New (APX related) assembler testsuite failures

New (APX related) assembler testsuite failures

[Nick Clifton]

Hi Cui, Hi Jan,

   I am seeing two new failures in the gas testsuite:

FAIL: x86_64 APX_F EVEX-Promoted insns
FAIL: x86_64 APX_F EVEX-Promoted insns (Intel disassembly)

   These only occur in x86_64 toolchains configured for PE targets
   (eg x86_64-pc-mingw64) and are happening because these targets
   pad out code sections in order to allow for section alignment.
   The testsuite log shows:

extra lines in tmpdir/dump.out starting with "^ b8c:	90                   	nop$"
EOF from /work/sources/binutils/upstream/current/gas/testsuite/gas/i386/x86-64-apx-evex-promoted-intel.d

   Could you take a look and maybe tweak something please ?

Cheers
   Nick

Re: New (APX related) assembler testsuite failures

[Jan Beulich]

On 19.01.2024 15:32, Nick Clifton wrote:
> Hi Cui, Hi Jan,
> 
>    I am seeing two new failures in the gas testsuite:
> 
> FAIL: x86_64 APX_F EVEX-Promoted insns
> FAIL: x86_64 APX_F EVEX-Promoted insns (Intel disassembly)
> 
>    These only occur in x86_64 toolchains configured for PE targets
>    (eg x86_64-pc-mingw64) and are happening because these targets
>    pad out code sections in order to allow for section alignment.
>    The testsuite log shows:
> 
> extra lines in tmpdir/dump.out starting with "^ b8c:	90                   	nop$"
> EOF from /work/sources/binutils/upstream/current/gas/testsuite/gas/i386/x86-64-apx-evex-promoted-intel.d
> 
>    Could you take a look and maybe tweak something please ?

I'll deal with that on Monday. I'm surprised though, as I thought I
had sorted such already. But it must have been some other testcase
then ...

Jan

Re: New (APX related) assembler testsuite failures

[H.J. Lu]

On Fri, Jan 19, 2024 at 6:35 AM Jan Beulich <jbeulich@suse.com> wrote:
>
> On 19.01.2024 15:32, Nick Clifton wrote:
> > Hi Cui, Hi Jan,
> >
> >    I am seeing two new failures in the gas testsuite:
> >
> > FAIL: x86_64 APX_F EVEX-Promoted insns
> > FAIL: x86_64 APX_F EVEX-Promoted insns (Intel disassembly)
> >
> >    These only occur in x86_64 toolchains configured for PE targets
> >    (eg x86_64-pc-mingw64) and are happening because these targets
> >    pad out code sections in order to allow for section alignment.
> >    The testsuite log shows:
> >
> > extra lines in tmpdir/dump.out starting with "^ b8c:  90                      nop$"
> > EOF from /work/sources/binutils/upstream/current/gas/testsuite/gas/i386/x86-64-apx-evex-promoted-intel.d
> >
> >    Could you take a look and maybe tweak something please ?
>
> I'll deal with that on Monday. I'm surprised though, as I thought I
> had sorted such already. But it must have been some other testcase
> then ...
>
> Jan

I pushed the fix into master branch and backported it to
binutils-2_42-branch.

-- 
H.J.

Re: New (APX related) assembler testsuite failures

[Jan Beulich]

On 19.01.2024 15:46, H.J. Lu wrote:
> On Fri, Jan 19, 2024 at 6:35 AM Jan Beulich <jbeulich@suse.com> wrote:
>>
>> On 19.01.2024 15:32, Nick Clifton wrote:
>>> Hi Cui, Hi Jan,
>>>
>>>    I am seeing two new failures in the gas testsuite:
>>>
>>> FAIL: x86_64 APX_F EVEX-Promoted insns
>>> FAIL: x86_64 APX_F EVEX-Promoted insns (Intel disassembly)
>>>
>>>    These only occur in x86_64 toolchains configured for PE targets
>>>    (eg x86_64-pc-mingw64) and are happening because these targets
>>>    pad out code sections in order to allow for section alignment.
>>>    The testsuite log shows:
>>>
>>> extra lines in tmpdir/dump.out starting with "^ b8c:  90                      nop$"
>>> EOF from /work/sources/binutils/upstream/current/gas/testsuite/gas/i386/x86-64-apx-evex-promoted-intel.d
>>>
>>>    Could you take a look and maybe tweak something please ?
>>
>> I'll deal with that on Monday. I'm surprised though, as I thought I
>> had sorted such already. But it must have been some other testcase
>> then ...
> 
> I pushed the fix into master branch and backported it to
> binutils-2_42-branch.

Thanks for taking care of this. Nevertheless I have to admit I'm a
little disappointed that you didn't check the other APX testcases
as well. I'll commit an adjustment to the PUSH2/POP2 testcases in a
few minutes. Which gets me to the 2nd point: Please can you "blame"
the correct patch when writing titles / descriptions of new ones
(if you feel like doing any blaming in the first place)? It was not
my patch which introduced the issue. Instead when the testcases
were first added the trap was set out for someone to fall into.
Much like iirc was the case for the USER_MSR testcases.

Jan