[Bug target/94282] New: [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[Bug target/94282] New: [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[tschwinge at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

            Bug ID: 94282
           Summary: [amdgcn] ld: error: undefined symbol:
                    __gxx_personality_v0
           Product: gcc
           Version: unknown
            Status: UNCONFIRMED
          Keywords: openacc, openmp
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: tschwinge at gcc dot gnu.org
                CC: ams at gcc dot gnu.org, jules at gcc dot gnu.org
  Target Milestone: ---
            Target: amdgcn

This is with GCC commit b73f69020f08208d2d969fcf8879bd294a6e3596 sources, with
commit b12fb08bd95d7dbb9fd6af120ee0b4c658a3deb1, and commit
65858e1869be4a54f8e0e2073a4aa99b15471273 cherry-picked, and PR94248 fixed with
Jakub's 'reg_overlap_mentioned_p' patch.

Relative to a GCC nvptx offloading configuration, when adding amdgcn
offloading, I see the following regression with already XFAILed
'libgomp.c++/../libgomp.c-c++-common/function-not-offloaded.c':

    ld: error: undefined symbol: foo()
    >>> referenced by /tmp/ccNzknBD.o:(main._omp_fn.0)
    >>> referenced by /tmp/ccNzknBD.o:(main._omp_fn.0)

    ld: error: undefined symbol: __gxx_personality_v0
    >>> referenced by /tmp/ccNzknBD.o:(.data+0x13)
    collect2: error: ld returned 1 exit status
    mkoffload: fatal error:
[...]/build-gcc/./gcc/x86_64-pc-linux-gnu-accel-amdgcn-amdhsa-gcc returned 1
exit status

The first error is expected (thus the XFAIL); the second error is not expected,
appears for the 'libgomp.c++' variant only.

This is somewhat related to PR92713.

Assuming that's not yet done, can something similar be done as done for nvptx,
where we force-disable all C++ exceptions code/tables generation, see PR86660
GCC trunk r263265, and then GCC trunk r263287?

Same for 'libgomp.oacc-*' (with all "execution test" etc. FAILs ignored as
"[n/a]"; "libgomp: device type gcn not supported" due to running on unsuitable
hardware):

    FAIL: libgomp.oacc-c++/../libgomp.oacc-c-c++-common/routine-1.c
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  (test
for excess errors)
    UNRESOLVED: libgomp.oacc-c++/../libgomp.oacc-c-c++-common/routine-1.c
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0 
compilation failed to produce executable
    PASS: libgomp.oacc-c++/../libgomp.oacc-c-c++-common/routine-1.c
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  (test
for excess errors)
    [n/a]: libgomp.oacc-c++/../libgomp.oacc-c-c++-common/routine-1.c
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2 
execution test

    FAIL: libgomp.oacc-c++/pr71959.C -DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0
-foffload=amdgcn-amdhsa  -O0  (test for excess errors)
    UNRESOLVED: libgomp.oacc-c++/pr71959.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  compilation failed to produce
executable
    PASS: libgomp.oacc-c++/pr71959.C -DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0
-foffload=amdgcn-amdhsa  -O2  (test for excess errors)
    [n/a]: libgomp.oacc-c++/pr71959.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  execution test

    FAIL: libgomp.oacc-c++/routine-1-auto.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  (test for excess errors)
    UNRESOLVED: libgomp.oacc-c++/routine-1-auto.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  compilation failed to produce
executable
    PASS: libgomp.oacc-c++/routine-1-auto.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  (test for excess errors)
    [n/a]: libgomp.oacc-c++/routine-1-auto.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  execution test

    FAIL: libgomp.oacc-c++/routine-1-template-auto.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  (test for excess errors)
    UNRESOLVED: libgomp.oacc-c++/routine-1-template-auto.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0 
compilation failed to produce executable
    PASS: libgomp.oacc-c++/routine-1-template-auto.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  (test for excess errors)
    [n/a]: libgomp.oacc-c++/routine-1-template-auto.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  execution test

    FAIL: libgomp.oacc-c++/routine-1-template-trailing-return-type.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  (test
for excess errors)
    UNRESOLVED: libgomp.oacc-c++/routine-1-template-trailing-return-type.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0 
compilation failed to produce executable
    PASS: libgomp.oacc-c++/routine-1-template-trailing-return-type.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  (test
for excess errors)
    [n/a]: libgomp.oacc-c++/routine-1-template-trailing-return-type.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2 
execution test

    FAIL: libgomp.oacc-c++/routine-1-template.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  (test for excess errors)
    UNRESOLVED: libgomp.oacc-c++/routine-1-template.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  compilation failed to produce
executable
    PASS: libgomp.oacc-c++/routine-1-template.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  (test for excess errors)
    [n/a]: libgomp.oacc-c++/routine-1-template.C -DACC_DEVICE_TYPE_gcn=1
-DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  execution test

    FAIL: libgomp.oacc-c++/routine-1-trailing-return-type.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0  (test
for excess errors)
    UNRESOLVED: libgomp.oacc-c++/routine-1-trailing-return-type.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O0 
compilation failed to produce executable
    PASS: libgomp.oacc-c++/routine-1-trailing-return-type.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2  (test
for excess errors)
    [n/a]: libgomp.oacc-c++/routine-1-trailing-return-type.C
-DACC_DEVICE_TYPE_gcn=1 -DACC_MEM_SHARED=0 -foffload=amdgcn-amdhsa  -O2 
execution test

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

Tobias Burnus <burnus at gcc dot gnu.org> changed:

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

--- Comment #1 from Tobias Burnus <burnus at gcc dot gnu.org> ---
The symbol __gxx_personality_v0 is part of libsupc++ – which I believe is not
build to to lacking/restricted C++ support.

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
(In reply to Tobias Burnus from comment #1)
> The symbol __gxx_personality_v0 is part of libsupc++ – which I believe is
> not build to to lacking/restricted C++ support.

It is also exception handling related.  Does amdgcn have EH support?

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[ams at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

--- Comment #3 from Andrew Stubbs <ams at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #2)
> (In reply to Tobias Burnus from comment #1)
> > The symbol __gxx_personality_v0 is part of libsupc++ – which I believe is
> > not build to to lacking/restricted C++ support.
> 
> It is also exception handling related.  Does amdgcn have EH support?

No, it does not. No unwind support, nor sjlj either (the register file is very
large, and has variable size, so it's not trivial to implement, IIUC).

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[tschwinge at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

Thomas Schwinge <tschwinge at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
     Ever confirmed|0                           |1
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2020-04-22

--- Comment #4 from Thomas Schwinge <tschwinge at gcc dot gnu.org> ---
(In reply to Andrew Stubbs from comment #3)
> (In reply to Andrew Pinski from comment #2)
> > (In reply to Tobias Burnus from comment #1)
> > > The symbol __gxx_personality_v0 is part of libsupc++ – which I believe is
> > > not build to to lacking/restricted C++ support.
> > 
> > It is also exception handling related.  Does amdgcn have EH support?
> 
> No, it does not. No unwind support, nor sjlj either (the register file is
> very large, and has variable size, so it's not trivial to implement, IIUC).

Unexpectedly, the same as for nvptx.

So:

(In reply to Thomas Schwinge from comment #0)
> Assuming that's not yet done

(It isn't.)

> can something similar be done as done for
> nvptx, where we force-disable all C++ exceptions code/tables generation, see
> PR86660 GCC trunk r263265, and then GCC trunk r263287?

Seems so.  I successfully tested the following change to
'gcc/common/config/gcn/gcn-common.c'; OK to commit?  If approving this patch,
please respond with "Reviewed-by: NAME <EMAIL>" so that your effort will be
recorded in the commit log, see <https://gcc.gnu.org/wiki/Reviewed-by>.

    +static enum unwind_info_type
    +gcn_except_unwind_info (struct gcc_options *opts ATTRIBUTE_UNUSED)
    +{
    +  return UI_NONE;
    +}

    +#undef  TARGET_EXCEPT_UNWIND_INFO
    +#define TARGET_EXCEPT_UNWIND_INFO gcn_except_unwind_info

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

--- Comment #5 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Cf. also https://gcc.gnu.org/pipermail/gcc-patches/2020-April/544368.html
["config/gcn/unwind-gcn.c (__gxx_personality_v0): New."]

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[ams at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

--- Comment #6 from Andrew Stubbs <ams at gcc dot gnu.org> ---
I think we've decided to with Thomas's approach.

Thomas, please go ahead and commit.

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[cvs-commit at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

--- Comment #7 from CVS Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Thomas Schwinge <tschwinge@gcc.gnu.org>:

https://gcc.gnu.org/g:7f1989249e25af6fc0f124452efa24b3796b767a

commit r10-8033-g7f1989249e25af6fc0f124452efa24b3796b767a
Author: Thomas Schwinge <thomas@codesourcery.com>
Date:   Tue Apr 21 22:39:33 2020 +0200

    [gcn] Set 'UI_NONE' for 'TARGET_EXCEPT_UNWIND_INFO' [PR94282]

    In libgomp offloading testing, this resolves all the 'ld: error: undefined
    symbol: __gxx_personality_v0' FAILs.

            gcc/
            PR target/94282
            * common/config/gcn/gcn-common.c (gcn_except_unwind_info): New
            function.
            (TARGET_EXCEPT_UNWIND_INFO): Define.
            libgomp/
            PR target/94282
            * testsuite/libgomp.c-c++-common/function-not-offloaded.c: Remove
            'dg-allow-blank-lines-in-output'.

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

--- Comment #8 from Tobias Burnus <burnus at gcc dot gnu.org> ---
This was committed to GCC mainline while it was GCC 10 – hence, GCC 10 and 11
have the fix.

Can this PR now be closed?

[Bug target/94282] [amdgcn] ld: error: undefined symbol: __gxx_personality_v0

[tschwinge at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94282

Thomas Schwinge <tschwinge at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|unassigned at gcc dot gnu.org      |tschwinge at gcc dot gnu.org
             Status|NEW                         |RESOLVED
           See Also|https://gcc.gnu.org/bugzill |
                   |a/show_bug.cgi?id=94248,    |
                   |https://gcc.gnu.org/bugzill |
                   |a/show_bug.cgi?id=92713,    |
                   |https://gcc.gnu.org/bugzill |
                   |a/show_bug.cgi?id=71959     |
         Resolution|---                         |FIXED

--- Comment #9 from Thomas Schwinge <tschwinge at gcc dot gnu.org> ---
.