Regression in asynchronous-0.4.5, Rust 1.17

[brson]

https://github.com/bcndanos/asynchronous

101 ec2-user@ip-10-145-56-122:~/triage/asynchronous$ git log -1
commit 5423c7e714e46117a75f8b89086b12f8e31da170
Author: bcndanos <[email protected]>
Date:   Fri May 22 19:38:07 2015 +0200

    Update Readme

ec2-user@ip-10-145-56-122:~/triage/aligner$ rustc +beta -Vv
rustc 1.17.0-beta.2 (b7c276653 2017-03-20)
binary: rustc
commit-hash: b7c27665307704a9b158fe242e88e83914029415
commit-date: 2017-03-20
host: x86_64-unknown-linux-gnu
release: 1.17.0-beta.2
LLVM version: 3.9

        error[E0282]: type annotations needed
 --> <anon>:3:1
  |
3 | asynchronous::Deferred::new(|| {
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^^ cannot infer type for `T`

error: aborting due to previous error(s)

thread 'rustc' panicked at 'Box<Any>', /checkout/src/librustc/session/mod.rs:203

---- src/lib.rs - Deferred<T, E>::then (line 175) stdout ----
        error[E0282]: type annotations needed
 --> <anon>:8:4
  |
8 |    Ok(34)
  |    ^^ cannot infer type for `E`

error: aborting due to previous error(s)

thread 'rustc' panicked at 'Box<Any>', /checkout/src/librustc/session/mod.rs:203

---- src/lib.rs - Promise<T, E>::chain (line 628) stdout ----
        error[E0282]: type annotations needed
  --> <anon>:12:4
   |
12 |    Ok(34)
   |    ^^ cannot infer type for `E`

error: aborting due to previous error(s)

thread 'rustc' panicked at 'Box<Any>', /checkout/src/librustc/session/mod.rs:203

---- src/lib.rs - Promise<T, E>::fail (line 685) stdout ----
        error[E0282]: type annotations needed
 --> <anon>:3:1
  |
3 | asynchronous::Promise::new(|| {
  | ^^^^^^^^^^^^^^^^^^^^^^^^^^ cannot infer type for `T`

error: aborting due to previous error(s)

thread 'rustc' panicked at 'Box<Any>', /checkout/src/librustc/session/mod.rs:203

---- src/lib.rs - Promise<T, E>::then (line 598) stdout ----
        error[E0282]: type annotations needed
 --> <anon>:8:4
  |
8 |    Ok(34)
  |    ^^ cannot infer type for `E`

error: aborting due to previous error(s)

thread 'rustc' panicked at 'Box<Any>', /checkout/src/librustc/session/mod.rs:203


failures:
    src/lib.rs - Deferred<T, E>::chain (line 214)
    src/lib.rs - Deferred<T, E>::fail (line 289)
    src/lib.rs - Deferred<T, E>::then (line 175)
    src/lib.rs - Promise<T, E>::chain (line 628)
    src/lib.rs - Promise<T, E>::fail (line 685)
    src/lib.rs - Promise<T, E>::then (line 598)

test result: FAILED. 22 passed; 6 failed; 0 ignored; 0 measured

error: test failed

Not on nightly.

cc @bcndanos

[alexcrichton]

Looks like it also affects dotfiler

cc @matthunz

[brson]

Also affects error-chain cc @Yamakaky

[alexcrichton]

Like also affects error_defs

cc @canndrew

[brson]

Also fiber cc @TileHalo

[brson]

Also gimli cc @fitzgen

[alexcrichton]

Also futures 0.1.0

[alexcrichton]

Also hematite_server

cc @bvssvni

[alexcrichton]

Also libimagstore

cc @matthiasbeyer

[fitzgen]

@brson:

Also gimli cc @fitzgen

I pinged @aturon about this a few weeks ago. Then I landed a commit to work around the regression within gimli and a new version published to crates.io. Are you saying that the most recent version on crates.io is still affected (or that perhaps this is a different-but-similar regression)? Are the crates listed here the results of a crater (or its new replacement whose name I forgot)?

[alexcrichton]

@fitzgen these results are from cargobomb (a crater equivalent-ish) and the full results are posted online with the two regression logs for gimli, downloaded from crates.io.

Do you remember why this fix was needed? It'd be good to track down why inference started failing to see if we want to revert that change.

[Yamakaky]

I modified the broken test from error-chain so that it doesn't fail anymore, BTW. But thanks for the ping.

[alexcrichton]

@Yamakaky can you point to the fix? It's not clear to me at least currently why this regression happened and affected so many crates.

[nikomatsakis]

This may be fixed by #40636 -- which has yet to be backported to beta.

[fitzgen]

Here is the gimli fix I had made: gimli-rs/gimli@dbec032

[nikomatsakis]

This does work on nightly (I just checked) -- nightly includes my PR that changes how we handle dead-code type-checking (which subsumes and improves upon #40636). I haven't gotten a beta build that includes #40636 yet to test, but I suspect it will be fixed by that.

[Yamakaky]

@alexcrichton rust-lang-deprecated/error-chain@3fab94a

[alexcrichton]

Thanks for looking into this Niko! Thanks also for the links to fixes, let's land the backport and go from there

[nikomatsakis]

OK, so, the backport landed. @alexcrichton, when do you think we'll spin off a new beta so we can most easily test if this is fixed? (That is still a manual process, right?)

[alexcrichton]

@nikomatsakis we'll try to get a new beta scheduled today

[alexcrichton]

New beta is here and bug appears fixed! Closing.