Thanks for looking into this. The fix addresses a real problem, but the approach in match_func() is fragile because is_defined_match(COL_TARGET) is not a reliable indicator that --target was used. set_match(COL_TARGET, ...) is also called from:

• get_next_fs() — source/target swap fallback when source-only match fails
• poll_match() — same swap logic for --poll
• positional args: findmnt <source> <target> (without FL_NOSWAPMATCH)

The swap cases would cause false shadowed-entry filtering on what was originally a source spec.

The good news is that this is already solved by existing infrastructure — we just need to enable it for --target. The mount table is read from /proc/self/mountinfo where later mounts appear later in the file. With MNT_ITER_BACKWARD, add_matching_lines() iterates from the end, so the last-mounted (topmost/visible) entry is seen first. Combined with FL_FIRSTONLY, it stops after that first match — returning exactly the visible filesystem. No changes to match_func() needed.

The fix would be something like:

case 'M':
    findmnt.flags |= FL_STRICTTARGET;
    FALLTHROUGH;
case 'T':
    set_match(COL_TARGET, optarg);
    force_target = 1;
    break;

where force_target is a local int in main(), and then after the getopt loop:

if (force_target) {
    direction = MNT_ITER_BACKWARD;
    findmnt.flags |= FL_NOSWAPMATCH | FL_FIRSTONLY;
}

— assisted by Claude Code