Peter Ebden

I'd vote in favour of an explicit delete in preference to overloading UpdateActionResult. Agreed that this is a useful thing in practice. We might want to consider extending ActionCacheUpdateCapabilities to...

Can you please check whether this has some impact on what the Go package driver should be doing with these paths? Feels like it might break it again (although I...

I see this as two somewhat separate questions - what does plz itself need in order to run, and what do the various build rules expect (as you say the...

We could probably express this as an in-repo test rather than a separate Github Action?

I don't think this is a security issue - if you import a plugin you pretty implicitly trust that plugin (it's running a bunch of build actions on your behalf...

I might argue for the name being `--out_dir` to be clearer about what it'll do. Like the idea otherwise though!

I'm not convinced we need this. The build language is meant to be minimal and it isn't meant to need to do a lot of introspection - the syntax is...

No, this isn't intended, a test can return nonzero and produce the results file, which should then get interpreted - generally the expectation is that the exit code matches the...

Hush, bot. I'm not sure what the immediate future is here - the main issue is that I'm not sure how to update the C++ plugin to work in the...

I don't think we do hardcode any paths like `/bin/mv` or similar anywhere. The normal answer would be to set the PATH in the config: ``` [build] path = /usr/local/bin:/usr/bin:/bin:/nix/store/nonsense/usr/bin...