date: expand GNU-compatible parsing and formatting

[anastygnome]

Handle extended years, GNU debug diagnostics, epoch seconds, normalization, DST gaps, and strftime modifier precedence

refs #14705

[codspeed]

Merging this PR will degrade performance by 12.58%

⚠️ 1 benchmark spent significant time in system calls

System calls cannot be consistently instrumented, so they are not included in the measure, which understates the real cost. Please switch to the Walltime instrument to accurately measure system calls.

Measurement and system calls

• single_date_now

⚠️ Different runtime environments detected

Some benchmarks with significant performance changes were compared across different runtime environments,
which may affect the accuracy of the results.

Open the report in CodSpeed to investigate

❌ 5 regressed benchmarks
✅ 7 untouched benchmarks
⏩ 444 skipped benchmarks¹

Warning

Please fix the performance issues or acknowledge them on CodSpeed.

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
❌	Simulation	file_custom_format	30 ms	36.5 ms	-17.9%
❌	Simulation	file_iso_dates	31.5 ms	38.2 ms	-17.52%
❌	Simulation	file_localized_names	41.9 ms	48.6 ms	-13.73%
❌	Simulation	file_tz_abbreviations	35.4 ms	38.1 ms	-7.07%
❌	Simulation	complex_relative_date	292.8 µs	311.4 µs	-5.97%

Tip

Investigate this regression by commenting @codspeedbot fix this regression on this PR, or directly use the CodSpeed MCP with your agent.

---

_{Comparing anastygnome:test-util (2b5fe3a) with main (38336e9)}

Footnotes

1. 444 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. ↩

[anastygnome]

extra work => additional cost.

[github-actions]

GNU testsuite comparison:

Congrats! The gnu test tests/date/date is no longer failing!
Congrats! The gnu test tests/date/date-tz is no longer failing!

[anastygnome]

CICD and Android are related to tail/cp, nothing related to this PR

[sylvestre]

sorry, but we can't take this: emit_gnu_debug hardcodes the expected output for specific GNU test inputs, which doesn't fix anything and looks taken from the GNU tests. If you want to keep going, please split it into small PRs that fix real behavior, one at a time.

[anastygnome]

@sylvestre would making the non debug tests into one PR be ok?

[sylvestre]

just the tests or tests+fixes?

[anastygnome]

I mean the code & test that fix tests/date/date & date/date-tz are very similar (I blieve matching debug output isn"t really wise but I had matched it to see if we could and discuss it here

[sylvestre]

Sure but this pr is huge. I would prefer smaller changes

[anastygnome]

I reduced the scope @sylvestre

[anastygnome]

@cakebaker @xtqqczze if one od you may review this it'd be appreciated

[cakebaker]

I'll likely wrap up the review tomorrow. The tests look fine.

[anastygnome]

No issues, take your time, I'll wait

[anastygnome]

@cakebaker thanks, I believe your suggestions make sense. Will implement them shortly.

[sylvestre]

same flag/width parsing as in substitute_epoch_seconds_value, could be dedup, no?

[sylvestre]

where does this list (wednes, tues, thurs...) come from?

[anastygnome]

GNU documents these forms in its "Day of week items" section.

[anastygnome]

didn't see the last line there, so I corrected it as well.

[sylvestre]

the patch is still too big

[cakebaker]

Done