timeout: add all signal handlers, pass on signals to child, add ignored signal handling #10254

ChrisDryden · 2026-01-15T02:04:05Z

There was quite a bunch of different features missing in timeout, to start off, now that we have a mechanism to read the SIGPIPE handlers before they are overwritten by the rust runtime, it means that we can not propagate this signal down to the child processes if the signal is set to ignore.

This also includes all of the latest changes since 9.9 where the specific signal sent to timeout will be propagated instead of just defaulting to a TERM signal.

github-actions · 2026-01-15T02:25:28Z

GNU testsuite comparison:

Skip an intermittent issue tests/misc/tee (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/shuf/shuf-reservoir (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/sort/sort-stale-thread-mem (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/timeout/timeout (passes in this run but fails in the 'main' branch)
Congrats! The gnu test tests/timeout/timeout-group is now passing!

github-actions · 2026-01-15T03:25:00Z

GNU testsuite comparison:

Skip an intermittent issue tests/misc/tee (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/timeout/timeout (passes in this run but fails in the 'main' branch)
Congrats! The gnu test tests/timeout/timeout-group is now passing!

ChrisDryden · 2026-01-15T21:20:49Z

This might seem like a fairly large PR, but its actually just added ontop of this PR: #10166 and this PR: #10194, when those are in this would be much smaller

ChrisDryden · 2026-01-15T21:22:07Z

Another note is that we have a bunch of over-rides in the build-gnu script that override the yes and timeout command that we can remove once this is removed.

codspeed-hq · 2026-01-15T22:29:06Z

CodSpeed Performance Report

Merging this PR will not alter performance

_{Comparing ChrisDryden:timeout-fixes-squashed (6d8901c) with main (2839f78)}

Summary

✅ 142 untouched benchmarks
⏩ 180 skipped benchmarks¹

180 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

src/uu/yes/src/yes.rs

src/uu/timeout/src/timeout.rs

ChrisDryden · 2026-01-19T12:28:10Z

@Ecordonnier Thanks for the feedback! This PR is a bit too complicated, I was thinking that the best approach here would be to add that pipe signal logic in a smaller PR to every utility. That would resolve 5+ pr's in the queue, then rebasing this one after addressing your feedback and then it would be better to understand

Ecordonnier · 2026-01-19T13:00:05Z

@Ecordonnier Thanks for the feedback! This PR is a bit too complicated, I was thinking that the best approach here would be to add that pipe signal logic in a smaller PR to every utility. That would resolve 5+ pr's in the queue, then rebasing this one after addressing your feedback and then it would be better to understand

you can do it that way if it makes sense short-term, but then you will have the same code duplicated 100 times (one time for each utility), which is not great for long-term maintenance.

ChrisDryden · 2026-01-19T13:17:40Z

Whoops, what I meant to say is to add it to every utility by adding it once in the bin macro to not have to have the code duplicated.

ChrisDryden · 2026-01-21T17:15:54Z

I'm rebasing it now after the changes to add the pipeline signal handlers to all of the utilities, should be a much smaller diff now

github-actions · 2026-01-21T21:08:17Z

GNU testsuite comparison:

Skipping an intermittent issue tests/misc/tee (passes in this run but fails in the 'main' branch)
Skipping an intermittent issue tests/timeout/timeout (passes in this run but fails in the 'main' branch)
Congrats! The gnu test tests/timeout/timeout-group is now passing!

ChrisDryden · 2026-01-21T21:12:43Z

I think the flakiness with the tee test should be resolved with this PR now. The latest change to make the yes utility have the PIPE signal handler, caused the test to fail because of this line:

        Err(err) if err.kind() == io::ErrorKind::BrokenPipe => Ok(()),

Now we can remove that line, and it makes the tee test pass and I believe now with the timeout changes, it removes the source of the flakness

…atibility

github-actions · 2026-01-21T21:27:15Z

GNU testsuite comparison:

Skipping an intermittent issue tests/misc/tee (passes in this run but fails in the 'main' branch)
Skipping an intermittent issue tests/timeout/timeout (passes in this run but fails in the 'main' branch)
Congrats! The gnu test tests/timeout/timeout-group is now passing!

Ecordonnier · 2026-01-21T22:38:51Z

Do you still want #10166 to be merged? It overlaps with this PR.

ChrisDryden · 2026-01-21T22:47:02Z

This one covers everything, I can close that one

Ecordonnier · 2026-01-22T09:51:13Z

src/uu/yes/src/yes.rs

    match exec(&buffer) {
        Ok(()) => Ok(()),
+        // On Windows, silently handle broken pipe since there's no SIGPIPE
+        #[cfg(windows)]


let's move this to another PR, as it's unrelated to the rest. and please also remove this from the PR description "I also had to enable to pipe signal handlers to the yes utility ".

The timeout gnu tests validate that when timeout yes is run with sigpipe ignored that it exits with broken pipe, I kept it because that line needs to be removed for unix to pass the gnu integ tests but it's required for windows to pass the rust integ tests we've made for it.

Added this in another PR here: #10429

Which rust integ test do you mean?

Answering my own question: test_yes.rs calls check_termination(), this calls "run_stdout_starts_with" in util.rs, and this calls "child.close_stdout()" and triggers a BrokenPipe error on windows.

Hence several tests are triggering BrokenPipe error on windows: test_yes::test_args, test_yes::test_long_input,
test_yes::test_long_odd_output, test_yes::test_long_output, test_yes::test_simple

github-actions · 2026-01-26T12:41:34Z

GNU testsuite comparison:

Congrats! The gnu test tests/timeout/timeout-group is now passing!

ChrisDryden · 2026-01-26T12:43:19Z

Removed the yes.rs change since it will be handled in the other PR, the timeout-group passes but the timeout gnu test fails since it relies on that yes change. Its split into the other PR

Ecordonnier · 2026-01-26T16:43:59Z

@ChrisDryden I had restarted the one job which had failed 1 minute ago, but the merge of main cancelled the job and restarted all CI jobs ;-)

github-actions · 2026-01-26T16:52:04Z

GNU testsuite comparison:

Skipping an intermittent issue tests/misc/tee (passes in this run but fails in the 'main' branch)
Skipping an intermittent issue tests/timeout/timeout (passes in this run but fails in the 'main' branch)

ChrisDryden · 2026-01-26T16:54:24Z

So timeout-group passed in the GNU test but isn't showing in the comments?

Ecordonnier · 2026-01-26T19:37:27Z

So timeout-group passed in the GNU test but isn't showing in the comments?

At least in this PR the logs of the job "Run GNU tests (native)" clearly says "PASS: tests/timeout/timeout-group.sh" in the step "Run GNU Tests".
In main (e.g. https://github.com/uutils/coreutils/actions/runs/21367773234/job/61504073540 ) the log says "timeout-group.sh: skipped test: timeout returned 142. SIGALRM not handled?"

But you are right something seems buggy in the comparison script. The step "Aggregate GNU test results" -> "compare test failures vs reference" says this in your PR:

Total tests in current run: 646
Total tests in reference: 645
New regressions: 0
Intermittent regressions: 0
Fixed tests: 0
Intermittent fixes: 2
Newly skipped tests: 0
Newly passing tests (previously skipped): 0
Newly failing tests (previously skipped): 0
Intermittent newly failing: 0
Notice: No new test failures detected
...

So somehow this is detected as a new test, and not as a test which was previously skipped.

Ecordonnier · 2026-01-27T09:10:44Z

So timeout-group passed in the GNU test but isn't showing in the comments?

This is also what coreutils-tracking sees btw ( https://github.com/uutils/coreutils-tracking/blob/accf571e14c51cdcbd7f1a57521498506d061ed6/gnu-result.json#L20822C1-L20839C4 ):

  "Mon, 26 Jan 2026 17:52:50 +0000": {
    "sha": "6124866934900219a4fba24c2ff3e3dead0c654c",
    "total": "646",
    "pass": "615",
    "skip": "8",
    "fail": "22",
    "xpass": "0",
    "error": "1"
  },
  "Tue, 27 Jan 2026 01:24:25 +0000": {
    "sha": "1c09c368db1e50d77c199a6583210e1038ba8557",
    "total": "646",
    "pass": "619",
    "skip": "7",
    "fail": "19",
    "xpass": "0",
    "error": "1"
  }

ChrisDryden force-pushed the timeout-fixes-squashed branch from 93ea128 to 228824e Compare January 15, 2026 02:10

ChrisDryden requested a review from Ecordonnier January 15, 2026 02:37

Ecordonnier self-assigned this Jan 15, 2026

Ecordonnier reviewed Jan 17, 2026

View reviewed changes

src/uu/timeout/src/timeout.rs Show resolved Hide resolved

ChrisDryden force-pushed the timeout-fixes-squashed branch 2 times, most recently from fa027b0 to 293ded3 Compare January 21, 2026 20:56

timeout: fix signal handling, SIGPIPE preservation, and GNU test comp…

4421852

…atibility

ChrisDryden force-pushed the timeout-fixes-squashed branch from 293ded3 to 4421852 Compare January 21, 2026 21:16

Ecordonnier reviewed Jan 22, 2026

View reviewed changes

ChrisDryden mentioned this pull request Jan 22, 2026

yes: Silently handle broken pipe on windows #10429

Merged

Remove yes.rs changes (moved to separate PR uutils#10429)

69a6e61

Ecordonnier changed the title ~~timeout: adding all signal handlers, passing on signals to child, adding ignored signal handling~~ timeout: add all signal handlers, pass on signals to child, add ignored signal handling Jan 26, 2026

Merge branch 'main' into timeout-fixes-squashed

6d8901c

ChrisDryden mentioned this pull request Jan 26, 2026

GNU coreutils 9.9: Detailed Test report 01/20 #10390

Open

Ecordonnier merged commit 67046f5 into uutils:main Jan 26, 2026
130 of 131 checks passed

Ecordonnier mentioned this pull request Jan 27, 2026

build-gnu.sh: remove workaround for timeout/yes #10505

Merged

Uh oh!

timeout: add all signal handlers, pass on signals to child, add ignored signal handling #10254

timeout: add all signal handlers, pass on signals to child, add ignored signal handling #10254

Conversation

ChrisDryden commented Jan 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions bot commented Jan 15, 2026

Uh oh!

github-actions bot commented Jan 15, 2026

Uh oh!

ChrisDryden commented Jan 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ChrisDryden commented Jan 15, 2026

Uh oh!

codspeed-hq bot commented Jan 15, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CodSpeed Performance Report

Merging this PR will not alter performance

Summary

Footnotes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

ChrisDryden commented Jan 19, 2026

Uh oh!

Ecordonnier commented Jan 19, 2026

Uh oh!

ChrisDryden commented Jan 19, 2026

Uh oh!

ChrisDryden commented Jan 21, 2026

Uh oh!

github-actions bot commented Jan 21, 2026

Uh oh!

ChrisDryden commented Jan 21, 2026

Uh oh!

github-actions bot commented Jan 21, 2026

Uh oh!

Ecordonnier commented Jan 21, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

ChrisDryden commented Jan 21, 2026

Uh oh!

Ecordonnier Jan 22, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

ChrisDryden Jan 22, 2026

Choose a reason for hiding this comment

Uh oh!

ChrisDryden Jan 22, 2026

Choose a reason for hiding this comment

Uh oh!

Ecordonnier Jan 22, 2026

Choose a reason for hiding this comment

Uh oh!

Ecordonnier Jan 24, 2026

Choose a reason for hiding this comment

Uh oh!

github-actions bot commented Jan 26, 2026

Uh oh!

ChrisDryden commented Jan 26, 2026

Uh oh!

Ecordonnier commented Jan 26, 2026

Uh oh!

github-actions bot commented Jan 26, 2026

Uh oh!

ChrisDryden commented Jan 26, 2026

Uh oh!

Ecordonnier commented Jan 26, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Ecordonnier commented Jan 27, 2026

Uh oh!

ChrisDryden commented Jan 15, 2026 •

edited

Loading

ChrisDryden commented Jan 15, 2026 •

edited

Loading

codspeed-hq bot commented Jan 15, 2026 •

edited

Loading

Ecordonnier commented Jan 21, 2026 •

edited

Loading

Ecordonnier Jan 22, 2026 •

edited

Loading

Ecordonnier commented Jan 26, 2026 •

edited

Loading