Re: [PATCH] hw/display/virtio-gpu: Fix empty blob discrimination

On 2026/07/28 16:47, Michael Tokarev wrote:
> On 7/28/26 09:54, Akihiko Odaki wrote:
> ...
>> The consequence of the bug this patch fixes is a NULL pointer 
>> dereference and denial-of-service. docs/devel/stable-process.rst says:
>>
>>  > Generally, the following patches are considered stable material:
>>  >
>>  > * Patches that fix severe issues, like fixes for CVEs
>>  > * Patches that fix regressions
> 
> In these two cases, changes are *definitely* for-stable.  Other changes -
> fixing bugs but without regressions or CVEs - might also be for stable,
> but it is more case-by-case material.
> 
> You haven't quoted the very next paragraph from this document though:
> 
>    If you think the patch would be important for users of the current 
> release
>    (or for a distribution picking fixes), it is usually a good candidate
>    for stable.
> 
> Without "regular bug fixes" in stable we wont survive ;)
> 
>> And I don't think this match with the description. It still does fix a 
>> bug, and I think you can cleanly backport it and you tend to backport 
>> such a patch.
> What I'm thinking is more like - all at once - does the bug affects users
> (is the subsystem in question actually used or not (it might be some
> very niche thing); is it a fast-moving target so only very recent code
> makes sense in the first place and so there's no reason to back-port
> fixes; can the bug be triggered in real life or just theoretical (like,
> it is not a problem in current code but future changes might trigger it);
> is it more a cosmetic thing or something like a crash; and so on.  And
> on the other hand, whenever this is a trivial fix or something more
> complex which might cause bugs by its own.
> 
> In this particular case, the fix is trivial and the subsystem is in
> use, but if this bug can be seen in real life I still don't know.
> 
> At least now I know why you haven't considered it for-stable in
> the first place - because you assumed only CVE and regression fixes
> should go there :)

Indeed I forgot to quote the paragraph. This bug can only be triggered 
by a malfunctional guest and does not have a practical impact. I usually 
consider a fix that address a user-provided bug report for stable.

Regards,
Akihiko Odaki