On 4/6/2026 05:15, Paul Guyot wrote:
Hello,
Should we try to circumvent GitHub actions 6 hours limit?
I don't know how we could do that, do you? We use GitHub's free hosted
runners, and AFAIK that limit cannot be bypassed unless you either pay
or supply your own runners.
The problem arises with clang updates that time out, but these timeouts may
hide issues with flang.
Alternatively we could split the llvm subports into separate files so CI only
rebuilds what was updated.
That may be the simplest way, though it comes with its own problems
since each port would have to be updated in a separate PR.
I tried a prototype of waves, but I don't think it's very pretty:
https://github.com/pguyot/macports-ports/actions/workflows/wave-build.yml
Also should we also have intel runners on GHA? I wish we had a build bot so
that intel Tahoe binaries were available, but at least we can ensure everything
compiles on Intel?
I think once again that we would have to supply our own runner, and it
would have to be on the small subset of Intel Mac hardware that can run
Tahoe, unless there is a VM that can get around that.
- Josh
