@witty-crayon-22786 Can I get a sanity check on this? https://github.com/pantsbuild/pants/pull/14352

yea, seems fine… although I think that maybe moving the awareness of the prefix into

JvmProcess

, and then converting into either a

Process

(with an empty prefix) or an

InteractiveProcess

(with a non-empty prefix) might avoid the

__

assumption?

hrm.

it would mean processes would need to be written with an awareness that they might not be in the same directory as their arguments… but i think that that would likely be good anyway (to the point where I’m a bit surprised that REAPI doesn’t support it natively)?

That seems like a big deal for now?

maybe. it’s just that changing a filename or adding a new filename in

JdkSetup

might silently break run without good testing.

If we’re including random shell scripts in places, it seems like it’d be hard to avoid that?

having said that, if

JvmProcess

did have support for the template variable, you could still forget to include the prefix on a relative argument and break a test silently

that’s really the trick. But testing the

run

infrastructure should be enough to make sure that everything is written properly

k.

ah right. I think I did that properly somewhere else, too