stat: fix precision when rendering mtime (%Y)

[jfinkels]

Support precision when rendering the time of last data modification with stat. For example, after this commit

$ stat --printf='%.1Y\n' f
1668645806.7

Previously, the precision in the format specification was ignored. This is implemented with a custom renderer because GNU stat seems to truncate the number as opposed to rounding the number as would happen when using format! with a specified number of digits of precision.

Fixes #3233

[sylvestre]

Sweet, i was working on it but your implementation is better than mine :)

Could you please add this test? It covers more cases

#[cfg(feature = "touch")]
#[test]
fn test_timestamp_format() {
    let ts = TestScenario::new(util_name!());

    // Create a file with a specific timestamp for testing
    ts.ccmd("touch")
        .args(&["-d", "1970-01-01 18:43:33.023456789", "k"])
        .succeeds()
        .no_stderr();

    let test_cases = vec![
        // Basic timestamp formats
        ("%Y", "67413"),
        ("%.Y", "67413.023456789"),
        ("%.1Y", "67413.0"),
        ("%.3Y", "67413.023"),
        ("%.6Y", "67413.023456"),
        ("%.9Y", "67413.023456789"),
        // Width and padding tests
        ("%13.6Y", " 67413.023456"),
        ("%013.6Y", "067413.023456"),
        ("%-13.6Y", "67413.023456 "),
        // Longer width/precision combinations
        ("%18.10Y", " 67413.0234567890"),
        ("%I18.10Y", " 67413.0234567890"),
        ("%018.10Y", "0067413.0234567890"),
        ("%-18.10Y", "67413.0234567890 "),
    ];

    for (format_str, expected) in test_cases {
        let result = ts
            .ucmd()
            .args(&["-c", format_str, "k"])
            .succeeds()
            .stdout_move_str();

        assert_eq!(
            result.trim(),
            expected,
            "Format '{}' failed.\nExpected: '{}'\nGot: '{}'",
            format_str,
            expected,
            result.trim()
        );
    }
}

[jfinkels]

Yes, I'll definitely add those tests, thank you. Sorry for the overlapping work.

…

On Saturday, January 11th, 2025 at 4:39 AM, Sylvestre Ledru ***@***.***> wrote: Sweet, i was working on it but your implementation is better than mine :) Could you please add this test? It covers more cases #[cfg(feature = "touch")] #[test] fn test_timestamp_format() { let ts = TestScenario::new(util_name!()); // Create a file with a specific timestamp for testing ts.ccmd("touch") .args(&["-d", "1970-01-01 18:43:33.023456789", "k"]) .succeeds() .no_stderr(); let test_cases = vec![ // Basic timestamp formats ("%Y", "67413"), ("%.Y", "67413.023456789"), ("%.1Y", "67413.0"), ("%.3Y", "67413.023"), ("%.6Y", "67413.023456"), ("%.9Y", "67413.023456789"), // Width and padding tests ("%13.6Y", " 67413.023456"), ("%013.6Y", "067413.023456"), ("%-13.6Y", "67413.023456 "), // Longer width/precision combinations ("%18.10Y", " 67413.0234567890"), ("%I18.10Y", " 67413.0234567890"), ("%018.10Y", "0067413.0234567890"), ("%-18.10Y", "67413.0234567890 "), ]; for (format_str, expected) in test_cases { let result = ts .ucmd() .args(&["-c", format_str, "k"]) .succeeds() .stdout_move_str(); assert_eq!( result.trim(), expected, "Format '{}' failed.\nExpected: '{}'\nGot: '{}'", format_str, expected, result.trim() ); } } — Reply to this email directly, [view it on GitHub](#7115 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/AAA5XG7RCQCTXITVFXXFVQ32KDRD3AVCNFSM6AAAAABU7QB25SVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKOBVGE3TOMBUGE). You are receiving this because you authored the thread.Message ID: ***@***.***>

[jfinkels]

In order to pass the new test cases, I needed to update the code a bit. There is now a Precision enum that distinguishes between NotSpecified (as in %Y), NoNumber (as in %.Y), and Number(usize) (as in %.3Y). I also made some slight adjustments to the proposed tests: (1) the number of spaces was off by one in the padding test cases, and (2) the result.trim() was trimming the spaces.

There is still a few aspects of the code that are not quite right, but I think they can be fixed in separate pull requests. First, there is some numerical inaccuracy between getting the Metadata::mtime_nsec() and the formatting code, so what should be .7639 is actually .7640. Second, the handling of %.Y does not quite match the documentation, which specifies 9 digits of precision.

[sylvestre]

Yeah, it is surprisingly complex :)

[github-actions]

GNU testsuite comparison:

GNU test failed: tests/misc/stdbuf. tests/misc/stdbuf is passing on 'main'. Maybe you have to rebase?
Skip an intermittent issue tests/tail/inotify-dir-recreate (fails in this run but passes in the 'main' branch)
Congrats! The gnu test tests/chmod/symlinks is no longer failing!