You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Tried to reproduce this with controlled atimes (via `os.utime`) rather than relying on whatever the filesystem happened to record during a `cargo run` build, and the exact-match semantics for `-amin N` (floor of elapsed-seconds/60) matched GNU find's documented behavior exactly:
So I don't think there's a semantic bug in the age-comparison logic itself. My read is the original repro (`find -amin 1` with no path, right after `cargo build`) just depends on which files in the tree happen to have an atime in the [60s, 120s) window at that exact moment — which is inherently timing-dependent and not something a test can pin down reliably. GNU's output (`./target/debug/find.d`) reflects a file whose atime happened to be ~1 min old on ITS particular run, not a difference in matching logic.
Happy to be pointed at a more deterministic repro if there's a real discrepancy I'm missing (e.g. a specific boundary case), but wanted to flag this before spending more time chasing a timing artifact.
Activity
MsfPablo commentedon Aug 7, 2026
Tried to reproduce this with controlled atimes (via `os.utime`) rather than relying on whatever the filesystem happened to record during a `cargo run` build, and the exact-match semantics for `-amin N` (floor of elapsed-seconds/60) matched GNU find's documented behavior exactly:
So I don't think there's a semantic bug in the age-comparison logic itself. My read is the original repro (`find -amin 1` with no path, right after `cargo build`) just depends on which files in the tree happen to have an atime in the [60s, 120s) window at that exact moment — which is inherently timing-dependent and not something a test can pin down reliably. GNU's output (`./target/debug/find.d`) reflects a file whose atime happened to be ~1 min old on ITS particular run, not a difference in matching logic.
Happy to be pointed at a more deterministic repro if there's a real discrepancy I'm missing (e.g. a specific boundary case), but wanted to flag this before spending more time chasing a timing artifact.
find: match GNU rounding for -amin/-cmin/-mmin