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

Pants

test_python_eval.py seems to do exactly what you’re describing

Alright, I’ll take a look at that, thanks (to both suggestions)!

<@U04S45AHA>: afaik, the relative classpath stuff is in zinc analysis

most of the tests appear to be passing, but i did need one small bugfix in jmake

btw, <@U0545CGNV> there was a question raised about vex on the pants summit agenda doc (<https://docs.google.com/document/d/1tBVOaK2u1Q_-RmL2gG9qIoLpArHXiiEy-XcGRQHEElw/edit#>) - would you be a good person to answer that?

k - thats had a ton of refactoring, so may have existed in IvyResolve 3 years ago.  I'm pretty sure `.raw` was used with feeling and there was in fact a non-raw

Yeah - here's the history - LHS of this diff: <https://github.com/pantsbuild/pants/commit/458dfda61cb92c3a22ad7c09af9dbbad71071d72>

And that used this helper critically: <https://github.com/pantsbuild/pants/blob/458dfda61cb92c3a22ad7c09af9dbbad71071d72/src/python/twitter/pants/tasks/ivy_utils.py#L122>

So the raw classpath file was turned into a classpath file of symlinks and that classpath file of symlinks was cached

In that diff the RHS code still uses the util, but only after it cached the raw file

That looks like a bug way back then, though I'm sure it must have been fixed in between then and now and then re-introduced - that was Dec. 2013!

Basically - the logic is complex and I'm not even sure of muy analysis of the change given Benjy's green comment on the RHS, simplicity would come if the actual resolved jars were cached

that's where i'm doing the relativization

Thats a huge cache blob in lots of cases though!

need to split ivy_resolve from ivy_fetch. expect <@U053HNE90> would be down.

...where ivy_fetch is a new pants step - you can't actually resolve without fetching jars to the ivy cache using ivy IIUC

hm.. lame. thought it was `resolve` vs `retrieve`

...wait, i think you can? we generate reports

retrieve copies or symlinks from the cache to where you say

...but the report actually fetches... argh.

yeah - you'd need a mode where ivy would do a GET on metadata, HEAD on the artifacts