QA and armhf

QA and armhf

[Andreas Enge]

Hello,

after the core-updates merge, QA is not keeping up; on
   https://qa.guix.gnu.org/patches
I see the message
   Builds for new patch series suspended as master branch substitute availability is low for: armhf-linux 

Maybe buying new ARM machines could be part of the solution, but I also
think that armhf (and i686) should not hold up progress of Guix, or, as it
stands, risk the Guix quality since people end up pushing regardless of
the QA status (if I remember well, the manual suggests a one-week waiting
period, which right now causes frustrating busy work - create an issue,
send the patch series if there are several commits, wait a week while
knowing that QA will not handle the patch, push, close the issue).
Notice that ci has already completely dropped armhf substitutes:
   https://qa.guix.gnu.org/branch/master

Could we drop this suspension of QA depending on armhf and i686 weather
status on the master branch, and only keep it for aarch64 and x86_64?

Andreas

Re: QA and armhf

[Andreas Enge]

PS: At https://git.guix-patches.cbaines.net/guix-patches/
    the latest issue branch seems to be issue-71974
    Is this related to the master weather, or a bug?

Re: QA and armhf

[John Kehayias]

Hi Andreas,

On Fri, Sep 13, 2024 at 12:37 PM, Andreas Enge wrote:

> Hello,
>
> after the core-updates merge, QA is not keeping up; on
>    <https://qa.guix.gnu.org/patches>
> I see the message
>    Builds for new patch series suspended as master branch substitute
> availability is low for: armhf-linux
>
> Maybe buying new ARM machines could be part of the solution, but I also
> think that armhf (and i686) should not hold up progress of Guix, or, as it
> stands, risk the Guix quality since people end up pushing regardless of
> the QA status (if I remember well, the manual suggests a one-week waiting
> period, which right now causes frustrating busy work - create an issue,
> send the patch series if there are several commits, wait a week while
> knowing that QA will not handle the patch, push, close the issue).
> Notice that ci has already completely dropped armhf substitutes:
>    <https://qa.guix.gnu.org/branch/master>
>
> Could we drop this suspension of QA depending on armhf and i686 weather
> status on the master branch, and only keep it for aarch64 and x86_64?
>
> Andreas

Thanks for bringing this up, as I was wondering the same (our level of
expectations for substitute coverage before merging new branches) and
sent a message now to the list about mesa-updates.

As I don't (currently) use Guix on anything but x86_64-linux, I know I
am biased in wanting to allow for more speed and momentum. However, I
worry that moving like that would risk making thing unmaintainable for
other architectures as they get left behind/have unnoticed issues,
which I wouldn't want to impose either.

So, the practical question is what we have resources for and that
users can rely on. Definitely x86_64, i686 too so far (though we are
quickly becoming the only distro), and then after...? Whatever we
decide we should clearly communicate this on the website and manual.
And if we do allow for other architectures to fall behind in coverage,
I hope that is temporary until we have more hardware to provide
similar support.

John