umputun / rlb-stats / 38139198471

84%

master: 84%

Pull #67

github

Address revmux r2 findings

- Test for the drain-before-reopen fix now actually covers the race window.
  The r1 test passed with the drain removed because the top-of-loop drain
  always runs before stat under Run's ticker loop. The new test calls
  tl.open, tl.drain, writes to the old inode after the drain returned,
  rotates the file, and calls tl.checkRotation directly. Mutation-verified:
  removing the drain block in checkRotation breaks the test as expected.

- A Host header with the reserved value "all" would collide with AllNode:
  Candle.Update iterates ["all", "all"] and double-counts the request.
  parser.go now remaps that value to "all-host" at the ingestion boundary.
  POST /api/insert already rejects dest="all"; this is the nginx-side twin.

- README: the CF-* fields are client-controlled on anything that is not
  CF-only origin. Document the spoof risk and the two mitigations
  (restrict origin to CF IP ranges, or hard-code "-" in log_format).

The ReadBytes-pre-allocates-before-maxLineBytes-check finding is accepted
as immaterial; the 1 MiB cap plus typical ~1-2 KB access-log lines bound
growth in practice. A ReadSlice-based bounded reader is left for a
follow-up if a malicious writer ever shows up in anger.

345 of 447 new or added lines in 10 files covered. (77.18%)

983 of 1169 relevant lines covered (84.09%)

79.79 hits per line