bug: no time information for files before UNIX epoch

[yerlotic]

Info

• Version: v0.23.4
• OS: Arch Linux x86_64

Steps to reproduce:

1. Create files with random timestamps:

   mkdir a && cd a
   touch a -d '1 Jan 1970 00:00'
   touch b -d '2 Jan 1970 00:00'
   touch c -d '00:00'
   touch d -d '20000 years ago 00:00'

2. Run ls and eza with -l --time-style full-iso:
   1. On ext4 filesystem it produces this output:

      a $ ls -l --time-style full-iso
      total 0
      -rw-r--r-- 1 user user 0 1970-01-01 00:00:00.000000000 +0100 a
      -rw-r--r-- 1 user user 0 1970-01-02 00:00:00.000000000 +0100 b
      -rw-r--r-- 1 user user 0 2025-11-23 00:00:00.000000000 +0000 c
      -rw-r--r-- 1 user user 0 1901-12-13 20:45:52.000000000 +0000 d
      a $ eza -l --time-style full-iso
      .rw-r--r-- 0 user -                                   a
      .rw-r--r-- 0 user 1970-01-01 23:00:00.000000000 +0000 b
      .rw-r--r-- 0 user 2025-11-23 00:00:00.000000000 +0000 c
      .rw-r--r-- 0 user -                                   d
      

   2. On tmpfs (in /tmp):

      a $ ls -l --time-style full-iso
      total 0
      -rw-r--r-- 1 user user 0 1970-01-01 00:00:00.000000000 +0100 a
      -rw-r--r-- 1 user user 0 1970-01-02 00:00:00.000000000 +0100 b
      -rw-r--r-- 1 user user 0 2025-11-23 00:00:00.000000000 +0000 c
      -rw-r--r-- 1 user user 0 -17975-11-23 00:00:00.000000000 -0001 d
      a $ eza -l --time-style full-iso
      .rw-r--r-- 0 user -                                   a
      .rw-r--r-- 0 user 1970-01-01 23:00:00.000000000 +0000 b
      .rw-r--r-- 0 user 2025-11-23 00:00:00.000000000 +0000 c
      .rw-r--r-- 0 user -                                   d
      

3. Output is different and eza doesn't show time before the unix epoch

Here is code generated by Github Copilot that fixes this issue. It should probably be
reviewed.

    fn systemtime_to_naivedatetime(st: SystemTime) -> Option<NaiveDateTime> {
        // Try the usual case first: SystemTime >= UNIX_EPOCH
        if let Ok(dur) = st.duration_since(SystemTime::UNIX_EPOCH) {
            let secs = i64::try_from(dur.as_secs()).ok()?;
            let nsecs = u32::try_from(dur.subsec_nanos()).ok()?;
            // Convert to local DateTime and return its naive_local representation.
            // timestamp_opt returns a LocalResult; single() yields the DateTime if valid.
            return Local.timestamp_opt(secs, nsecs).single().map(|dt| dt.naive_local());
        }

        // Handle times before the UNIX epoch.
        let dur_before = SystemTime::UNIX_EPOCH.duration_since(st).ok()?;
        let secs = u64::try_from(dur_before.as_secs()).ok()?;
        let nsecs = u32::try_from(dur_before.subsec_nanos()).ok()?;

        // For timestamps before the epoch we need to produce a negative seconds value.
        // If there are subseconds (nsecs > 0) we borrow one second and invert the
        // nanoseconds: timestamp = -(secs + 1) + (1_000_000_000 - nsecs) nanoseconds.
        let (secs_adj, nsecs_adj) = if nsecs == 0 {
            (-(secs as i64), 0u32)
        } else {
            (-(secs as i64) - 1, 1_000_000_000u32 - nsecs)
        };

        Local.timestamp_opt(secs_adj, nsecs_adj).single().map(|dt| dt.naive_local())
    }

Output with this code replacing systemtime_to_naivedatetime() in src/fs/file.rs:

1. On ext4:

   $ eza -l --time-style full-iso
   .rw-r--r-- 0 user 1970-01-01 00:00:00.000000000 +0000 a
   .rw-r--r-- 0 user 1970-01-02 00:00:00.000000000 +0000 b
   .rw-r--r-- 0 user 2025-11-23 00:00:00.000000000 +0000 c
   .rw-r--r-- 0 user 1901-12-13 20:45:52.000000000 +0000 d
   

2. On tmpfs:

   $ eza -l --time-style full-iso
   .rw-r--r-- 0 user 1970-01-01 00:00:00.000000000 +0000   a
   .rw-r--r-- 0 user 1970-01-02 00:00:00.000000000 +0000   b
   .rw-r--r-- 0 user 2025-11-23 00:00:00.000000000 +0000   c
   .rw-r--r-- 0 user -17975-11-23 00:00:00.000000000 +0000 d
   

The output still differs from ls in timezone but I'm not sure if it is a bug in ls or eza