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

Pants

<@U06A03HV1>: I mean, we use whatever is built into publishing. We don't have any other references

yea, so the deprecation you were supposed to get there applies

Is there a way to do something like `os.path.dirname(__file__)` that will behave consistently between run.py and running a pex?  With run: `/home/csb/gvs/ops/2ops/.pants.d/python-setup/chroots/54755f3eeed3326c8c51aafcd165a3e1d0df2512/certifi/cacert.pem` with pex `/home/csb/gvs/ops/2ops/tools/cobbling-time/bin/cobbling-time.shoe-sizer.pex/certifi/cacert.pem` (The obvious right answer is to use `pkgutil.get_data` but I'm working with a 3rd party library that really really wants a file on disk)

What happens if you build the pex with `zip_safe=False`?

Well the relevant snippet is in a `python_library`  but the pex itself has several dependencies that are not zip_safe, which I think makes the pex itself not zip safe?

Hm, I'm not entirely sure if that's true--it might only copy those particular deps to the cache dir

But I can't say a library is `zip_safe=False`, so anyone making a python_binary who had a transitive dep on this library would need to know to set zip_safe, which does not feel great

`python_library` can set `zip_safe=False`

Did that change since 0.0.76?: ```PythonLibrary received unknown arguments: 
  zip_safe = False
```

Just have to make them transitively sticky

Thanks.  Will work around for now and open a ticket

<@U053HNE90>: do you guys use `JvmDependencyCheck` still?

<@U053HHEDG>: <@U06A03HV1>: we also hit that publishing bug - I reverted your commit internally. We also use those credentials to publish things that are not jars (we have pom-publish, which publishes a group of jars with a common version)

I haven't had a chance to dig at it yet, though

the message just says "use netrc_credentials"

Yeah - the other task consumes the username and password in a probably weird way - I just was past the time I had to look at it

<@U053H6LQS>: how would you feel about not creating a report if there is no associated resolve?

I am ok with that as long as we write to the console or something.

<@U051221NF>: Was it intentional that `list` option types now return `None` by default if their `default` isn’t explicitly set to `[]`?

having good default zero-values for options would be really nice, but it wasn't guaranteed even before that patch.  Pants has been bit elsewhere by this.