Erik Sipsma

@vito shit sorry you are right, I was going through my 1000 open tabs too quickly and misread this, re-opening!

I have not seen this one since July I think, so closing for now

@jedevc after my commit for https://github.com/dagger/dagger/pull/9730#discussion_r1984336471 I saw a flake in `TestCall/TestGitMod`, which I assumed at first was from my commit. It turns out it was happening before my commit...

> My issue with an array of exit codes, is that it makes it difficult to say "catch all errors". I think it's cool to have that level of control,...

> Yeah but at that point, why have the array at all? If you let me catch all error codes, and know what error code I caught, I can implement...

> Re: deep in the buildkit guts. That's something I meant to ask. Don't we already have the ability to do this in the shim, instead of buildkit itself? Not...

Still working on root cause, just a few more significant things in meantime: I looked at the overlay layers for the /src mount where the problematic go.mod file is, I...

Picking back up today, thankfully was able to eventually repro again (though it took a long time). Added this to the milestone because whatever is happening here seems deeply wrong,...

Noticed that the behavior seems to generally be that once this get hit once, it's easy to hit again for a while. But it will go through long periods (literally...

Just confirmed that when the error happens, the `TestExecWithEntrypoint` client that hit the error does indeed end up re-using a local sync cache ref from a client that originally ran...