1.28.3 Release Planning

[rami3l]

@rust-lang/rustup is welcomed to propose outstanding issues that would be urgent enough to address before cutting the upcoming patch release.
Other issues on the current milestone will be postponed to the milestone representing the next release.

Outstanding Issues

• illumos: rustup 1.28.{0,1,2} are miscompiled #4420
• Document the behavior change WRT auto-installing the default profile v.s. the currently default profile (Rustup 1.28.2 uses a different profile to install rust-toolchain.toml files #4337).
• ci(macos): run x64 workflows with Rosetta 2 #4428

Note

It could be a good opportunity to test out rust-lang/promote-release#84, but it would not be a dealbreaker for cutting this release.

[djc]

I'd like to go to 1.29.0 and drop curl and native-tls support. I have one more change I'd like to make before that.

Might be good to also write up a list of (notable) changes that we have on master right now to get a handle on what the release would look like.

[rami3l]

I'd like to go to 1.29.0 and drop curl and native-tls support.

@djc That sounds... risky since there are still remaining problems in #3790, and I would originally expect OpenSSL to last for at least a few more patches.

This doesn't mean we cannot do 1.29.0 here and now; it's just that it requires more attention into justifying/advertising the change and the possible consequences, and I expect that would take longer than a normal patch release... combined with #4420 I'm afraid that some people might never get a working 1.28 release and now they are forced to face 1.29 🙈

[djc]

@djc That sounds... risky since there are still remaining problems in #3790, and I would originally expect OpenSSL to last for at least a few more patches.

It's not obvious to me what problems/remaining risks you are referring to. As I understand it, the change to rustls by default has been pretty smooth and has not led to a lot of issues, even with the explicit warnings in 1.28.2. The one issue I'm aware of is #4325, which is arguably a pretty specific environment problem, and we now have better error handling for that via rustls-platform-verifier 0.6 and I'm working on a fix that includes the roots required for verifying static.rust-lang.org certificates as part of the rustup binary.

[rami3l]

It's not obvious to me what problems/remaining risks you are referring to. As I understand it, the change to rustls by default has been pretty smooth and has not led to a lot of issues, even with the explicit warnings in 1.28.2. The one issue I'm aware of is #4325, which is arguably a pretty specific environment problem, and we now have better error handling for that via rustls-platform-verifier 0.6 and I'm working on a fix that includes the roots required for verifying static.rust-lang.org certificates as part of the rustup binary.

@djc I'm also aware of #4233 and #4305 (IIRC what has landed was not really a fix but a mitigation?), although they are indeed not particularly easy to reproduce.

[rami3l]

Might be good to also write up a list of (notable) changes that we have on master right now to get a handle on what the release would look like.

@djc After another look, I agree that that would be the best way to decide whether we are going with a patch or a minor. I'll do that shortly :D