ICE: trimmed_def_paths called, diagnostics were expected but none were emitted
#135289
Comments
djc
added
C-bug
rustbot
added
the
needs-triage
|
What are the steps to reproduce? cargo check on quinn? |
|
CI ran |
|
Thanks you saved me a click 👍.
This is still happening on nightly with Bisection points at #132658 but it's unclear whether that's actually correct yet -- I'll cc the author @estebank for due diligence -- but such issues are sometimes easy so I'll see if I have the time to look into it soon. |
lqd
added
A-diagnostics
|
Some context for esteban:
|
|
I think that we can side-step the issue quickly by removing the call to |
|
There’s no rush, right? |
|
I couldn't build a minimal repro to add to the test suite, but #135310 solves the ICE. |
|
The ICE occurs due to some incorrect behavior in reachability checking of the pattern at |
Always force non-trimming of path in `unreachable_patterns` lint Creating a "trimmed DefID path" when no error is being emitted is an ICE (on purpose). If we create a trimmed path for a lint that is then silenced before being emitted causes a known ICE. This side-steps the issue by always using `with_no_trimmed_path!`. This was verified to fix https://github.com/quinn-rs/quinn/, but couldn't write a repro case for the test suite. Fix rust-lang#135289.
Always force non-trimming of path in `unreachable_patterns` lint Creating a "trimmed DefID path" when no error is being emitted is an ICE (on purpose). If we create a trimmed path for a lint that is then silenced before being emitted causes a known ICE. This side-steps the issue by always using `with_no_trimmed_path!`. This was verified to fix https://github.com/quinn-rs/quinn/, but couldn't write a repro case for the test suite. Fix rust-lang#135289.
Rollup merge of rust-lang#135310 - estebank:issue-135289, r=Nadrieril Always force non-trimming of path in `unreachable_patterns` lint Creating a "trimmed DefID path" when no error is being emitted is an ICE (on purpose). If we create a trimmed path for a lint that is then silenced before being emitted causes a known ICE. This side-steps the issue by always using `with_no_trimmed_path!`. This was verified to fix https://github.com/quinn-rs/quinn/, but couldn't write a repro case for the test suite. Fix rust-lang#135289.
|
Reopening to track beta backport |
|
Let's keep it open even after backport for |
estebank
added
the
E-needs-test
lqd
added
the
E-needs-mcve
|
@djc btw, in the piece of code where I found the ICE and linked earlier: the |
|
This ICE resisted minimization a lot. Obviously any error is a no no, but it was the same for many warnings as well: any unused variable, useless mut, unused import, etc and the ICE disappears like a frightened tortoise 🐢. Once I discovered that, things became less unpleasant.
#[macro_export]
macro_rules! assert_matches {
( $e:expr , $($pat:pat)|+ ) => {
match $e {
$($pat)|+ => (),
_ => (),
}
};
}
const _A: u64 = 0;
pub fn f() -> u64 {
0
}
fn main() {
fake_matches::assert_matches!(f(), _non_existent);
}I've opened #135833 to add this as a test for now, but the fix should now be easier to clean up. |
lqd
added
S-has-mcve
Add fixme and test for issue rust-lang#135289 This PR: - adds a test minimizing issue rust-lang#135289 for PR rust-lang#135310 - adds a fixme about the suboptimal fix for the ICE I've verified the test indeed ICEs with 3f2f695 reverted. r? `@estebank`
…iaskrgr Rollup of 8 pull requests Successful merges: - rust-lang#132232 (CI: build FreeBSD artifacts on FreeBSD 13.4) - rust-lang#135706 (Move `supertrait_def_ids` into the elaborate module like all other fns) - rust-lang#135750 (Add an example of using `carrying_mul_add` to write wider multiplication) - rust-lang#135793 (Ignore `mermaid.min.js`) - rust-lang#135810 (Add Kobzol on vacation) - rust-lang#135821 (fix OsString::from_encoded_bytes_unchecked description) - rust-lang#135824 (tests: delete `cat-and-grep-sanity-check`) - rust-lang#135833 (Add fixme and test for issue rust-lang#135289) Failed merges: - rust-lang#135816 (Use `structurally_normalize` instead of manual `normalizes-to` goals in alias relate errors) r? `@ghost` `@rustbot` modify labels: rollup
Rollup merge of rust-lang#135833 - lqd:add-ice-test, r=compiler-errors Add fixme and test for issue rust-lang#135289 This PR: - adds a test minimizing issue rust-lang#135289 for PR rust-lang#135310 - adds a fixme about the suboptimal fix for the ICE I've verified the test indeed ICEs with 3f2f695 reverted. r? `@estebank`
Creating a "trimmed DefID path" when no error is being emitted is an ICE (on purpose). If we create a trimmed path for a lint that is then silenced before being emitted causes a known ICE. This side-steps the issue by always using `with_no_trimmed_path!`. This was verified to fix https://github.com/quinn-rs/quinn/, but couldn't write a repro case for the test suite. Fix rust-lang#135289. (cherry picked from commit 93a1950)
|
Closing since #135310 was merged |
beta triggered an ICE in Quinn's CI today:
https://github.com/quinn-rs/quinn/actions/runs/12691355015/job/35374229601?pr=2130
Maybe similar/related to:
Maybe a fix that could be backported?
The text was updated successfully, but these errors were encountered: