[Bug target/67417] New: powerpc64 bootstrap with -mcmodel=small results in linker error

[Bug target/67417] New: powerpc64 bootstrap with -mcmodel=small results in linker error

[amodra at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67417

            Bug ID: 67417
           Summary: powerpc64 bootstrap with -mcmodel=small results in
                    linker error
           Product: gcc
           Version: 5.2.1
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: amodra at gmail dot com
  Target Milestone: ---

To test recent powerpc64 linker changes I attempted to bootstrap gcc-5 with
-mcmodel=small, and ran into

/home/amodra/gnu/powerpc64-linux/bin/ld: libbackend.a(sel-sched-ir.o): In
function `loop_iterator::loop_iterator(loop**, unsigned int)':
sel-sched-ir.c:(.text+0x116f0): call to `loop_iterator::loop_iterator(loop**,
unsigned int)' lacks nop, can't restore toc; (-mcmodel=small toc adjust stub)

On investigating, I see there really isn't a nop, and the call is to a group
with a different toc pointer.  This despite the destination being defined in
the same file.  However, it is weak and there's another definition.

nm -o gcc/*.o | grep _ZN13loop_iteratorC1EPP4loopj
gcc/ipa-inline-analysis.o:0000000000000420 W _ZN13loop_iteratorC1EPP4loopj
gcc/sel-sched-ir.o:00000000000013e0 W _ZN13loop_iteratorC1EPP4loopj

[Bug target/67417] powerpc64 bootstrap with -mcmodel=small results in linker error

[amodra at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67417

Alan Modra <amodra at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Target|                            |powerpc64-linux
             Status|UNCONFIRMED                 |ASSIGNED
   Last reconfirmed|                            |2015-09-01
           Assignee|unassigned at gcc dot gnu.org      |amodra at gmail dot com
     Ever confirmed|0                           |1

[Bug target/67417] powerpc64 bootstrap with -mcmodel=small results in linker error

[amodra at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67417

--- Comment #1 from Alan Modra <amodra at gmail dot com> ---
BTW, the linker error message doesn't identify the correct function, which is
looking like a linker bug.

[Bug target/67417] powerpc64 bootstrap with -mcmodel=small results in linker error

[amodra at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67417

--- Comment #3 from Alan Modra <amodra at gmail dot com> ---
Yes, exactly.  When a COMDAT group in another file is chosen, the rs6000
predicate current_file_function_operand gets the wrong answer.  COMDAT group
functions use weak symbols, and weak symbols more generally have the behaviour
exhibited here by COMDAT group functions: A symbol defined in the same object
file might be overridden by the same symbol defined in another file.

[Bug target/67417] powerpc64 bootstrap with -mcmodel=small results in linker error

[amodra at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67417

--- Comment #4 from Alan Modra <amodra at gcc dot gnu.org> ---
Author: amodra
Date: Tue Sep  1 23:03:19 2015
New Revision: 227386

URL: https://gcc.gnu.org/viewcvs?rev=227386&root=gcc&view=rev
Log:
[RS6000] Weak functions may not be file local

A weak symbol defined in the current object file may not turn out to
be the definition used at link time, if other copies of the symbol
exist.  This means they can't be considered file local.

        PR target/67417
        * config/rs6000/predicates.md (current_file_function_operand): Don't
        return true for weak symbols.
        * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Likewise.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config/rs6000/predicates.md
    trunk/gcc/config/rs6000/rs6000.c

[Bug target/67417] powerpc64 bootstrap with -mcmodel=small results in linker error

[amodra at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67417

--- Comment #5 from Alan Modra <amodra at gcc dot gnu.org> ---
Author: amodra
Date: Tue Sep  1 23:04:58 2015
New Revision: 227387

URL: https://gcc.gnu.org/viewcvs?rev=227387&root=gcc&view=rev
Log:
[RS6000] Weak functions may not be file local

A weak symbol defined in the current object file may not turn out to
be the definition used at link time, if other copies of the symbol
exist.  This means they can't be considered file local.

        PR target/67417
        * config/rs6000/predicates.md (current_file_function_operand): Don't
        return true for weak symbols.
        * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Likewise.


Modified:
    branches/gcc-5-branch/gcc/ChangeLog
    branches/gcc-5-branch/gcc/config/rs6000/predicates.md
    branches/gcc-5-branch/gcc/config/rs6000/rs6000.c

[Bug target/67417] powerpc64 bootstrap with -mcmodel=small results in linker error

[amodra at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67417

--- Comment #6 from Alan Modra <amodra at gcc dot gnu.org> ---
Author: amodra
Date: Tue Sep  1 23:25:48 2015
New Revision: 227389

URL: https://gcc.gnu.org/viewcvs?rev=227389&root=gcc&view=rev
Log:
[RS6000] Weak functions may not be file local

A weak symbol defined in the current object file may not turn out to
be the definition used at link time, if other copies of the symbol
exist.  This means they can't be considered file local.

        PR target/67417
        * config/rs6000/predicates.md (current_file_function_operand): Don't
        return true for weak symbols.
        * config/rs6000/rs6000.c (rs6000_function_ok_for_sibcall): Likewise.


Modified:
    branches/gcc-4_9-branch/gcc/ChangeLog
    branches/gcc-4_9-branch/gcc/config/rs6000/predicates.md
    branches/gcc-4_9-branch/gcc/config/rs6000/rs6000.c

[Bug target/67417] powerpc64 bootstrap with -mcmodel=small results in linker error

[amodra at gmail dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67417

Alan Modra <amodra at gmail dot com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |FIXED
   Target Milestone|---                         |4.9.4

--- Comment #7 from Alan Modra <amodra at gmail dot com> ---
Fixed all active branches.