[Bug middle-end/114912] New: [15 regression] SIGBUS in wi::copy<> on SPARC

[Bug middle-end/114912] New: [15 regression] SIGBUS in wi::copy<> on SPARC

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

            Bug ID: 114912
           Summary: [15 regression] SIGBUS in wi::copy<> on SPARC
           Product: gcc
           Version: 15.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: ro at gcc dot gnu.org
                CC: aldyh at gcc dot gnu.org
  Target Milestone: ---
            Target: sparc-sun-solaris2.11

Between 20240430 (0b2735e0797fee9b4ec5cd74f22afe0483f888dd) and 20240501
(c3bc2787b8beb7aae67fdf2a7f7271a9a4edca7c),
Solaris/SPARC bootstrap began to fail with a SIGBUS in cc1 compiling stage 1
libgcc.

E.g.

/var/gcc/reghunt/sigbus-range/288807/./gcc/xgcc
-B/var/gcc/reghunt/sigbus-range/288807/./gcc/
-B/usr/local/sparc-sun-solaris2.11/bin/ -B/usr/local/sparc-sun-solaris2.11/lib/
-isystem /usr/local/sparc-sun-solaris2.11/include -isystem
/usr/local/sparc-sun-solaris2.11/sys-include    -g -O2 -O2  -g -O2 -DIN_GCC  
-W -Wall -Wno-narrowing -Wwrite-strings -Wcast-qual -Wstrict-prototypes
-Wmissing-prototypes -Wold-style-definition  -isystem ./include  -fPIC -g
-DIN_LIBGCC2 -fbuilding-libgcc -fno-stack-protector  -fPIC -I. -I.
-I../.././gcc -I/var/gcc/reghunt/master/libgcc
-I/var/gcc/reghunt/master/libgcc/. -I/var/gcc/reghunt/master/libgcc/../gcc
-I/var/gcc/reghunt/master/libgcc/../include     -o _muldi3.o -MT _muldi3.o -MD
-MP -MF _muldi3.dep -DL_muldi3 -c /var/gcc/reghunt/master/libgcc/libgcc2.c
-fvisibility=hidden -DHIDE_EXPORTS
during GIMPLE pass: evrp
/var/gcc/reghunt/master/libgcc/libgcc2.c: In function ‘__muldi3’:
/var/gcc/reghunt/master/libgcc/libgcc2.c:538:1: internal compiler error: Bus
Error
  538 | }
      | ^
0xf26783 crash_signal
        /var/gcc/reghunt/master/gcc/toplev.cc:319
0x723e58 void wi::copy<wide_int_storage,
generic_wide_int<wide_int_ref_storage<true, false> > >(wide_int_storage&,
generic_wide_int<wide_int_ref_storage<true, false> > const&)
        /var/gcc/reghunt/master/gcc/wide-int.h:2191
0x723e58 wide_int_storage&
wide_int_storage::operator=<wi::hwi_with_prec>(wi::hwi_with_prec const&)
        /var/gcc/reghunt/master/gcc/wide-int.h:1247
0x723e58 generic_wide_int<wide_int_storage>&
generic_wide_int<wide_int_storage>::operator=<wi::hwi_with_prec>(wi::hwi_with_prec
const&)
        /var/gcc/reghunt/master/gcc/wide-int.h:1002
0x723e58 irange_bitmask::set_unknown(unsigned int)
        /var/gcc/reghunt/master/gcc/value-range.h:163
0x723e58 irange::set_varying(tree_node*)
        /var/gcc/reghunt/master/gcc/value-range.h:1067
0x13680db gimple_range_global(vrange&, tree_node*, function*)
        /var/gcc/reghunt/master/gcc/value-query.cc:419
0x136954f global_range_query::range_of_expr(vrange&, tree_node*, gimple*)
        /var/gcc/reghunt/master/gcc/value-query.cc:436
0x1ad000b fur_stmt::get_operand(vrange&, tree_node*)
        /var/gcc/reghunt/master/gcc/gimple-range-fold.cc:162
0x1ad5edb fold_using_range::range_of_range_op(vrange&,
gimple_range_op_handler&, fur_source&)
        /var/gcc/reghunt/master/gcc/gimple-range-fold.cc:673
0x1ad831f fold_using_range::fold_stmt(vrange&, gimple*, fur_source&,
tree_node*)
        /var/gcc/reghunt/master/gcc/gimple-range-fold.cc:604
0x1ad8927 fold_range(vrange&, gimple*, range_query*)
        /var/gcc/reghunt/master/gcc/gimple-range-fold.cc:324
0x1ac8727 ranger_cache::get_global_range(vrange&, tree_node*, bool&)
        /var/gcc/reghunt/master/gcc/gimple-range-cache.cc:1054
0x1abefaf gimple_ranger::range_of_stmt(vrange&, gimple*, tree_node*)
        /var/gcc/reghunt/master/gcc/gimple-range.cc:323
0x1367d2f range_query::value_of_stmt(gimple*, tree_node*)
        /var/gcc/reghunt/master/gcc/value-query.cc:133
0x1315c3f rvrp_folder::value_of_stmt(gimple*, tree_node*)
        /var/gcc/reghunt/master/gcc/tree-vrp.cc:1001
0x11ab12f substitute_and_fold_dom_walker::before_dom_children(basic_block_def*)
        /var/gcc/reghunt/master/gcc/tree-ssa-propagate.cc:820
0x1a4df83 dom_walker::walk(basic_block_def*)
        /var/gcc/reghunt/master/gcc/domwalk.cc:311
0x11a9d07 substitute_and_fold_engine::substitute_and_fold(basic_block_def*)
        /var/gcc/reghunt/master/gcc/tree-ssa-propagate.cc:999
0x13132f7 execute_ranger_vrp(function*, bool, bool)
        /var/gcc/reghunt/master/gcc/tree-vrp.cc:1066

A reghunt identified

commit c60b3e211c555706cdc2dc8bfcdd540152cff350
Author: Aldy Hernandez <aldyh@redhat.com>
Date:   Tue Apr 30 19:39:00 2024 +0200

    Reduce startup costs for Value_Range.

as the culprit.

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

Rainer Orth <ro at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |15.0

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #1 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
Since this happens while building libgcc during stage1, perhaps this can be
reproduced with a cross?  Would it be possible to get the preprocessed file
that's failing?

You could try /var/gcc/reghunt/sigbus-range/288807/./gcc/xgcc -save-temps [blah
blah], and attach the libgcc2.i file that gets generated.

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
What compiler version are you starting with?
It could be that compiler is miscompiling stage 1 here; especially when it
comes to C++ usage is becoming more and more.

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #3 from Rainer Orth <ro at gcc dot gnu.org> ---
(In reply to Aldy Hernandez from comment #1)
> Since this happens while building libgcc during stage1, perhaps this can be
> reproduced with a cross?  Would it be possible to get the preprocessed file
> that's failing?

I doubt this can be seen in a cross: running the compilation under truss
reveals
the SIGBUS as

2834:       Incurred fault #5, FLTACCESS  %pc = 0x00723E58
2834:         siginfo: SIGBUS BUS_ADRALN addr=0xFFBFC954
2834:       Received signal #10, SIGBUS [caught]
2834:         siginfo: SIGBUS BUS_ADRALN addr=0xFFBFC954

i.e. an unaligned access.  While SPARC is a strict-alignment target, x86 cares
little if any about alignment at all.

> You could try /var/gcc/reghunt/sigbus-range/288807/./gcc/xgcc -save-temps
> [blah blah], and attach the libgcc2.i file that gets generated.

Sure, please find _muldi3.i attached.

cc1 invocation is

cc1 -fpreprocessed _muldi3.i -quiet -dumpbase _muldi3.c -dumpbase-ext .c
-mcpu=v9 -g -g -g -O2 -O2 -O2 -Wextra -Wall -Wno-narrowing -Wwrite-strings
-Wcast-qual -Wstrict-prototypes -Wmissing-prototypes -Wold-style-definition
-version -fbuilding-libgcc -fno-stack-protector -fPIC -fvisibility=hidden -o
_muldi3.s

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #4 from Rainer Orth <ro at gcc dot gnu.org> ---
Created attachment 58081
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58081&action=edit
preprocessed input

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #5 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I see the issue then.
char m_buffer[sizeof (int_range_max)];

Needs _Align to get the alignment correct.

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
I might see if I can figure out a patch for some to try later tonight.

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #7 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Created attachment 58084
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58084&action=edit
Something like this

This should cause the char array be on the correct alignment ...

[Bug middle-end/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Last reconfirmed|                            |2024-05-02
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW

--- Comment #8 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
The other way of fixing this is to use an union and I think since we are using
C++11, it might work correctly.

I do think we should prefer the union rather than having it as a char array too
...

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #9 from Richard Biener <rguenth at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #8)
> The other way of fixing this is to use an union and I think since we are
> using C++11, it might work correctly.
> 
> I do think we should prefer the union rather than having it as a char array
> too ...

If a union works then indeed that's what we should use.

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
If Aldy does not fix it by Saturday, I will give the union a try then. I will
also try out the solaris machine on the compile farm if possible.

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #11 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #10 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
> If Aldy does not fix it by Saturday, I will give the union a try then. I will

Great, thanks.  Your alignas patch also worked fine btw.

> also try out the solaris machine on the compile farm if possible.

Otherwise, you can just send it my way and I'll give it a whirl.

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #12 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #10)
> If Aldy does not fix it by Saturday, I will give the union a try then. I
> will also try out the solaris machine on the compile farm if possible.

Sorry, didn't mean for you to pick this up.  Thanks for the analysis.  I can
take it from here.

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #13 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
BTW, I'm waiting for a review, or at least a nod from a C++ savvy person here:

https://gcc.gnu.org/pipermail/gcc-patches/2024-May/650634.html

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #14 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #13 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
> BTW, I'm waiting for a review, or at least a nod from a C++ savvy person here:
>
> https://gcc.gnu.org/pipermail/gcc-patches/2024-May/650634.html

I can give the patch a whirl, thanks.

I had Andrew's patch in my tree to avoid the issue.  Unfortunately,
Solaris/SPARC bootstrap is broken again due to PR ipa/114985.

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #15 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
Created attachment 58136
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=58136&action=edit
proposed patch in testing

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #16 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
(In reply to ro@CeBiTec.Uni-Bielefeld.DE from comment #14)
> > --- Comment #13 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
> > BTW, I'm waiting for a review, or at least a nod from a C++ savvy person here:
> >
> > https://gcc.gnu.org/pipermail/gcc-patches/2024-May/650634.html
> 
> I can give the patch a whirl, thanks.

I've attached a rebased patch against current mainline.  Let me know if it
works on your end, and I'll commit.

> 
> I had Andrew's patch in my tree to avoid the issue.  Unfortunately,
> Solaris/SPARC bootstrap is broken again due to PR ipa/114985.

I have provided a patch for that PR as well, but the IPA folk need to say if
this is the correct approach.

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #17 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Aldy Hernandez <aldyh@gcc.gnu.org>:

https://gcc.gnu.org/g:d7ff8ae5313bea755f5960786b33a7b151e7b663

commit r15-336-gd7ff8ae5313bea755f5960786b33a7b151e7b663
Author: Aldy Hernandez <aldyh@redhat.com>
Date:   Fri May 3 11:17:32 2024 +0200

    [ranger] Force buffer alignment in Value_Range [PR114912]

    gcc/ChangeLog:

            PR tree-optimization/114912
            * value-range.h (class Value_Range): Use a union.

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

--- Comment #18 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
> --- Comment #16 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
> (In reply to ro@CeBiTec.Uni-Bielefeld.DE from comment #14)
>> > --- Comment #13 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
>> > BTW, I'm waiting for a review, or at least a nod from a C++ savvy
>> > person here:
>> >
>> > https://gcc.gnu.org/pipermail/gcc-patches/2024-May/650634.html
>> 
>> I can give the patch a whirl, thanks.
>
> I've attached a rebased patch against current mainline.  Let me know if it
> works on your end, and I'll commit.

I've included both this patch ...

>> I had Andrew's patch in my tree to avoid the issue.  Unfortunately,
>> Solaris/SPARC bootstrap is broken again due to PR ipa/114985.
>
> I have provided a patch for that PR as well, but the IPA folk need to say if
> this is the correct approach.

... and that one in last night's SPARC bootstraps, which worked just
fine again.  Thanks.

[Bug tree-optimization/114912] [15 regression] SIGBUS in wi::copy<> on SPARC since r15-88-gc60b3e211c5557 since char array is not aligned to what it needs to be

[aldyh at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=114912

Aldy Hernandez <aldyh at gcc dot gnu.org> changed:

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

--- Comment #19 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
(In reply to ro@CeBiTec.Uni-Bielefeld.DE from comment #18)
> > --- Comment #16 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
> > (In reply to ro@CeBiTec.Uni-Bielefeld.DE from comment #14)
> >> > --- Comment #13 from Aldy Hernandez <aldyh at gcc dot gnu.org> ---
> >> > BTW, I'm waiting for a review, or at least a nod from a C++ savvy
> >> > person here:
> >> >
> >> > https://gcc.gnu.org/pipermail/gcc-patches/2024-May/650634.html
> >> 
> >> I can give the patch a whirl, thanks.
> >
> > I've attached a rebased patch against current mainline.  Let me know if it
> > works on your end, and I'll commit.
> 
> I've included both this patch ...
> 
> >> I had Andrew's patch in my tree to avoid the issue.  Unfortunately,
> >> Solaris/SPARC bootstrap is broken again due to PR ipa/114985.
> >
> > I have provided a patch for that PR as well, but the IPA folk need to say if
> > this is the correct approach.
> 
> ... and that one in last night's SPARC bootstraps, which worked just
> fine again.  Thanks.

Thanks for testing.  I'll close this PR as fixed in mainline then.  FWIW, I
also retested on x86-64 Linux.