RFC: PATCH: PR gas/12049: Unnecessary relaxation

[H.J. Lu]

On Tue, Oct 26, 2010 at 9:50 AM, Dave Korn <dave.korn.cygwin@gmail.com> wrote:
> On 25/10/2010 10:19, H.J. Lu wrote:
>
>>       * gas/i386/i386.exp: Run relax-1 and relax-2 for all 32bit
>>       targets.  Run x86-64-relax-1.
>
>  The relax-1 and relax-2 tests fail on i686-pc-cygwin, because the objdump
> output has labels in it, and no "0x" prefix on the immediates:
>
>> $ /gnu/binutils/obj/gas/testsuite/../../binutils/objdump  -d dump.o
>>
>> dump.o:     file format pe-i386
>>
>>
>> Disassembly of section .text:
>>
>> 00000000 <.text>:
>>    0:   90                      nop
>>    1:   90                      nop
>           [ ... snip ... ]
>>    d:   90                      nop
>>    e:   e9 8d 00 00 00          jmp    a0 <.text+0xa0>
>>   13:   90                      nop
>
> ... while the dump script files don't:
>
>> $ cat /gnu/binutils/src/gas/testsuite/gas/i386/relax-1.d
>> #name: i386 relax 1
>> #objdump: -dw
>>
>> .*: +file format .*
>>
>>
>> Disassembly of section .text:
>>
>> 0+ <.text>:
>> #...
>>    e:   e9 8d 00 00 00          jmp    0xa0
>> #...
>>   21:   eb 7d                   jmp    0xa0
>> #...
>>   a0:   90                      nop
>> #pass
>
>  The attached patch makes it pass on i686-pc-cygwin, and I checked on
> i686-pc-linux-gnu (both native and cross to x86_64-pc-linux-gnu) that it
> didn't break the test there.  OK?
>
> gas/testsuite/ChangeLog:
>
>        * i386/relax-1.d: Generalize regexes a bit.
>        * i386/relax-2.d: Likewise.
>

OK.  Thanks.


-- 
H.J.