Ergonomic, hermetic, user-friendly software build system for Python, Go, Java, Kotlin, Protobuf, Scala, Shell. Pants lets you fearlessly scale up your codebase!

Pants

it's possible that a change afterward broke that.

ah, so pants list fails correctly, but test does not

looks like john changed it quite a bit over the weekend

<https://github.com/pantsbuild/pants/blob/master/build-support/bin/native/bootstrap.sh#L69> doesn't die... unclear why not

but it doesn't seem like that codepath is used, so

Codepath is used to generate linux binaries in CI: <https://github.com/pantsbuild/pants/blob/master/build-support/bin/native/generate-bintray-manifest.sh#L78-L79>

Working my real job today, but presumably just `|| die ...` instead of `|| echo ...` - likely refactor bug on my part

yea, sorry... shouldn't have pinged. <@U066LC126> : easiest for you just fix whatever isn't dieing currently probably.

possible that the exit is in a subshell, and so it doesn't kill the whole process...? unknown

<@U06A03HV1> if I have the tuple that’s used to construct a root node, how do I get the root node through to trace? Is zipping together the execution_request’s roots with the raw roots from self._native.lib.execution_roots(self._scheduler) the only way?

<@U066LC126> : imo, would remove the ability to trace individual nodes

can restore it later, but it's not clearly useful yet.

The scheduler has roots. Are those persistent across runs?

so I could just iterate over those to produce traces?

yes... it will be a behaviour change in a few places in the python code

but the problem is that the trace is overlaid

working on switching to producing one trace per root

converting a graph to a path when there’s multiple starting points is tricky

<@U066LC126> : see the linked tickets... if you're going to change the behaviour, would recommend one trace per unique _failure_