Work around CI problem from ghcup switch to GHC 9.2 #8503
Conversation
Mikolaj
changed the title
Mikolaj
force-pushed
the
update-CI-to-GHA-on-GHC-9.4
branch
from
23b560f
to
24bb731
Compare
Mikolaj
changed the title
|
If we change "recommended" to x.y.z, then we better have some reminder to upgrade it once in a while. I don't know how to reliably implement such a reminder. I'd much rather we leave recommended in place and deal with upgrades when ghcup bumps it. I also don't get why this patch seemingly both rolls back GHC to 8.10 and bumps the bounds that would allow to build with 9.2? |
|
Oh, I see a comment about doctest now. Well, we could allow-newer, open an issue against doctest, and when upstream fixes the issue remove allow-newer. |
|
Yes, at least a ticket about upgrading doctest would be a good idea. I think we should investigate whether we should stick to the current doctest, change it (there are a couple of competing products), fork it, etc. I don't know which doctest we are using, but if it fails for GHC 9.2, it may not be worth sticking to. Or perhaps we somehow use it wrong or from a wrong branch. Outside the scope of this ticket, certainly. |
|
I will |
|
I'm going to emergency-merge RSN. |
|
Thank you very much, @ulysses4ever and @athas. Merging. |
|
@mergify rebase |
☑️ Nothing to do
|
|
@mergify backport |
❌ No backport have been createdNo destination branches found |
|
@mergify backport 3.8 |
✅ Backports have been created
|
Work around CI problem from ghcup switch to GHC 9.2 (backport #8503)
This is a workaround for #7798 based on the theory that the recent CI breakage is caused by GHA bumping GHC to 9.4 (I'm quite surprised, but OTOH 9.4 is a good release and IIRC an LTS release).
Edit: a correction: it seems to be GHC 9.2. Much less surprising.
Edit2: actually, it seems this was not GHA upgrading GHC, but ghcup changing the "recommended" version and our scripts install that version