Peter Ebden

Seems legit, although I would rather not add more features to most of the built-in rules at this point. Future development is going to happen in https://github.com/please-build/cc-rules - would you...

I'm not sure I see how this is different from having two separate build actions - one that outputs the `performance_caches` and another one that consumes that as an input?

I don't like the idea of making the build model more complex for this - at present targets get built exactly once (at most once if you include things that...

plz watch does have some debouncing, see https://github.com/thought-machine/please/blob/270c0ece7d2465ea20457afeae5056bfea243709/src/watch/watch.go#L67 It seems reasonable to make that configurable though (currently it's a constant 100ms). > This would be very helpful for avoid building...

> I'm using vscode with autosave, and its saves every ~ 250ms leading to this behaviour OK, I'm fine with providing a flag to make this configurable to work around...

I think this is as expected; depending on the language, we may need to recompile the code in order to instrument it for coverage (e.g. for Go, C or C++)....

`remote_file` supports multiple URLs passed to it, which are tried in sequence. That's the underlying mechanism used to back `python_wheel`. To me that seems sufficient; you simply put your cache...

FYI you can also have more complex structures in the config, for example ``` [lint "go"] target = //linters:go pattern = *.go ```

No stalebot, we're just thinking it over still.

I can see it's finding `cpu_darwin_nocgo.go`. I suspect this is because of cross-compilation (we don't have a `darwin_arm64` worker on CircleCI) although not certain of all the logic there -...