Josh Reed

> Question on this: Is this for use cases where pre-commit is already in use for the repo? Yeah that was the intention > Secondary question: Do we have anywhere...

Btw, how do I add a category label? Is that a right given only to maintainers?

Is this really because Union return types (other than possibly the degenerate case of `Optional`) are a bad idea design-wise? By which I mean that the return type should usually...

I guess my question is more about whether this is the kinda documentation we want. Returning unions tells mypy that users need to check the return values of everything before...

Note: this is reproducible in the pants repo itself on https://github.com/pantsbuild/pants/releases/tag/release_2.13.0a1

[changed-dep.redacted.trace.log](https://github.com/pantsbuild/pants/files/9079974/changed-dep.redacted.trace.log) Trace logs... command was ``` ./pants --no-pantsd list --changed-since=main --changed-dependees=transitive -ltrace 2> /var/tmp/changed-deps.trace.log ``` and then some company specific stuff was redacted

So `./pants dependencies ::` takes about the same amount of time, so I think that assumption is valid. I dug deeper into the "gap" in the DEBUG log by looking...

I meant trace level logging, which is what I was posting excerpts from. But I can give a try to a profiler sometime later. I'm on vacation atm so it...

[pants-dependencies.speedscope.txt](https://github.com/pantsbuild/pants/files/9104354/pants-dependencies.speedscope.txt) Uploaded a speedscope profile of the run, trying to make sense of it myself. The command was ``` ./pants --no-pantsd dependencies :: &> /dev/null & py-spy record -p $last_pid...

For the record, the improvement is _dramatic_. Like we get 200-300% worse performance with `content_only` (the default) than we do with `never` (which I've now added in our default `pants.toml`...