On 11/08/2024 01:17, Blair Noctis wrote: > On 10/08/2024 18:04, Paul Gevers wrote: (...) >> https://ci.debian.net/packages/r/rust-just/testing/s390x/50327770/ >> >> 149s failures: >> 149s >> 149s ---- choose::status_error stdout ---- >> 149s thread 'choose::status_error' panicked at tests/choose.rs:209:3: >> 149s assertion failed: Regex::new("^error: Chooser `exit-2` failed: exit >> (code| >> status): 2\n$").unwrap().is_match(str::from_utf8(&output.stderr).unwrap()) > (...) > > Hi Paul, > > Thanks for bringing attention to it. I ran the tests a few times on zelenka > and > on debci, which all passed. Since it's kind of a test-Heisenbug with no report > in real usage, I'll mark the tests ignored as a temporary workaround.
This particular error didn't give much information other than stderr didn't match what's expected, but two others said "Broken pipe". A few searches revealed it's because Rust ignores SIGPIPE [1] by default and the println! macro just panics when an error occurred [2]. Setting SIGPIPE handler back to SIG_DFL works in a simple repro. I'll bring the severity down to important after uploading the fix, please revert if the problem happens again. On 11/08/2024 03:21, Paul Gevers wrote: > On 10-08-2024 19:17, Blair Noctis wrote: >> On the infra side, I'm pondering the idea of a "debug" patch that captures >> more >> information the next time it goes south. However, "fixing" (or rather >> frankly, >> ignoring) the test failures makes following runs pass. Is there a way to >> monitor >> such an output in passed runs, other than downloading every run and grepping >> through them? > > Not really. > > If you think that might be helpful, I can give you access to a testbed on the > host that runs our tests. Guess this is no longer needed ;) still, thanks for offering to help. [1]: https://github.com/rust-lang/rust/issues/46016 [2]: https://github.com/BurntSushi/advent-of-code/issues/17 -- Sdrager, Blair Noctis
OpenPGP_signature.asc
Description: OpenPGP digital signature
