mirror of
https://github.com/storytold/lightcraft.git
synced 2026-11-13 17:31:09 -06:00
9c07f427f0
* Arm the window renderer marker before probing the driver The issue's own table says the window renderer is created without a crash marker, so a launch that dies in the driver writes nothing and repeats itself. The window does have a marker, but it is armed only after the risky probe has already run, which is exactly the case the reporter's Windows launch hit. Arm the window-init.marker with the renderer name before each probe in pick_probed, so the marker on disk names the renderer being tried while the driver is touched, and re-arm it for the next renderer when a probe fails. The next launch then reads that renderer via take_failed and reports it (the existing 'the window could not start with its ... renderer last time' notice) instead of failing silently again. Test: the_window_marker_is_written_while_the_renderer_is_probed failed before the change and passes after. * Window marker: fix the test's closing brace, skip it under forced glow, sync the marker to disk Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> * Format Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> --------- Co-authored-by: eliets2 <287273517+eliets2@users.noreply.github.com> Co-authored-by: FredApps <153173854+FredApps@users.noreply.github.com> Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>