https://bz.apache.org/bugzilla/show_bug.cgi?id=70309
--- Comment #1 from abernal <[email protected]> --- Confirmed all four points against the real current trunk source: 1. acme_get_ari()'s random selection: with c (0-255) computed as (end-start)*(c-128)/256, exactly 128/256 values go negative (before the window start), and the positive half never exceeds 49.6% of the window -- checked exhaustively over all 256 byte values, matches your analysis exactly. 2. A window with start == end is indeed accepted (only start > end is rejected). 3. md_time_parse_rfc3339()'s timezone offset: confirmed the exact bug -- the hour component is computed into tm_gmtoff and then overwritten (should be added to) by the minutes computation ("=" instead of "+="). Reproduced live against the real function: "+02:30" parsed as 11:30 UTC instead of 09:30 UTC, and "-05:00" parsed as no offset at all instead of 17:00 UTC. 4. Fractional seconds are indeed discarded (explicit "skip for now" in the code) -- left as-is for now, negligible practical impact on renewal timing. Fix for 1-3 and regression tests: https://github.com/apache/httpd/pull/969 Thanks for the detailed, well-analyzed report! -- You are receiving this mail because: You are the assignee for the bug. --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
