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

Pants

Thanks for the heads up on the VM exiting, I'll dive into that

Which leaves only the hung process problem!  Which is itself pretty bad

I'm mildly surprised this isn't showing up for other shops

I'm trying to get this to work: `export PYTHONFAULTHANDLER=true`

And these processes aren't responding to `kill -3`, only abort

what issues would folks anticipate if the templated workdirs in pants.ini became relative rather than absolute?

(in the case of `buildroot`, `pants_workdir`, etc)

Relative to what is the question that comes to my mind. Isn’t it important that we be able to change these for testing?  For tests we change buildroot and pants_workdir to temp directories.

Maybe we could so some kind of generic scrubbing of options values in the Task fingerprinting that would identify and strip `buildroot` and `pants_workdir` prefixes from values

templating only to un-template later feels awkward.

`pants_workdir`, `pants_supportdir`, `pants_distdir` are the only ones i really care about, and would become relative to the `buildroot`

the issue i'm seeing is that when those variables are used in options, they change the fingerprint unnecessarily between hosts/workspaces

<@U053HNE90>: there was an issue very early in the initial NailgunExecutor refactoring where nailgun startups would kill each other due to java version checkins + symlinks, but that was fixed; if you can capture a -ldebug output during a repro that would probably tell us a lot about whats happening. is the nailgun still running in your case?

hello -- following along with pants contributors guide here: <https://pantsbuild.github.io/howto_contribute.html>

i've got a shipit from benjy here: <https://rbcommons.com/s/twitter/r/2670/>

and assume I don't have write rights for this pull request: <https://rbcommons.com/s/twitter/r/2670/>

what's my best bet for getting help getting it merged?

<@U097BRCBV>: the diff on the pull request looks confused (which your comment indicates you saw), otherwise i would have merged already probably

the pull itself looks ok on github, the diff in the rb is borked

so maybe i should kill that and start fresh?

no need to kill it... just get a clean branch

`rbt` uses `git diff master $branch`, iirc.

<@U054MQUPL>: I'm running with a periodic thread dumper now to try to repro

I'm not sure if the nailguns are still running, my repro process is old and they might have been nuked