Rust 1.18 regression, constrained-0.1.0, cannot infer type

[brson]

https://github.com/dylanede/constrained

101 brian@ip-10-145-43-250:~/dev/constrained⟫ git log -1
commit d3dfd78ebf46b8a5e03e8f1ef803506c07a2b15d
Merge: 1a03281 d288d01
Author: Dylan Ede <dylanede@googlemail.com>
Date:   Sun Apr 10 13:57:34 2016 +0100

    Merge branch 'master' of github.com:dylanede/constrained

brian@ip-10-145-43-250:/mnt2/dev⟫ rustc +beta -Vv
rustc 1.18.0-beta.1 (4dce67253 2017-04-25)
binary: rustc
commit-hash: 4dce67253d67785c74c8d270c38d774edd0deed8
commit-date: 2017-04-25
host: x86_64-unknown-linux-gnu
release: 1.18.0-beta.1
LLVM version: 3.9

brian@ip-10-145-43-250:~/dev/constrained⟫ cargo +beta test
   Compiling libc v0.2.4
   Compiling winapi-build v0.1.1
   Compiling rustc-serialize v0.3.16
   Compiling winapi v0.2.5
   Compiling advapi32-sys v0.1.2
   Compiling rand v0.3.13
   Compiling num v0.1.30
   Compiling dense_mats v0.4.0
   Compiling sprs v0.3.0
error[E0282]: type annotations needed
   --> /home/brian/.cargo/registry/src/gh.neting.cc-1ecc6299db9ec823/sprs-0.3.0/src/sparse/binop.rs:158:19
    |
158 |       let mut res = match rhs.ordering() {
    |  ___________________^
159 | |         StorageOrder::C => MatOwned::zeros(rhs.shape()),
160 | |         StorageOrder::F => MatOwned::zeros_f(rhs.shape()),
161 | |         _ => unreachable!(),
162 | |     };
    | |_____^ cannot infer type for `dense_mats::Tensor<_, [usize; 2], std::vec::Vec<N>>`

error: aborting due to previous error

error: Could not compile `sprs`.

To learn more, run the command again with --verbose.

cc @dylanede

[brson]

Oh the failure is actually in sprs-0.3.0 cc @vbarrielle.

[vbarrielle]

Hello,

I can fix this with a simple type annotation, but the trouble is, this comes from a fairly old version of sprs (0.3.0, but I'm developping over version 0.5.0 with several breaking changes). I cannot patch version 0.3.0 and submit a fixed version on crates.io, since there are already versions 0.3.x published.

On solution I can see is for constrained to depend upon version 0.3.3 or later, and I could fix the type inference in a version 0.3.4.

How does that sound?

[Mark-Simulacrum]

84be2df -- #41236 (second time)

[brson]

@vbarrielle let's see if it's a bug on our side before you work around it.

[nikomatsakis]

Possibly related to #41727

[nikomatsakis]

Good news, this seems to build on my PR #41716, which is a fix for #41716.

[brson]

Still repros on 1.18.0-beta.3. I'll bump beta to .4.

[nikomatsakis]

@brson this was backported in #42182, which landed 2 days ago -- not sure which beta that is though

[nikomatsakis]

This seems to work for me now. (I just pulled the latest beta and did a build.) I'm going to close this as a dup of #41716. @dylanede let me know if you disagree.