CI: add GHC 9.6.1 #8840
CI: add GHC 9.6.1 #8840
Conversation
|
@ulysses4ever: would perhaps bumping cabal on master to 3.11 help with this? If so, I can prioritize. |
|
@Mikolaj oh it hasn't been? I think that's the reason of one of the failures, yes (number three in the "Usual blockers" list in OP). There's no rush though: do it when you feel it's the time. This PR can wait. It seems that putting cache-restore in front did help with haskell/actions#203 |
|
Version bumps on master: #8844 |
ulysses4ever
force-pushed
the
ci-9.6.1
branch
2 times, most recently
from
c8a9dd7
to
28ff5e7
Compare
|
I rolled back cabal install in CI to 3.8 because on this branch the versions have not been updated and therefore cabal-testsuite requires 3.8. But I get a failure "All install methods for cabal 3.8.1.0 failed" foir some reason...: It didn't happen with 3.10... |
|
This seems broken upstream. Let's see if it works fine (again with 3.10) once the version bump PR lands. |
|
@ulysses4ever wrote:
Well this is exactly haskell/actions#202, as analysed by @Mistuke: The setup action installs cabal 3.10.1.0 already with ghc, and then the subsequent attempt to install 3.8.1.0 chokes when trying to handle the 3.10.1.0 generated config file. |
|
@andreasabel well, I must say: you put it much more clear than I was able to distill from that thread. Thank you! I thought some sort of workaround had been deployed though? Mistuke said that it should be 3.8 now? I can try to explicitly |
|
Otherwise, thanks for the review! I'll look into |
|
The master version bump PR is merged. |
|
@ulysses4ever : I fixed the haskell setup action |
|
Oh no, why is it trying to build Cabal-syntax-3.8.1.0... https://github.com/haskell/cabal/actions/runs/4409245230/jobs/7728945807#step:13:788 |
How about rebasing on master (with the version bumps)? |
Well, this is a good point and a possibility I had in mind. But I thought that moving one thing at a time (in this case, the version of GHC) may be easier. I can try to rebase, yes. I may not have time to for a day or two though. |
|
No rush. BTW, yet another possibility for the CI cache problem resolution may be to reset the whole cache (first only in the PR, to verify) and so switch to XDG completely. Then the result of |
ulysses4ever
force-pushed
the
ci-9.6.1
branch
2 times, most recently
from
3f8fa8c
to
d668620
Compare
ulysses4ever
force-pushed
the
ci-9.6.1
branch
2 times, most recently
from
ab06a99
to
1f8606c
Compare
|
All right, useless GitHub wouldn't run CI because of conflicts with master (TIL...), so I had to rebase on master and, therefore, pull the versions bump from #8844. Now let's see... |
|
@Mikolaj on the topic of bumping minor versions. Our GHC_FOR_RELEASE is currently at 9.2.3. Do you want me to bump it to 9.2.7 or leave the 9.2 series alone. We could also have two 9.2.x compilers, I presume. |
|
Yes, I think bumping it all makes sense and I'd keep only the newest version. I waited with touching the gitlab script until after the release, so this is a good moment now. |
haskell/acitons support prereleases since recently, and so, a separate workflow is not necessary.
It hasn't been used for a while now.
Partially revert 78cbeba.
mergify
bot
added
the
merge delay passed
|
@mergify backport 3.10 |
✅ Backports have been created
|
|
Hurray! |
|
Yep, this was a long ride. Not sure when I have energy to attend to a backport... |
Backport of haskell#8840
Backport of haskell#8840
Backport of haskell#8840
Time to rock 🤘
Usual Blockers
base,unix, etc.allow-newer:windnscabal-testsuiteneeds the latest release of cabal-install (more discussion: Default cabal.project in cabal repo fails with cabal-install older than latest release; hard to diagnose #8133).Unusual Blockers
cabal updateis not run? actions#203 (fixed upstream)TODO
create an issue for the "double directory" issue and make sure *-suite validation is reenabled after it's solved.
Bump minor versions of the previous computer series. Don't forget about
GHC_FOR_RELEASEand the friends.