subtree-push nightly-2024-04-08

[ytmimi]

trafficstars

Bumping the toolchain version as part of a git subtree push

current toolchain (nightly-2023-12-28):

• 1.77.0-nightly (89e2160c4 2023-12-27)

latest toolchain (nightly-2024-04-08):

• 1.79.0-nightly (9d5cdf75a 2024-04-07)

r? @calebcartwright

[ytmimi]

@calebcartwright I think we need to remove or disable the generic-simd feature. bytecount uses packed_simd, which is where the error is occurring on our nightly builds.

[calebcartwright]

@calebcartwright I think we need to remove or disable the generic-simd feature. bytecount uses packed_simd, which is where the error is occurring on our nightly builds.

I see you're working on getting that fixed upstream, happy to consider alternatives in parallel (e.g. adjusting feature specification on our end)

[ytmimi]

Yeah, I thought I'd take a stab at resolving the issue in https://github.com/llogiq/bytecount/pull/92. CI is passing so I'm guessing I ported things over from packed_simd to std::simd correctly. Depending on how urgently we want to get the sync in I think we have 3 options:

1. Wait for the fix to be released upstream
2. Temporarily vendor the changes locally so we can still offer the generic-simd feature
3. (Temporarily) drop the generic-simd feature.

Let me know if anything else comes to mind or if you'd prefer to move forward with one of these options.

[calebcartwright]

I don't think there's any particular urgency to push a sync through at the moment. It's been too long since the last one, but that was the case last week and that will remain the case next week.

As such I'd rather take a pause and seek a resolution to the root cause as opposed to trying to work around symptoms

[ytmimi]

Works for me. Hopefully the bytecount maintainers can help get the upstream PR reviewed soon🤞🏼

[ytmimi]

@calebcartwright bytecount pushed a new minor release with fixes for generic-simd 🎉

[ytmimi]

@calebcartwright is it fine to remove error-chain from our integration tests? The project is archived and it seems that the tests don't build anymore with the nightly version we're bumping to.

[calebcartwright]

@calebcartwright is it fine to remove error-chain from our integration tests? The project is archived and it seems that the tests don't build anymore with the nightly version we're bumping to.

If the project is archived then yes, though would you mind sharing some more details about you mean regarding the tests not building?

[ytmimi]

@calebcartwright is it fine to remove error-chain from our integration tests? The project is archived and it seems that the tests don't build anymore with the nightly version we're bumping to.

If the project is archived then yes, though would you mind sharing some more details about you mean regarding the tests not building?

If you take a look at integration.sh, we start by reformatting the crate: https://github.com/rust-lang/rustfmt/blob/c97996fab61fc09432abcf64dbe3ce2d84f4d679/ci/integration.sh#L21

And then we clone the repo, and build and run the test suite for that crate to make sure that the code still builds after formatting. It's when we try to build the tests that things fail: https://github.com/rust-lang/rustfmt/blob/c97996fab61fc09432abcf64dbe3ce2d84f4d679/ci/integration.sh#L107-L114

You can confirm the failure by cloning the repo and trying to run the tests with the nighty version that we're bumping to:

And yeah, the repo is archived

[ytmimi]

closing in favor of #6188