Reviewers: The first commit includes the very minor manual changes and can use human scrutiny. The second is the very large number of changes created by generate-lockfiles.

We didn't lose anything noteworthy in those 10k lines?

Also, did this come with any version updates on that generation?

We didn't lose anything noteworthy in those 10k lines?

No, the info in the files is equivalent (pex lockfiles have a few more lines of metadata, but that's not all that significant). It's more that the uv format crams more info per line.

Also, did this come with any version updates on that generation?

It did. I should probably mention those in the release notes.

Added to the release notes.

I think dogfooding with python-default makes a lot of sense as a next step.

I'm less sure about switching all of the bundled tools lockfiles. That seems like a broader change. Do we know for example if this is faster or slower when used?

I think dogfooding with python-default makes a lot of sense as a next step.

I'm less sure about switching all of the bundled tools lockfiles. That seems like a broader change. Do we know for example if this is faster or slower when used?

I don't think we're cherry picking this into 2.32, so I specifically think we SHOULD try this now and revert it later if things go off the rails.

Definitely not for cherry-picking.

Noting that this failed - spectacularly - on main due to issues users have already pointed out, but were seemingly ignored. The current treatment does not work with remote caching: #23344 (comment)

Yikes, that is bad (of me). Revert here: #23360

I did not remember that we use remote caching (only on intra-fork merges).
But in any case, this would harm users too so I should never have merged it.