Hello,
Christopher Baines <[email protected]> writes:
> Say ci.guix.gnu.org did start doing builds based off of Pull Requests
> and sending that data to data.qa.guix.gnu.org, […]
I don’t think that’s a reasonable plan.
Cuirass works well for Guix-Science & co. on a single architecture, but
I suspect it would not scale well to build all of Guix for all the pull
requests on all the architectures. (Would be interesting to try if we
could get a sufficiently beefy test instance somewhere.)
The better plan I think is to:
1. Get Codeberg webhooks sent to data.qa.guix.gnu.org. This is what
<https://codeberg.org/guix/data-service/pulls/2> implements but it
has yet to be deployed. If I’m not mistaken, that’s about all we
need to get pull requests built.
2. Adjust QA-Frontpage so it can display info about pull requests in
addition to Debbugs entries. I haven’t checked what it would take.
3. Adjust the Data Service (?) so it can send “reviews” to pull
requests with a summary of what built. (This is a capability
Cuirass recently gained¹, but the information it has is less
precise than what the Data Service has.)
I think this is all within reach, with few unknowns, and can be done
incrementally (#2 and #3 are nice to have but #1 is what matters most).
Ludo’.
¹ Example:
<https://codeberg.org/guix-science/guix-science/pulls/93#issuecomment-5088387>.