Fix last compare-mode false negatives in tests #136504
Conversation
- duplicates of crateresolve1 are used in a couple error-codes tests - also fix the note in crateresolve1 to link to these other duplicates, now that E0523 has been merged into E0464
There was a problem hiding this comment.
The test changes themselves look reasonable to me. However, I don't quite understand the underlying problem. These normalizations look very easy to forgor. Do we have an issue somewhere to track the compare-mode false-negative problems? That is, adding new tests look super easy to neglect doing these compare-mode normalizations, right?
|
r? jieyouxu |
|
The problem is described in my ted talk in #136328 (comment) and I don't have a satisfying fix in mind between compiletest/rustc/tests. Maybe you will have one? Though, it'd be cool not to be blocked on a complete compiletest fix in the meantime.
Absolutely, unfortunately. |
Ah right, I completely failed to make the connection. Opened a new issue #136510. Haven't looked deeply into this yet, but that doesn't need to block this PR that is an immediate relief. You can r=me after PR CI is green of course. |
|
(To be brutally honest: I have no idea what a compare-mode is, this is a specific corner of compiletest that I have not yet looked into lmao) |
|
@bors r=jieyouxu |
Rollup of 6 pull requests Successful merges: - rust-lang#136398 (add UnsafeCell direct access APIs) - rust-lang#136465 (Some `rustc_middle` cleanups) - rust-lang#136479 (std::fs: further simplify dirent64 handling) - rust-lang#136504 (Fix last compare-mode false negatives in tests) - rust-lang#136511 (Add `cast_signed` and `cast_unsigned` methods for `NonZero` types) - rust-lang#136518 (Add note about `FnPtr` trait being exposed as public bound) r? `@ghost` `@rustbot` modify labels: rollup
Rollup merge of rust-lang#136504 - lqd:more-compare-modes, r=jieyouxu Fix last compare-mode false negatives in tests This PR is a continuation of rust-lang#136310 and fixes the last remaining cases of false negatives when running tests under a compare-mode. With these normalizations, all the compare-mode failures in `next-solver` (and `polonius`) should now be real, actual differences in diagnostics.
This PR is a continuation of #136310 and fixes the last remaining cases of false negatives when running tests under a compare-mode.
With these normalizations, all the compare-mode failures in
next-solver(andpolonius) should now be real, actual differences in diagnostics.