for those not subscribed to the

pants-devel@

list, and using Python: https://groups.google.com/d/topic/pants-devel/b-28CClQSH4/discussion

Is there some summary of improvements or changes on the new backend?

Just followed up that email with a short summary

Unrelated: has anyone ever seen issues where the shader munges string literals?

It looks like it’s acting on the wrong entries in the constant pool

Something shady is going on there

pun intended

Ugh, there’s always something...

For posterity: This is related to the fact that the most recent version of the jar being shaded contains a single classfile (module-info.class) compiled with version 53 (i.e., java9) which ASM chokes on. But the other class files in the jar are version 52, which apparently causes jarjar to do something weird instead of just fail.

😟

We do check jarjar’s return value

so it’s somehow swallowing the error and continuing

Does anyone know what module-info.class in the package root is for and what causes it to be generated?

It’s something to do with some new Java modules spec

one that doesn’t appear to solve a real problem, as far as I can tell...

hah it's new to me I'm afraid

Hmm, it’s related to the jdk somehow

interesting

Still, I can exclude it from the shader with

Shader.exclude_package(’’)

I think

Yep, that works

phew

that's not particularly obvious

(to a lay observer)

random thought: why don’t we turn on

—fail-fast

in CI?

Does this look right?

ProductMapping(pex_archives) {
  PythonBinary(BuildFileAddress(BuildFile(docker/monitoring_sidecar/BUILD, FileSystemProjectTree(/Users/kyle.derr/Repositories/allclearid/odin/mono)), monitoring_sidecar)) => /Users/kyle.derr/Repositories/allclearid/odin/mono/.pants.d/binary/python-binary-create/0550b754d96b/docker.monitoring_sidecar.monitoring_sidecar/current
    [u'monitoring_sidecar.pex']
  PythonBinary(BuildFileAddress(BuildFile(docker/monitoring_sidecar/BUILD, FileSystemProjectTree(/Users/kyle.derr/Repositories/allclearid/odin/mono)), monitoring_sidecar)) => /Users/kyle.derr/Repositories/allclearid/odin/mono/.pants.d/binary/python-binary-create/0550b754d96b/tools.templating.configure/current
    [u'configure.pex']
}

It would appear that there are two Product entries for the same target. One of them is correct, but the other is a file produced in accordance with a different target. Am I confused about what dictates the target key in a ProductMapping?

(pants 1.2.1)

@happy-kitchen-89482 The shading problem you saw happens with Java 9 jars that have module-info.class at the top level of the jar. I saw the same problem with the errorprone jar and created https://github.com/pantsbuild/pants/issues/4288

Yeah, that’s exactly what I saw

Thanks for the reference

I think this is a bug in jarjar