Do y’all know why we use a

tool.pex

,

requirements.pex

, and

runner.pex

for things like Pytest? I think the

runner.pex

can be consolidated into the

tool.pex

. Iirc, it was to avoid having to recreate the

runner.pex

when you do something like bump a tool’s version. But that doesn’t seem like we save much; all the

runner.pex

ever does is set up an entry-point and

--pex-path

, which is trivial to do directly on the

tool.pex

in the first place.

sounds right. but not sure what the contents of tool.pex would even be…

pytest.pex

vs.

pytest_runner.pex

.

mypy.pex

vs.

mypy_runner.pex

ah.

pytest_runner.pex

contains nothing. It’s simply the metadata for the entry point, +

--pex-path

There presumably was a reason

I think that might be what we were thinking, but it doesn’t seem to change unless the user for some reason sets a custom entry point I do know that we expected the tool Pex to be rebuilt a lot from version bumps. And that we didn’t want the runner Pex to have to change. But I think our miscalculation is that adding the metadata to the original tool Pex has a trivial cost