Pierrick Bouvier <[email protected]> writes:

> On 7/14/2026 3:47 AM, Ilya Leoshkevich wrote:
>> 
>> 
>> On 7/10/26 23:43, Pierrick Bouvier wrote:
>>> Tested-by: Alex Bennée <[email protected]>
>>> Signed-off-by: Pierrick Bouvier <[email protected]>
>>> ---
>>>   tests/tcg/s390x/float_convd.ref | 988 ++++++++++++++++++++++++++++++++
>>>   tests/tcg/s390x/float_convs.ref | 748 ++++++++++++++++++++++++
>>>   tests/tcg/s390x/float_madds.ref | 768 +++++++++++++++++++++++++
>>>   3 files changed, 2504 insertions(+)
>>>   create mode 100644 tests/tcg/s390x/float_convd.ref
>>>   create mode 100644 tests/tcg/s390x/float_convs.ref
>>>   create mode 100644 tests/tcg/s390x/float_madds.ref
>> Would it be possible to add an explanation to the commit message why
>> this is needed now and wasn't needed before?
>
> Before the series, we had only some arch that had those files. I have no
> clue why to be very honest, out of "people forgot to add them".

It was totally that - generally because we didn't have access to
reference hardware at the time.

>
> Makefiles were "smart" and skipped the output compare when such files
> were not available. This results in arch being very inconsistent with
> some providing all files, some providing no files, and some providing
> some files. There was even one case of a misnamed file that result in a
> skipped check.
>
> For consistency, the current series adds all missing reference files for
> all arch, and make their presence mandatory.
>
> Does it look like a good reason to you?
>
> Regards,
> Pierrick

-- 
Alex Bennée
Virtualisation Tech Lead @ Linaro

Reply via email to