[Bug libgcc/113700] New: libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[Bug libgcc/113700] New: libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[youremailsarecrap at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

            Bug ID: 113700
           Summary: libgcc_s does not include symbols for _Float16 and
                    __bf16 on Solaris/Illumos even though gcc generates
                    code for _Float16 and __bf16
           Product: gcc
           Version: 14.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: libgcc
          Assignee: unassigned at gcc dot gnu.org
          Reporter: youremailsarecrap at gmail dot com
  Target Milestone: ---

Created attachment 57277
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=57277&action=edit
_Float16 and __bf16 that breaks on Illumos

When compiling the code included in the .ii
files on illumos the output claims that it
can't find the symbols __extendhfdf2,
__extendbfsf2 and __truncsfbf2 .

Command line and output:

g++ -march=native -shared -shared-libgcc -fPIC -Wl,--no-undefined -o
f16-bf16.so f16-bf16.cc
Undefined                       first referenced
 symbol                             in file
__extendhfdf2                       f16-bf16.so-f16-bf16.o
__extendbfsf2                       f16-bf16.so-f16-bf16.o
__truncsfbf2                        f16-bf16.so-f16-bf16.o
ld: fatal: symbol referencing errors. No output written to f16-bf16.so
collect2: error: ld returned 1 exit status

The -Wl,--no-undefined is used here to not have create a executable that links
to the so files
to cause the error.

-march=native is used since sse2 seems to be needed for f16 and bf16 on x86
platforms if I
understand the half-precision doc in gcc correctly, I used a system with sse2
when compiling this.

My own fix, that may not be correct:
I tried to add the gcc12 and up parts of
libgcc/config/i386/libgcc-glibc.ver to
libgcc/config/i386/libgcc-sol2.ver rebuilt
gcc and then it linked correctly.
Some maintainer could take a look at that.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
>I tried to add the gcc12 and up parts of

It is correct except it should just use GCC 14 I think.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[youremailsarecrap at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #2 from Niclas Rosenvik <youremailsarecrap at gmail dot com> ---
(In reply to Andrew Pinski from comment #1)
> >I tried to add the gcc12 and up parts of
> 
> It is correct except it should just use GCC 14 I think.

I forgot to mention that the problem with _Float16 aka
__extendhfdf2 has happened on gcc12 as well as gcc14.
If it was possible to not just choose one version of
gcc then I would have marked gcc12 to 14.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2024-02-01
     Ever confirmed|0                           |1

--- Comment #3 from Rainer Orth <ro at gcc dot gnu.org> ---
(In reply to Niclas Rosenvik from comment #2)
> (In reply to Andrew Pinski from comment #1)
> > >I tried to add the gcc12 and up parts of
> > 
> > It is correct except it should just use GCC 14 I think.
> 
> I forgot to mention that the problem with _Float16 aka
> __extendhfdf2 has happened on gcc12 as well as gcc14.
> If it was possible to not just choose one version of
> gcc then I would have marked gcc12 to 14.

You cannot add symbols to symbol versions of older releases retroactively.
Since those versions (GCC_12.0.0, GCC_13.0.0, GCC_14.0.0) were overlooked
before the respective releases, all symbols that are listed in
i386/libgcc-glibc.ver in those versions need to be added to
i386/libgcc-sol2.ver in version GCC_14.0.0 (after verifying that they are
actually defined on Solaris/x86).

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[youremailsarecrap at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #4 from Niclas Rosenvik <youremailsarecrap at gmail dot com> ---
(In reply to Rainer Orth from comment #3)
> (In reply to Niclas Rosenvik from comment #2)
> > (In reply to Andrew Pinski from comment #1)
> > > >I tried to add the gcc12 and up parts of
> > > 
> > > It is correct except it should just use GCC 14 I think.
> > 
> > I forgot to mention that the problem with _Float16 aka
> > __extendhfdf2 has happened on gcc12 as well as gcc14.
> > If it was possible to not just choose one version of
> > gcc then I would have marked gcc12 to 14.
> 
> You cannot add symbols to symbol versions of older releases retroactively.
> Since those versions (GCC_12.0.0, GCC_13.0.0, GCC_14.0.0) were overlooked
> before the respective releases, all symbols that are listed in
> i386/libgcc-glibc.ver in those versions need to be added to
> i386/libgcc-sol2.ver in version GCC_14.0.0 (after verifying that they are
> actually defined on Solaris/x86).

OK, thanks for explaining.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |andreast at gcc dot gnu.org,
                   |                            |iains at gcc dot gnu.org,
                   |                            |jakub at gcc dot gnu.org
             Target|x86_64-*-solaris*           |x86_64-*-solaris*,
                   |                            |i?86-*-solaris*
   Target Milestone|---                         |14.0

--- Comment #5 from Rainer Orth <ro at gcc dot gnu.org> ---
The problem is even more widespread than Solaris/x86: beside
i386/libgcc-sol2.ver,
there are i386/libgcc-darwin.ver (where the versions end at GCC_12.0.0) and
i386/libgcc-bsdver (stops at GCC_7.0.0, used by both t-freebsd and t-dragonfly;
the latter has no listed maintainer).

I wonder if we should add a prominent note to the end of i386/libgcc-glibc.ver
to notify other affected maintainers about additions.  Those are way too easy
to overlook right now.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[ro at CeBiTec dot Uni-Bielefeld.DE]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #6 from ro at CeBiTec dot Uni-Bielefeld.DE <ro at CeBiTec dot Uni-Bielefeld.DE> ---
When looking at the 64-bit libgcc_s.so.1 on both Solaris/x86 and
Linux/i686, I noticed that all the new GCC_14.0.0 symbols from
libgcc-glibc.ver (and now libgcc-sol2.ver) have been demoted to local.

IIUC, this happens because the __PFX__ handling (substitution by
$(LIBGCC_VER_GNU_PREFIX) as needed) is only applied to libgcc-std.ver.in
by Makefile.in.  In the i386/libgcc-*.ver files, this substitution
doesn't happen, the literal "__PFX__fixxfbitint" etc. symbols are not
found in any object, so the unprefixed ones are turned local.

From what I could see in config.host, LIBGCC_VER_GNU_PREFIX only applies
to non-x86 targets.  Maybe we can just remove __PFX__ from
i386/libgcc-*.ver?  Jakub?

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #7 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
I didn't know what it is.
But seems you're right, on x86_64-linux
   252: 000000000001c1c0  1974 FUNC    LOCAL  DEFAULT   13 __floatbitintbf
   253: 000000000001b200  1972 FUNC    LOCAL  DEFAULT   13 __fixxfbitint
   262: 000000000001b9c0  2046 FUNC    LOCAL  DEFAULT   13 __floatbitinthf
   266: 000000000001c980  1934 FUNC    LOCAL  DEFAULT   13 __floatbitintxf
   271: 0000000000019220  2271 FUNC    LOCAL  DEFAULT   13 __fixtfbitint
   276: 0000000000019b00  2166 FUNC    LOCAL  DEFAULT   13 __floatbitinttf
aren't exported even when they should be.  Let me change it.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[jakub at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The reason tests work is that we default to -static-libgcc in C, which contains
the symbols.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #9 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Jakub Jelinek <jakub@gcc.gnu.org>:

https://gcc.gnu.org/g:9f5caef53e7808fef21111baf8e6ffac230b9863

commit r14-8750-g9f5caef53e7808fef21111baf8e6ffac230b9863
Author: Jakub Jelinek <jakub@redhat.com>
Date:   Fri Feb 2 11:46:34 2024 +0100

    libgcc: Export XF, TF, HF and BFmode specific _BitInt symbols from
libgcc_s.so.1 [PR113700]

    Rainer pointed out that __PFX__ and __FIXPTPFX__ prefix replacement is done
    solely for libgcc-std.ver.in and not for the *.ver files in config.
    I've used the __PFX__ prefix even in config/i386/libgcc-glibc.ver because
it
    was used for similar symbols in libgcc-std.ver.in, and that results in
those
    symbols being STB_LOCAL in libgcc_s.so.1.  Tests still work because gcc by
    default uses -static-libgcc when linking (unlike g++ etc.), but would
    have failed when using -shared-libgcc (but I see nothing in the testsuite
    actually testing with -shared-libgcc, so am not adding tests).

    With the patch, libgcc_s.so.1 now exports
    __fixtfbitint@@GCC_14.0.0 FUNC GLOBAL DEFAULT
    __fixxfbitint@@GCC_14.0.0 FUNC GLOBAL DEFAULT
    __floatbitintbf@@GCC_14.0.0 FUNC GLOBAL DEFAULT
    __floatbitinthf@@GCC_14.0.0 FUNC GLOBAL DEFAULT
    __floatbitinttf@@GCC_14.0.0 FUNC GLOBAL DEFAULT
    __floatbitintxf@@GCC_14.0.0 FUNC GLOBAL DEFAULT
    on x86_64-linux which it wasn't before.

    2024-02-02  Jakub Jelinek  <jakub@redhat.com>

            PR target/113700
            * config/i386/libgcc-glibc.ver (GCC_14.0.0): Remove __PFX prefixes
            from symbol names.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
                URL|                            |https://gcc.gnu.org/piperma
                   |                            |il/gcc-patches/2024-Februar
                   |                            |y/645035.html

--- Comment #10 from Rainer Orth <ro at gcc dot gnu.org> ---
Patch posted.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[iains at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #11 from Iain Sandoe <iains at gcc dot gnu.org> ---
(In reply to Rainer Orth from comment #10)
> Patch posted.

FWIW Darwin is, indeed, also affected and I have patches in progress to resolve
it.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

--- Comment #12 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Rainer Orth <ro@gcc.gnu.org>:

https://gcc.gnu.org/g:c5f48b5fdde849759d0e3b4effd9352a2399d6f9

commit r14-8820-gc5f48b5fdde849759d0e3b4effd9352a2399d6f9
Author: Rainer Orth <ro@CeBiTec.Uni-Bielefeld.DE>
Date:   Tue Feb 6 10:20:30 2024 +0100

    libgcc: Export i386 symbols added after GCC_7.0.0 on Solaris [PR113700]

    As reported in the PR, all libgcc x86 symbol versions added after
    GCC_7.0.0 were only added to i386/libgcc-glibc.ver, missing all of
    libgcc-sol2.ver, libgcc-bsd.ver, and libgcc-darwin.ver.

    This patch fixes this for Solaris/x86, adding all of them
    (GCC_1[234].0.0) as GCC_14.0.0 to not retroactively change history.

    Since this isn't the first time this happens, I've added a note to the
    end of libgcc-glibc.ver to request notifying other maintainers in case
    of additions.

    Tested on i386-pc-solaris2.11.

    2024-02-01  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>

            libgcc:
            PR target/113700
            * config/i386/libgcc-sol2.ver (GCC_14.0.0): Added all symbols from
            i386/libgcc-glibc.ver (GCC_12.0.0, GCC_13.0.0, GCC_14.0.0).
            * config/i386/libgcc-glibc.ver: Request notifications on updates.

[Bug target/113700] libgcc_s does not include symbols for _Float16 and __bf16 on Solaris/Illumos even though gcc generates code for _Float16 and __bf16

[ro at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113700

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

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
             Status|NEW                         |RESOLVED
           Assignee|unassigned at gcc dot gnu.org      |ro at gcc dot gnu.org

--- Comment #13 from Rainer Orth <ro at gcc dot gnu.org> ---
Mine.  Fixed for GCC 14.0.0.  Thanks for the report.