Closed
Description
I've found some kind of regression in error reporting. In short, the same code compiled by nightly-2019-11-25 yields different error messages on Linux (Ubuntu) and Windows. On Ubuntu:
error[E0283]: type annotations needed: cannot resolve `_: error_mismatch::Trait`
--> tests/fail.rs:4:13
|
4 | let _ = collect();
| ^^^^^^^
|
::: /home/runner/work/test-error-mismatch/test-error-mismatch/src/lib.rs:9:19
|
9 | pub fn collect<T: Trait>() -> T {
| ----- required by this bound in `error_mismatch::collect`
On Windows:
error[E0283]: type annotations needed: cannot resolve `_: error_mismatch::Trait`
--> tests\fail.rs:4:13
|
4 | let _ = collect();
| ^^^^^^^
Reproduction is here: https://github.com/Cerber-Ursi/test-error-mismatch/runs/325632093 (tests are failing intentionally - in real project this is handled by trybuild).