seems to be passing locally for me,

[final valley]

It's weird cause it failed over and over in two PRs I have (both of which are unrelated), then it failed locally for me. I also saw it fail in your git PR here, which is also unrelated

But I just re-ran locally a few mins ago and it passed...

Oh I just saw your edit

Cool

Maybe it only fails between the hours of 2 and 4 pm Pacific time

[tribal crag]

Nah, maybe it's linked to dang's sdk that's unpinned (this test installs old github.com/dagger/jest commits that has it unpinned). Alex has been pushing today on it, very suspicious

[final valley]

Ah! that makes more sense 😄

[lofty dock]

yep we need to pin all the toolchains 😅 i'll do that tomorrow

[tribal crag]

Ok, I confirm that this is indeed a transient Dang's SDK failure.

This was the error I got:

27  : ┆ ┆ [21.5s] | entrypoint/main.go:263:15: cannot use fnName (variable of type string) as *dang.Symbol value in struct literal
27  : ┆ ┆ [21.5s] | entrypoint/main.go:313:13: cannot use fmt.Sprintf("load%sFromID", modType.Named) (value of type string) as *dang.Symbol value in struct literal

The chain of events:

1. The jest module has "sdk": {"source": "github.com/vito/dang/dagger-sdk"} with no pin
2. Between 4:37 and 6:46pm, on Dang's repo, main was broken by https://github.com/vito/dang/commit/e5942b3, which changed Select.Field from string to *Symbol (same error I got)
3. But the entrypoint wasn't fixed until https://github.com/vito/dang/commit/0e40af1

Any CI run that fetched during that ~2 hour window got a compile error (which was the case of your PRs Erik). And, after that fix, no CI run had the problem

--

Update: Dang's CI was covering this edge case. We should totally pin the dependency:

1. Pin on dagger/jest
2. Bump the commit(s) on the test