
            current_retval_sentinel.

    gcc/testsuite/ChangeLog:

            * g++.dg/eh/return3.C: New test.

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

--- Comment #11 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The releases/gcc-12 branch has been updated by Jason Merrill
<jason@gcc.gnu.org>:

https://gcc.gnu.org/g:050fc885a6e324a03a6397d86134ab4b68aac944

commit r12-10112-g050fc885a6e324a03a6397d86134ab4b68aac944
Author: Jason Merrill <jason@redhat.com>
Date:   Tue Jan 23 15:41:09 2024 -0500

    c++: throwing cleanup after return [PR113347]

    Here we were assuming that current_retval_sentinel would be set if we have
    seen a throwing cleanup, but that's not the case if the cleanup is after
all
    returns.

    This change isn't needed on trunk, where current_retval_sentinel is set for
    all NRV functions.

            PR c++/113347

    gcc/cp/ChangeLog:

            * semantics.cc (finalize_nrv_r): Handle null
            current_retval_sentinel.

    gcc/testsuite/ChangeLog:

            * g++.dg/eh/return3.C: New test.

    (cherry picked from commit 713fbaff8b17a01d3b72110f89112851ed43a90a)

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

--- Comment #12 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The trunk branch has been updated by Jason Merrill <jason@gcc.gnu.org>:

https://gcc.gnu.org/g:aedfdd340cbb4a97a8f6d432720a4df374c1e467

commit r14-8401-gaedfdd340cbb4a97a8f6d432720a4df374c1e467
Author: Jason Merrill <jason@redhat.com>
Date:   Tue Jan 23 16:09:15 2024 -0500

    c++: add test [PR113347]

    The patch for this PR is unneeded on trunk, but let's add the test.

            PR c++/113347

    gcc/testsuite/ChangeLog:

            * g++.dg/eh/return3.C: New test.

[Bug c++/113347] [12/13 Regression] ICE during gimplification building TVM since r13-8079-gd237e7b291ff52

[jason at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113347

Jason Merrill <jason at gcc dot gnu.org> changed:

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

--- Comment #13 from Jason Merrill <jason at gcc dot gnu.org> ---
Fixed for 12.4/13.3.