Re: [PATCH v3 3/7] migration: Fix parsing of s390 stream

[Fabiano Rosas]

Fabiano Rosas <farosas@suse.de> writes:

> Michael Tokarev <mjt@tls.msk.ru> writes:
>
>> 13.01.2025 11:19, Thomas Huth wrote:
>>> On 13/01/2025 08.51, Michael Tokarev wrote:
>>
>>>> Picked up:
>>>>
>>>>   1/7 migration: Add more error handling to analyze-migration.py
>>>>   3/7 migration: Fix parsing of s390 stream
>>>>   7/7 s390x: Fix CSS migration
>>>>
>>>> but still getting the same error:
>>>>
>>>>   https://gitlab.com/mjt0k/qemu/-/jobs/8832218999 (9.2 branch)
>>>>   https://gitlab.com/mjt0k/qemu/-/jobs/8832224338 (9.1 branch)
>>> 
>>> Blindly guessing: You need now patch 4/7 and 5/7, too? ... hopefully 
>>> Fabiano could reply and point you to the exact set of patches that you 
>>> need...
>>
>> Yes, after picking up ALL 7 out of 7 in this patchset (b/c even the
>> trivial 2/7, "migration: Remove unused argument in vmsd_desc_field_end",
>> is needed or else the subsequent fixes doesn't apply) it is now green
>> finally, eg, https://gitlab.com/mjt0k/qemu/-/jobs/8832849813
>>
>> Fabiano, what do you think, - should the whole patchset be picked up
>> for 9.2 and 9.1?
>
> Yeah, sorry, I was focused on unbreaking the migration PR and added a
> bunch of patches without thinking of stable.
>
> So the s390x regression (1/7) is 9.1, but 9.0 already had the broken

7/7

> analyze-script.py (3/7) and the broken array compression code (4-6/7).
>
> We definitely need 1/7 for 9.1 and 9.2. The rest of the series is "just"

7/7

> to avoid breaking the tests. If you can apply it easily I think that's
> preferable. Otherwise maybe we could disable the analyze-migration.py
> test for stable? I can also work on a backport if needed. Let me know
> what you prefer.
>
>>
>> Thanks,
>>
>> /mjt