timeout: Replace busy loop with select(2)

[Alonely0]

Fixes issue #9099, where uu_timeout would have a static 100ms latency. This issue implements a self-pipe trick in order to run transparently on as many platforms as possible. There was a previous attempt of mine doing so with pselect(), but this failed due to macOS shenanigans and the lack of support from Rust's std for spawning with custom signal masks (which was ACP-accepted two or three years ago, but even patching it I couldn't get it to work properly…).

[github-actions]

GNU testsuite comparison:

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

[codspeed-hq]

CodSpeed Performance Report

Merging #9227 will not alter performance

_{Comparing Alonely0:timeout (a74ca55) with main (2da2c90)}

Summary

✅ 139 untouched
⏩ 37 skipped¹

Footnotes

1. 37 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. ↩

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/timeout/timeout (fails in this run but passes in the 'main' branch)
Skipping an intermittent issue tests/tail/overlay-headers (passes in this run but fails in the 'main' branch)

[Alonely0]

I think I have most likely addressed the last macOS non-POSIX conformance bug, at last. Elsewhere, I will just ignore the only test that spuriously fails exclusively on obsolete macOS x86_64, and address it in the future; it is just a spurious failure on retrieving a child's exit code, because of a POSIX noncompliance bug on outdated macOS. This is most likely only an issue on how Rust's std handles the exit code retrieval, and not on real-world scenarios, so this is absolutely a non-issue compared to what this PR is fixing, sigh. I'll fix the build issue on Redox after macOS CI decides it's done bugging me and other maintainers.

[Alonely0]

@cakebaker Do you have any suggestion/objection? I would like to get this issue fixed right away. I'll add some comments and rebase to latest main, then re-check CI because it was modified on main. Afterwards, unless CI says otherwise, this can be cleanly merged and close #9099.

[github-actions]

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/timeout/timeout (fails in this run but passes in the 'main' branch)

[github-actions]

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/timeout/timeout (fails in this run but passes in the 'main' branch)

[github-actions]

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/tail/overlay-headers (fails in this run but passes in the 'main' branch)
Skip an intermittent issue tests/timeout/timeout (fails in this run but passes in the 'main' branch)

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/timeout/timeout (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/tail/follow-name is no longer failing!

[github-actions]

GNU testsuite comparison:

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

[github-actions]

GNU testsuite comparison:

Congrats! The gnu test tests/basenc/bounded-memory is now passing!
Note: The gnu test tests/dd/no-allocate was skipped on 'main' but is now failing.

[ChrisDryden]

I don't know of a way that you can read the pipe signals to be able to handle the ignore signals and flags without it being overriden by the rust runtime without reading them in the .data_init section. Theres an implementation of it here: #9657

I think it would be much simpler to take this approach than whats in this PR.

[Alonely0]

I don't know of a way that you can read the pipe signals to be able to handle the ignore signals and flags without it being overriden by the rust runtime without reading them in the .data_init section. Theres an implementation of it here: #9657

I think it would be much simpler to take this approach than whats in this PR.

Oh, I didn't know about this. The thing is, timeout needs to scan all the signal space and block ignored ones to conform to GNU (something this PR is introducing), so I guess the best option is to just code in a special case for SIGPIPE that uses that approach. However, this PR has become quite expansive because it's fixing multiple bugs and adapting to new GNU behavior, so I guess it sticks.

[github-actions]

GNU testsuite comparison:

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

[github-actions]

GNU testsuite comparison:

Skip an intermittent issue tests/misc/tee (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/tail/follow-name is no longer failing!

[github-actions]

GNU testsuite comparison:

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

[github-actions]

GNU testsuite comparison:

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

[Alonely0]

I swear macOS CI is so (demoralizingly) petty it has internal failures in purpose when it knows it's gonna pass lmao.