[Bug bootstrap/97124] New: ICE when bootstrapping GCC on x86_64-w64-mingw32

[Bug bootstrap/97124] New: ICE when bootstrapping GCC on x86_64-w64-mingw32

[lh_mouse at 126 dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97124

            Bug ID: 97124
           Summary: ICE when bootstrapping GCC on x86_64-w64-mingw32
           Product: gcc
           Version: 10.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
          Assignee: unassigned at gcc dot gnu.org
          Reporter: lh_mouse at 126 dot com
  Target Milestone: ---

Created attachment 49242
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=49242&action=edit
build log

This happens inside MSYS2 with `makepkg-mingw` [1].
Last known working revision should be around 2020-08-26.

Reverting these commits locally did NOT help:

    7ad48f0a6b498e4fb89d0ab1208d8e74fe747101
    2cbd82044b21679fa533445f373374576f638da1
    436ee9fec2751826bc48069d4dd320207fd9dfa4
    a41439a06901dfff7c01d2455cb881ff939b57b5



[1]
https://github.com/lhmouse/MINGW-packages/blob/master/mingw-w64-gcc-git/PKGBUILD


```
In file included from
E:/GitHub/MINGW-packages/mingw-w64-gcc-git/src/build-x86_64-w64-mingw32/gcc/include/immintrin.h:55,
                 from
E:/GitHub/MINGW-packages/mingw-w64-gcc-git/src/build-x86_64-w64-mingw32/gcc/include/x86intrin.h:32,
                 from
C:/MinGW/MSYS2/mingw64/x86_64-w64-mingw32/include/winnt.h:1555,
                 from
C:/MinGW/MSYS2/mingw64/x86_64-w64-mingw32/include/minwindef.h:163,
                 from
C:/MinGW/MSYS2/mingw64/x86_64-w64-mingw32/include/windef.h:9,
                 from
C:/MinGW/MSYS2/mingw64/x86_64-w64-mingw32/include/windows.h:69,
                 from ../../../gcc-git/libgcc/libgcc2.c:2183:
E:/GitHub/MINGW-packages/mingw-w64-gcc-git/src/build-x86_64-w64-mingw32/gcc/include/avx512fintrin.h:
In function
'_mm512_mask_reduce_mul_ps':
E:/GitHub/MINGW-packages/mingw-w64-gcc-git/src/build-x86_64-w64-mingw32/gcc/include/avx512fintrin.h:15958:3:
internal
compiler error: Segmentation fault
15958 |   __A = _mm512_mask_mov_ps (_mm512_set1_ps (1.0f), __U, __A);
      |   ^~~
libbacktrace could not find executable to open
Please submit a full bug report,
with preprocessed source if appropriate.
See <https://gcc-mcf.lhmouse.com/> for instructions.
make[3]: *** [Makefile:501: _trampoline.o] Error 1
make[3]: *** Waiting for unfinished jobs....
make[3]: Leaving directory
'/e/GitHub/MINGW-packages/mingw-w64-gcc-git/src/build-x86_64-w64-mingw32/x86_64-w64-mingw32/libgcc'
make[2]: *** [Makefile:17959: all-stage1-target-libgcc] Error 2
make[2]: Leaving directory
'/e/GitHub/MINGW-packages/mingw-w64-gcc-git/src/build-x86_64-w64-mingw32'
make[1]: *** [Makefile:21867: stage1-bubble] Error 2
make[1]: Leaving directory
'/e/GitHub/MINGW-packages/mingw-w64-gcc-git/src/build-x86_64-w64-mingw32'
make: *** [Makefile:1009: all] Error 2
==> ERROR: A failure occurred in build().
```

[Bug bootstrap/97124] ICE when bootstrapping GCC on x86_64-w64-mingw32

[lh_mouse at 126 dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97124

--- Comment #1 from Liu Hao <lh_mouse at 126 dot com> ---
MSYS2 has this patch [1]. Should it be incorporated by GCC somehow?


[1]
https://github.com/msys2/MINGW-packages/commit/0b2f1460ea3b72aa6cf4f263ace52f61a3470afe

[Bug bootstrap/97124] ICE when bootstrapping GCC on x86_64-w64-mingw32

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97124

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

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

--- Comment #2 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I think more interesting than the build.log would be preprocessed source it
ICEs on, so that it can be (attempted to) reproduced with cross-compilers too.

[Bug bootstrap/97124] ICE when bootstrapping GCC on x86_64-w64-mingw32

[lh_mouse at 126 dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97124

--- Comment #3 from Liu Hao <lh_mouse at 126 dot com> ---
Did you reproduce the ICE when cross-compiling?

I suspect no, as the MSYS2 patch contains some twiddling about stack limits and
forces static linking of libgcc during bootstrapping, which should not affect
cross-compiling.

[Bug bootstrap/97124] ICE when bootstrapping GCC on x86_64-w64-mingw32

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97124

--- Comment #4 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
Of course not, I don't have any libraries/headers for that.
What I meant is if you upload preprocessed source, one might try to reproduce
it with a cross that doesn't have anything (binutils, libraries, headers), just
by trying ./cc1 ...

[Bug bootstrap/97124] ICE when bootstrapping GCC on x86_64-w64-mingw32

[lh_mouse at 126 dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97124

--- Comment #5 from Liu Hao <lh_mouse at 126 dot com> ---
I didn't keep a copy of the file. It was stage1 xgcc that ICEs, not the host
gcc.

[Bug bootstrap/97124] ICE when bootstrapping GCC on x86_64-w64-mingw32

[lh_mouse at 126 dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97124

--- Comment #6 from Liu Hao <lh_mouse at 126 dot com> ---
Today I got the same ICE when building mingw-w64. I am not clear why this error
only happened with GCC previously.

[Bug bootstrap/97124] ICE when bootstrapping GCC on x86_64-w64-mingw32

[lh_mouse at 126 dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97124

Liu Hao <lh_mouse at 126 dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |INVALID

--- Comment #7 from Liu Hao <lh_mouse at 126 dot com> ---
I received this response from MSYS2 later:

```
jeremyd2019 @jeremyd2019 Oct 26 10:52
how is that back?!?
that is caused by a messed up libgmp, and gcc segfaults when it tries to parse
a floating point constant
(specifically, a messed up libgmp import library, and used via mpfr)
gmp patch I sent upstream:
https://gmplib.org/list-archives/gmp-bugs/2020-October/004890.html
```

There's a detailed explanation about the cause in the mail linked above, which
also answers why this ICE does not always happen.

AFAICT this can be closed.