x86_64-sun-solaris target was removed without warning in Rust 1.52

[jhpratt]

#82216 removed the alias of x86_64-sun-solaris to x86_64-pc-solaris, but as far as I can tell did not remove the target itself.

After inquiring on Zulip, it was pointed out that the sun-solaris target was demoted to tier three, meaning builds are no longer provided. It appears this was done with no discussion, warning, or even a mention in the release notes.

Given that the pc-solaris target hasn't been around for that long (I think? feel free to tell me I'm wrong), I think this change should be reverted. My CI unexpectedly failed after being unable to download the target, and the only solution that wouldn't involve a notable increase in complexity of the CI script is to outright remove the checks for Solaris/Illumos — a let loss for everyone.

cc @Mark-Simulacrum

[Mark-Simulacrum]

I think my feeling right now is that we should likely re-enable the sun-solaris target alongside the pc-solaris, I don't think that should be a major problem for CI times or similar.

I will see if we can do this in time for 1.52.1, but if it ends up being harder than expected, I think it's likely that either a) we'll push it out to a 1.52.2 release later in the cycle, or simply leave it to 1.53. For the time being, tagging as 1.53.

[jhpratt]

I will note that the change itself is fine, of course, it's the suddenness and unexpectedness of it that caught me by surprise.

[Mark-Simulacrum]

I think it is likely true that we'll eventually drop the sun target, but it does seem like we should communicate that change better at least. Certainly the newly merged target tier policy suggests just that:

A tier 2 target may be demoted or removed if it no longer meets these requirements. Any proposal for demotion or removal will be CCed to the target maintainers, and will be communicated widely to the Rust community before being dropped from a stable release. (The amount of time between such communication and the next stable release may depend on the nature and severity of the failed requirement, the timing of its discovery, whether the target has been part of a stable release yet, and whether the demotion or removal can be a planned and scheduled action.)

[Mark-Simulacrum]

It is not clear to me what the impact of this change was. In particular, the produced target was not changed in #82216, but I guess the target definition is sort of https://github.com/rust-lang/rust/blob/1.52.0/compiler/rustc_target/src/spec/x86_64_unknown_illumos.rs#L12 for pc...

I think I am not sufficiently confident in moving in a particular direction here - and a full revert does not seem quite right either - so I will likely say this should be left for at least 1.52.2 or 1.53.0.

cc @nagisa as well, who reviewed #82216

[kulikjak]

Removal of the older target was certainly not my intent; if that happened due to #82216, I am sorry.

Both x86_64_sun_solaris.rs and x86_64_pc_solaris.rs are there and both targets are usable when compiling with rustc, but I guess that change to one of src/ci/docker/... files might have resulted in artifacts necessary for rustup no longer being available.

[kulikjak]

I created #85252, which I believe should bring back the ability to rustup the old target.

[Mark-Simulacrum]

Reopening to track the illumos target work. I'm not sure it's a release blocker, but would like to not lose sight.

[Mark-Simulacrum]

The remaining bug in this issue -- the change in the compiler used to build illumos to be based on a -pc- triple, not -sun- here https://github.com/rust-lang/rust/blob/master/src/ci/docker/scripts/illumos-toolchain.sh#L21 is not a release blocker.

[jhpratt]

Realistically I don't think there's anything actionable here, so I'm going to go ahead and close this.