powerpc build failures with GCC trunk

powerpc build failures with GCC trunk

[Joseph Myers]

My glibc bot is showing build failures for powerpc with GCC trunk.  For 
powerpc64 (BE and LE), glibc fails to build, with errors of the form:

../sysdeps/powerpc/powerpc64/power8/strcmp.S: Assembler messages:
../sysdeps/powerpc/powerpc64/power8/strcmp.S:55: Error: unrecognized opcode: `cmpb'

For 32-bit, the failures are in the testsuite:

/tmp/ccs8hquE.s:43: Error: unrecognized opcode: `mfppr32'
(building tst-set_ppr.o).

Without having specifically tested, I'm guessing that

2019-05-22  Alan Modra  <amodra@gmail.com>

        * config/rs6000/rs6000.h (ASM_OPT_ANY): Define.
        (ASM_CPU_SPEC): Conditionally add -many.
        * config/rs6000/rs6000.c (rs6000_machine): New static var.
        (rs6000_machine_from_flags, emit_asm_machine): New functions..
        (rs6000_file_start): ..extracted from here, and modified to
        test all ISA bits.
        (rs6000_output_function_prologue): Emit .machine as necessary.

is the change causing this.  Could the powerpc maintainers look at this 
and fix glibc / tests to build with current GCC (ensuring appropriate CPU 
options or assembler directives are passed to enable the instructions in 
question, I suppose)?

-- 
Joseph S. Myers
joseph@codesourcery.com

Re: powerpc build failures with GCC trunk

[Gabriel F. T. Gomes]

On Thu, May 23 2019, Joseph Myers wrote:
>
> is the change causing this.  Could the powerpc maintainers look at this 
> and fix glibc / tests to build with current GCC (ensuring appropriate CPU 
> options or assembler directives are passed to enable the instructions in 
> question, I suppose)?

Thanks for the heads-up, I'm working on a patch.