[prev in list] [next in list] [prev in thread] [next in thread] 

List:       binutils
Subject:    Re: [Committed][PATCH v1 1/1][Binutils] aarch64: Fix the 2nd operand in gcsstr and gcssttr instructi
From:       Srinath Parvathaneni <srinath.parvathaneni () arm ! com>
Date:       2024-02-29 21:39:12
Message-ID: 3061a86a-e94d-4e5f-bf3a-c8210da20781 () arm ! com
[Download RAW message or body]

Hi

On 2/14/2024 9:24 AM, Jan Beulich wrote:
> On 13.02.2024 19:03, srinath wrote:
>> The assembler wrongly expects plain register name instead of
>> memory-form 2nd operand for gcsstr and gcssttr instructions.
>> This patch fixes the issue.
>>
>> Regression testing for aarch64-none-elf target and found no regressions.
>>
>> Ok for binutils-master? Also ok to be backported to binutils-2.42 branch?
> Looks good to me, but I'm not an Arm64 maintainer. I'd prefer if one of
> the two (you didn't even Cc Marcus) would ack the patch, but in case you
> don't hear back within a week, feel free to put in on both branches.
> That said, ...

I have committed the patch to both "master" and "binutils-2_42-branch".

>> ---
>>   gas/testsuite/gas/aarch64/gcs-1-bad.l | 48 +++++++++++++--------------
> ... the need for you to alter the expectations here indicates that the
> testcase itself likely was having too specific expectations. For the
> intended purpose, the exact set of operands isn't of interest and hence
> would likely better have been omitted in the first place. I'm aware
> though that it's a common approach to simply not edit expectations from
> obtained tool output any further than to add the necessary escaping.
> Yet I'm trying to encourage people to put in a little more effort, in
> an attempt to limit follow-on effort for themselves or others.
>
> The more thoroughly one zaps irrelevant parts from expectations, the
> more likely it also is that one might spot actual issues - this isn't
> the case here, i.e. I'm merely trying to provide some further
> background, but I've seen it numerous times that expectations were
> derived largely blindly from tool output, without checking that the
> output is actually correct/sensible in the first place. (I'm afraid
> I, too, have been guilty of that in a few cases.)
>
> Jan

Thanks for the feedback, in my next set of patches (bug fixes to sve2p1) 
I shall explain

the logic I'm using to write the tests (covering most of the encoding). 
Also, I'm comparing

the outcome of the disassembler with other tools, to avoid these kind of 
errors.

Regards,

Srinath.

>>   gas/testsuite/gas/aarch64/gcs-1.d     | 48 +++++++++++++--------------
>>   gas/testsuite/gas/aarch64/gcs-1.s     |  2 +-
>>   opcodes/aarch64-tbl.h                 |  4 +--
>>   4 files changed, 51 insertions(+), 51 deletions(-)
>>
[prev in list] [next in list] [prev in thread] [next in thread]