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

Pants

I hope pants is just being smart and trying to compile targets that don't depend on the failed one

yes - isolated partitions partionable things and so a parallel failure can happen

<@U053H6LQS>: I've got a bit of cleanup to do and more tests, but everything looks fine now, although your example needed
-  service_writer='com.squareup.wire.SimpleServiceWriter',
+  service_factory='com.squareup.wire.java.SimpleServiceFactory',

Oh, that changed from 2.0.0-BETA3 i guess.

<@U053HHEDG>: I have a change in pants coming up for adding distributions and platforms into the export format.

I keep running into little gocha’s around jvm_platform so I had to make some changes.

Drat, the developers of the world are waking up and slowing down my travis-ci builds!

<@U053HNE90>: did you start from `pants.ini.isolated` when you created your isolated config?

cool. the analysis cache sizing is one important performance knob

but we haven't seen any memory issues at all since switching

I did bump that up to the value in pants.ini

so... that would need to be a pretty insane target

But after giving it more heap, it seemed less likely to OOM and more likely to get really stuck and CPU bound

because we have a few absymal ones, but they still barely use 2G of heap

My suspicion is that it's a function of the length of the classpath, and that it's when it's attempting to generate the name hashing specific analysis

an excellent question! i don't know yet know which zinc/scalac use most heavily

The length of the classpath looked to be several hundred or so target, I can reproduce today if you have some suggestions on how to debug/report

but i do know that it does a ton of class lookups.