Saving cache is skipped

Thank you for trying to make this useful action, but it does not seem to be working. At the first build attempt that failed, it cached something (as much as it built to that point), but all the following pushes on that branch do not update the incomplete cache even if the build finishes successfully. See for ex: https://github.com/cdcseacave/openMVS/actions/runs/3513334392/jobs/5886034774

Am I using it wrong or I do not understand the purpose of this action?

PS: I also noticed that the buildtrees and downloads folder are cached, which seems to me like a big waste of space (actually the build fails with run out of space error if I do not remove these folders manually every time). Why do you cache these?

[lukka]

@rcoup @cdcseacave i am taking my time to follow up on this, I won’t forget!

[lukka]

What do you mean by “cloud cache” — the GHA cache?

Yes, the GHA cache.

Workflows all running concurrently already clobber each other’s cache entries, that’s inevitable, and happens already if you push several commits requiring the (same) vcpkg rebuild. Caches are key-value stores though, so latest write “wins” — run-vcpkg/other actions don’t re-read the cache entries they’ve written earlier in a job, so doesn’t seem like a problem?

As far as I know the first wins, as the cache is immutable by design.