v1.09.57 still ran the quarantined tests — and win-x64 timed out
at 30 minutes. Root cause, verified against the doctest 2.4.11
source in-tree: the short flag for --test-case-exclude is
-tce, not -e (bare -e is silently ignored), and repeated
-tce occurrences do NOT accumulate (parseOption takes one).
So the repo's long-standing -e=*[slow]* excluded nothing: every
tier always ran everything, including the 240 s storm cases —
which also explains past 30-minute Test timeouts on loaded
runners. Both tiers now use a single comma-separated -tce
(verified: default tier runs 1571, slow tier 4, quarantined cases
still pass explicitly on demand).
No product-code changes versus v1.09.57 (server-side HRB UDF loading ships here unchanged).
- Local: tiers behave as designed; only failure anywhere is the documented pre-existing MinGW-only CDX alloc-tail case (fails identically on the clean tree; passes on MSVC/clang).
- Full matrix (Windows MSVC x64/x86, Linux, macOS) runs in CI on the tag push.
openads-1.09.58-windows-x64.zipopenads-1.09.58-windows-x86.zipopenads-1.09.58-linux-x64.tar.gz(HRB-enabled serverd)openads-1.09.58-macos-universal.tar.gz
Field triage, release engineering and validation: Pritpal Bedi.