add macOS x86_64 support highlight to release notes #22782
Conversation
At the last dev meeting it was felt that dropping support *in* 2.30 was too aggressive. Disccusion ref: pantsbuild#22711
|
|
||
| Python lockfile metadata can now be stored in a sibling file to the lockfile, instead of as a prepended header in the lockfile itself. | ||
|
|
||
| Pants 2.30 is scheduled to be the last version to provide binaries for `x86_64` macOS. This flows from the [reduced support](https://github.com/actions/runner-images/issues/13046) from GitHub and continued movement in the macOS ecosystem. If you are interested in maintaining or supporting `x86_64` macOS binaries, please [reach out](https://github.com/pantsbuild/pants/discussions/22711). |
There was a problem hiding this comment.
Since discussing this new information has come to light, to wit:
- We now have a working self-hosted x86 runner for macos 13, so can continue to run CI on it.
- GHA will provide a hosted x86 runner for macos 15 until mid-2027.
So with that said I wonder if we should change this so that this is the last version to support macos 13 on x86, and punt on when to stop x86 support altogether? The reasons we considered (lack of GHA support) are no longer an issue. And since macos 15 is the last version that will support x86, its sunsetting will be a natural point for us to sunset our own support.
Thoughts?
There was a problem hiding this comment.
I'm not following what the new information is. The macos-15-intel option is cross-linked and discussed at the top of #22711
I run an idiosyncratic Linux distro on self-assembled hardware. I'm inclined towards sympathy if someone wanted to do Pants on RISC-V or NetBSD. I think Pants should run fine on a Raspberry Pi? It would be cool to hear about it.
For macOS x86_64 I think the bar can be really low, like any sponsorship level, or interest at all from a human in helping, but our calls out have not gotten that.
The ongoing non-compute maintenance costs for a volunteer project to support a target that represents are real. For example, here are the (macOS13-x86_64) failures on the first page of the GitHub history for main when I looked Friday:
- https://github.com/pantsbuild/pants/actions/runs/18734971616/job/53439949030
- https://github.com/pantsbuild/pants/actions/runs/18727405282/job/53415896518
- https://github.com/pantsbuild/pants/actions/runs/18697429970/job/53318405350
- https://github.com/pantsbuild/pants/actions/runs/18670364147/job/53231048779
- https://github.com/pantsbuild/pants/actions/runs/18670359960/job/53231975715
- https://github.com/pantsbuild/pants/actions/runs/18574102534/job/52958259814
- https://github.com/pantsbuild/pants/actions/runs/18567022910/job/52930800000
- https://github.com/pantsbuild/pants/actions/runs/18567011470/job/52930761378
- https://github.com/pantsbuild/pants/actions/runs/18252174847/job/51968171772
- https://github.com/pantsbuild/pants/actions/runs/18252116189/job/51968026378
- https://github.com/pantsbuild/pants/actions/runs/18252116189/job/51968026378
- https://github.com/pantsbuild/pants/actions/runs/18252085855/job/51967952732
- https://github.com/pantsbuild/pants/actions/runs/18234948746/job/51926797338
That's a lot of failures. Are there 13 bugs? 1 bug? Someone has to do the triage and fix the bugs or the flakiness eats away at the rest of the project. Are all the deps needed for #22790 going to continue to support free threaded and macOS x86_64 for the next two years?
The -- admittedly imperfect -- download stats for put macOS x86_64 at around 0.1%, and with no new hardware being manufactured it is only going to down from there.
There was a problem hiding this comment.
Ah, I missed the part of the discussion that already took the macos-15 runner into account. I was thinking that the primary concern was the lack of CI resources.
Not sure I buy the Raspberry PI comparison, but the point about non-compute project costs is well taken.
Ok, so let's merge this, and I'll get it in to 2.30.0a0.
|
@cburroughs OK to merge this? |
4 participants
At the last dev meeting it was felt that dropping support in 2.30 was too aggressive.
Disccusion ref: #22711