John Sirois

@danxmoran I don't believe you / hrm. Can you get a sandbox on that version, even if just locally, and confirm the wrapper script has the `rm -rf ...` removed?

Alrighty - thanks. I am totally out of ideas then. That `rm -rf ...` truly was a race that definitely should cause errors like you saw. With that gone, the...

@danxmoran I'm stepping away from work until the 28th; so I'll un-assign for now and pick back up then if no one else has dug in.

@stuhood - check me on this. Here is the rule that creates the `__pants_df_parser.py` PEX: https://github.com/pantsbuild/pants/blob/7251de9bb8aef692b4ffcfc62c36943983a51145/src/python/pants/backend/docker/subsystems/dockerfile_parser.py#L73-L95 After f06bf22be that @rule has 0 changes to its argument values and 0 changes...

In short, I'm not giving up on the `rm -rf` explanation yet because it actually does explain this perfectly.

Right, rules aren't cached, only Processes are. That's an eternally confusing facet of this system.

Yeah, Eric's suggestion need not be followed. My late realization (for the umpteenth time) of only processes being cached, not @rules, makes that avenue make no sense. Apparently Eric also...

Just following up to @engnatha's comment. His issue was #17221 which was almost certainly fixed in #17177. The issue here in the PEX_ROOT in Pants' named_caches directory is still a...

Ok, back at this. I still can spot no bug or wrong assumption in the relevant Pex code, but I've got https://github.com/pantsbuild/pex/pull/1961 spun up to hopefully make errors more clear...

Thanks @danxmoran. When you get a chance you can try the following in `pants.toml` to pull in the release here: https://github.com/pantsbuild/pex/releases/tag/v2.1.112: ```toml [pex-cli] version = "v2.1.112" known_versions = [ "v2.1.112|macos_arm64|6b0be9f1dcf4d105aec06fd87d405ba46b45c266ef8549727b86541eea7c2f57|4065703",...