Run 32-bit codegen tests in PR CI?

[nikic]

New codegen or assembly tests have a very high chance of passing PR CI and then failing full checks. There are lots of reasons for that, but one of them is 64-bit vs 32-bit differences.

I wonder whether it would make sense to run these tests against i686-unknown-linux-gnu on PR CI, the same way we do for mir-opt tests. I'd expect codegen tests to run pretty quickly.

[jieyouxu]

I think that is reasonable. cc @marcoieni re. PR CI job time.

[marcoieni]

👋

Do you want to create a new job?
In this case, as long as it doesn't take longer than the other jobs and runs on standard runners, it should be fine I think

Do you want to add this task to an existing job?
In this case, which one? How much it will take to run these additional tests?

[RalfJung]

I don't think we care much either way (other than maybe not growing the PR CI list too long...). Which of the two options works better for you? :D

Who has an overview over how "full" the jobs are and which would be a good job to add this to?

[Kobzol]

I would suggest adding this to pr-check-2, as it already builds the stage 1 compiler and isn't a bottleneck currently.