Oh, hm. Didn't realize this implication till hooki...
# development
h
Oh, hm. Didn't realize this implication till hooking up Pex's lockfile support to Pants. It requires that you have the targeted interpreters available on your machine to generate a lockfile. Whereas with Poetry I can generate for Python 2.7 and 3.5-3.7 on my M1, I can't do that with Pex anymore. I believe it's a non-starter to change that given the design, right, @enough-analyst-54434? It seems like a reasonable limitation. And not everyone is generating lockfiles everyday, so scope is limited. But it does mean we probably need to deprecate changing from Poetry -> Pex because this is breaking I didn't do my homework well enough. Known that https://github.com/pantsbuild/pex/issues/1402 is a WIP
e
You shouldn't be hooking up Pex lockfiles at all since its not finished, and its orthoganal to multiple lockfile support IIUC - right?
The lock technology is irrelevant to supporting 2 or more of them.
h
Ah, miscommunication and I should have gone directly to the source (you + project board) rather than coordinating with Stu. I thought it was ready enough, mostly polish left with https://github.com/pantsbuild/pex/issues/1404 and https://github.com/pantsbuild/pex/issues/1405 Poetry is broken for
[python-repos]
and also leaking its cache, so there's motivation to switch But I see now that I missed https://github.com/pantsbuild/pex/issues/1402 is still WIP. My bad
and its orthoganal to multiple lockfile support IIUC - right?
Mostly, but
--platforms
support for
pex_binary
etc requires Pex, Still, I can make progress on other things in the meantime
e
Yeah, --platforms support is orthogonal. There are two things - fixing broken lock technology, which is the platforms bucket, and supporting multiple locks. It would probably have been better to not ship broken locks.