TST: avoid building indirect dependencies from the all extra via tool.uv.no-build-package#18949
Conversation
|
Thank you for your contribution to Astropy! 🌌 This checklist is meant to remind the package maintainers who will review this pull request of some common things to look for.
|
|
👋 Thank you for your draft pull request! Do you know that you can use |
558087b to
67a5d62
Compare
| "backcall", # can be removed once ipython<8.17 is dropped | ||
| "pickleshare", # can be removed once ipython<8.17 is dropped |
There was a problem hiding this comment.
Why can't we drop support for ipython<8.17 right now?
There was a problem hiding this comment.
Well, we can do that separately, I just don't think this is in scope for this PR (and I'm trying to bump as little dependencies as possible in this process)
There was a problem hiding this comment.
These constraints are workarounds, but bumping the dependencies would solve the problem.
There was a problem hiding this comment.
these constraints are temporary code that barely cost anything to maintain, and have no repercussions on users. Bumping optional dependencies is for ever and may incur a costs to users. I don't understand why temporary workarounds with clear expiration conditions should be avoided at all costs, or why we seem to be drawing diverging conclusions in face of what seems to me like a very imbalanced trade off.
| "psutil", | ||
| "pyzmq", | ||
| "tornado", |
There was a problem hiding this comment.
Why do some packages have comments documenting when they can be removed from this list and others do not?
There was a problem hiding this comment.
I include comments for the ones I was able to figure out an exit strategy for. The IPython/Jupyter ecosystem is a very intricate and deep dependency graph on its own, and while I did give it a try, I wasn't able to find the exact conditions when this constraints can be lifted.
…ool.uv.no-build-package`
67a5d62 to
c98d887
Compare
|
Overall, I don't see how disallowing non-binary for more stuff would hurt, as long as this does not just pull them in blindly (doesn't, right?). I can approve after CI passes. I want to see the Extra CI also, just in case.
|
pllim
left a comment
There was a problem hiding this comment.
Specific concerns like bumping minversion of IPython should be a follow-up issue/PR, should you decide to pursue it.
Thanks, all!
Closing the loop: it doesn't! |
Yup, the minimal deps job looked clean. :) |
Description
A first line of defense against unnecessary builds when trying to install with
--resolution=lowest.A bunch of these can be upstreamed later (and I've started the process already), but merging this will immediately help towards #18782