https://gitlab.synchro.net/main/sbbs/-/commit/859eb16f4937a4a66bbee5c3
Modified Files:
exec/tests/webmail/drafts.js
Log Message:
webmail tests: don't let a slow run fail the "left to recovery" sweep
drafts.js runs on a fixed clock (NOW, taken when the file starts), but
after a crash between release()'s mark and its rename, draft.json no
longer holds the claim time, so Drafts.sweep() dates the claim by its
files, which have real modification times. The check swept at NOW + 700
with a 600-second live window, so it only passed when the claim was
written within 100 seconds of the file starting. Under valgrind on a
busy runner the file takes over 200 seconds, and the check failed
("crash before the rename: left to recovery ... got [0,[],[]]") in the
FreeBSD valgrind CI job. Sweep from the time of the check instead.
Co-Authored-By: Claude Opus 5.5 <
noreply@anthropic.com>
--- SBBSecho 3.38-Linux
* Origin: Vertrauen - [vert/cvs/bbs].synchro.net (1:103/705)