chore: update rust to 1.82

[Stebalien]

And explicitly disable features we don't support.

[rvagg]

I have a local version of this that I've rebased to current master and got the lints all sorted out but it still has a couple of test failures that are a bit concerning:

---- activate_deals_one_sector stdout ----
thread 'activate_deals_one_sector' panicked at runtime/src/test_utils.rs:1184:9:
assertion `left == right` failed: send to f07 expected method 3621052141, was 3261979605
  left: 3621052141
 right: 3261979605
note: run with `RUST_BACKTRACE=1` environment variable to display a backtrace

---- sectors_fail_and_succeed_independently_during_batch_activation stdout ----
thread 'sectors_fail_and_succeed_independently_during_batch_activation' panicked at runtime/src/test_utils.rs:1184:9:
assertion `left == right` failed: send to f07 expected method 3621052141, was 3261979605
  left: 3621052141
 right: 3261979605

Why would the FRC42 method numbers be changed with this update? Aren't they supposed to be a stable hash value?

[Stebalien]

They are supposed to be stable. I tested this because @raulk reported that 1.82 enables some wasm features we don't support in the FVM but... that shouldn't affect the tests here.

[Stebalien]

Can you force-push your updated version?

[rvagg]

done .. and done again to match the new master from yesterday

[Stebalien]

Ah, ok. It's not computing the wrong number, it's calling the wrong function.

We're expecting a call to Balance on f07, but we're getting a call to TransferFrom. And I think I found the bug.

[rvagg]

oooops .. is_none, my bad

[Stebalien]

We should probably keep the code CID and actually log it here.

[Stebalien]

I'd prefer to keep the error internals. I assume the complaint is that they're unused? In that case, we probably need to log them somewhere (will only affect actor debugging but it's still useful).

[rvagg]

@Stebalien I've undone the dead code removals I originally pushed, one was easy to log, the other two I don't see an obvious place to log the output without either writing a new test just for that or spending more time that I have right now to figure out which tests might touch them and logging the specifics in there. Can you see a quick path to doing that, or should we just force override the linter on them?

[Stebalien]

Clippy is a filthy liar. And it even tells us it's lying.

= note: `PrecompileError` has a derived impl for the trait `Debug`, but this is intentionally ignored during dead code analysis

We log here:

builtin-actors/actors/evm/src/interpreter/instructions/call.rs

Line 194 in 649f09a

log::warn!(target: "evm", "Precompile failed: error {:?}", err);

[Stebalien]

I've improved this a bit. I adds a tiny bit of code bloat, but it shouldn't be noticeable.

[rvagg]

k, all lgtm, but maybe we should wait until after v16-final before merging.