Minor unsoundness fix with FnDef lifetimes

[addiesh]

This is a minor breaking(?) change that tries to fix some issues brought up in #70917 and #84366 . As far as I can tell, the behavior for closures is still unchanged, which I believe is another part of the unsoundness. I don't have enough experience working with the compiler yet to fix it though.

r? oli-obk

[workingjubilee]

Is it breaking for code that is not unsound?

[addiesh]

Is it breaking for code that is not unsound?

I'm not sure. Probably not? But worth doing a crater run to make sure.

[oli-obk]

@bors try

[rust-bors]

☀️ Try build successful (CI)
Build commit: e622d8d (e622d8d7bed4f2668d446e06c6c1436ecae15796, parent: b52edc25bfbaa955b4b83c10f998e5224c3478b2)

[oli-obk]

@craterbot check

[craterbot]

👌 Experiment pr-156776 created and queued.
🤖 Automatically detected try build e622d8d
🔍 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

[addiesh]

This is quickly becoming a major unsoundness fix. Issues I have my eyes on:

• 'static closures/FnDefs/futures with non-'static return type are unsound #84366
• Rust allows impl Fn(T<'a>) -> T<'b> to be : 'static, which is unsound #112905
• Unsoundness due to 'static coroutines that yield non-'static values. #144442
• closures are not required to outlive generic arguments mentioned in their body #153140
• and (orthogonally) Tracking Issue for early/late bound parameter cleanup and late-bound turbofishing #156581

[rust-log-analyzer]

The job x86_64-gnu-gcc failed! Check out the build log: (web) (plain enhanced) (plain)

Click to see the possible cause of the failure (guessed by this bot)

[RUSTC-TIMING] gix_transport test:false 11.960
error[E0309]: the parameter type `V` may not live long enough
  --> src/tools/cargo/src/cargo/util/local_poll_adapter.rs:60:13
   |
18 |   impl<'a, S, K, V, E> LocalPollAdapter<'a, S, K, Result<V, E>>
   |        -- the parameter type `V` must be valid for the lifetime `'a` as defined here...
...
60 | /             async move {
61 | |                 let v = f(self_parameter.deref(), &key).await;
62 | |                 (key, v)
63 | |             }
64 | |             .boxed_local()
   | |__________________________^ ...so that the type `V` will meet its required lifetime bounds
   |
help: consider adding an explicit lifetime bound
   |
22 |     V: Clone + 'a,
   |              ++++

error[E0309]: the parameter type `E` may not live long enough
  --> src/tools/cargo/src/cargo/util/local_poll_adapter.rs:60:13
   |
18 |   impl<'a, S, K, V, E> LocalPollAdapter<'a, S, K, Result<V, E>>
   |        -- the parameter type `E` must be valid for the lifetime `'a` as defined here...
...
60 | /             async move {
61 | |                 let v = f(self_parameter.deref(), &key).await;
62 | |                 (key, v)
63 | |             }
64 | |             .boxed_local()
   | |__________________________^ ...so that the type `E` will meet its required lifetime bounds
   |
help: consider adding an explicit lifetime bound
   |
18 | impl<'a, S, K, V, E: 'a> LocalPollAdapter<'a, S, K, Result<V, E>>
   |                    ++++

For more information about this error, try `rustc --explain E0309`.
[RUSTC-TIMING] cargo test:false 28.564
error: could not compile `cargo` (lib) due to 2 previous errors

For more information how to resolve CI failures of this job, visit this link.

[workingjubilee]

"Probably not" is
the chance that sound code breaks
Yet gcc fails 👀

[addiesh]

"Probably not" is

the chance that sound code breaks

Yet gcc fails 👀

oop. I'm new to this stuff 😅. The first error was forgetting to run x test and updating a known-bug to a regression test, the second one is a bit confusing though. I'll have to keep working before any of these changes get merged.

[craterbot]

🚧 Experiment pr-156776 is now running

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

[craterbot]

🎉 Experiment pr-156776 is completed!
📊 402 regressed and 4 fixed (952039 total)
📊 5319 spurious results on the retry-regressed-list.txt, consider a retry¹ if this is a significant amount.
📰 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

Footnotes

1. re-run the experiment with crates=https://crater-reports.s3.amazonaws.com/pr-156776/retry-regressed-list.txt ↩

[addiesh]

• The proptest failures are for versions that are 6 or more years old. Probably won't fix.
• arch-pkg-text has 3 dependents, should be easy to fix, too, but unsure if that breaks more ppl because it's captured in a returned iterator
• The id3 failures are recent, but have an easy fix. I propose we fix the latest version and wait a bit and then land this PR.
  • fixed: https://codeberg.org/polyfloyd/rust-id3/pulls/162
• bevy_channel_trigger only has one failure, but these kind of crates are only used in private projects mostly. So we should fix it and probably just wait for the next bevy release, when everyone updates anyway
  • fixed: Fix from a rustc breaking change rustunit/bevy_channel_trigger#3
• wagon-lexer is only used by the author themselves, also looks straight forward to fix

[workingjubilee]

bevy updated on the 19th, and the new version of bevy-channel-trigger containing both the fix and the bevy 0.19 update is now on crates.io https://crates.io/crates/bevy-channel-trigger

[rust-bors]

☔ The latest upstream changes (presumably #159246) made this pull request unmergeable. Please resolve the merge conflicts by rebasing.