Reject static async closures in AST lowering

[compiler-errors]

Bug reported in https://rust-lang.zulipchat.com/#narrow/channel/122651-general/topic/static.20async.20closures/with/535095775.

r? fmease

[compiler-errors]

Rejecting this pre-expansion should be covered by #145604, but I guess we don't necessarily need to parse the combination at all.

[compiler-errors]

Let's see what the fallout is tho.

@bors2 try

[compiler-errors]

TODO: add a test for static gen || {} too.

[compiler-errors]

I'll bless the spans when bors is back

[GrigorenkoPV]

I think there is one test in rust-analyzer that this will break too

[fmease]

I think there is one test in rust-analyzer that this will break too

rust-analyzer doesn't depend on rustc like that, so no, that shouldn't happen.

[rust-bors]

☀️ Try build successful (CI)
Build commit: 0e1f5e2 (0e1f5e292c80a395d02835ac656d141e992caf89, parent: 8365fcb2b840c95eeb0bc377af8bd498fad22245)

[compiler-errors]

@craterbot check

[craterbot]

👌 Experiment pr-145605 created and queued.
🤖 Automatically detected try build 0e1f5e2
🔍 You can check out the queue and this experiment's details.

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[craterbot]

🚧 Experiment pr-145605 is now running

ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[rustbot]

This PR was rebased onto a different master commit. Here's a range-diff highlighting what actually changed.

Rebasing is a normal part of keeping PRs up to date, so no action is needed—this note is just to help reviewers.

[craterbot]

🎉 Experiment pr-145605 is completed!
📊 5 regressed and 10 fixed (685389 total)
📰 Open the summary report.

⚠️ If you notice any spurious failure please add them to the denylist!
ℹ️ Crater is a tool to run experiments across parts of the Rust ecosystem. Learn more

[fmease]

All spurious as far as I can tell.

[fmease]

TODO: add a test for static gen || {} too.

.

but I guess we don't necessarily need to parse the combination at all.

Do you mean rejecting static || {} (and async, gen) in the parser unless it's #[coroutine] static || {}? That'd be quite surprising and also unprecedented if I'm not mistaken (namely, looking through the attributes at parse time), tho I guess it's technically possible since built-in attributes can never be shadowed by user-defined attributes (we reject such cases with ambiguity).

---

Regarding the 'gen || {} doesn't parse <-> static gen || {} parses and "works"' situation, I guess we want to do that in a separate PR? Esp. if it needs further discussions with oli. E.g, we also parse async gen || {} but bug! during AST lowering. I don't know what the untold plans are for all of these features.

---

r=me with static gen closure test added unless you also want to address some of the other things I mentioned above in this PR.

[compiler-errors]

Do you mean rejecting static || {} (and async, gen) in the parser unless it's #[coroutine] static || {}?

No, i mean making $KW static || {} illegal, since we don't really have a meaning yet for the combination of the two modifiers on closures yet.

[fmease]

Ah, sounds good! Let's do that once the results for #145604 are back!

Edit: Results are back and they're clean! Feel free to proceed :)