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.