- Lots of bug triage and user support in Launchpad and upstreams.
- Triple buffering for GNOME 48:
- Too much time spent investigating a check-code-style bug causing style warnings. It’s a design limitation in Mutter’s style checking script and not something that can be simply fixed without embarking on a larger overhaul of the style checker.
- Merged the gnome-shell fix upstream that was causing CI failures for triple buffering. Seems like it was just bad luck and a race condition being exploited that made it look like triple buffering’s fault.
- Fixed reported frame drops and runtime check failures caused by triple buffering (yes someone found a real bug). Now fixed in both the 47 and 48 branches. 46 should be unaffected.
- Finally unified the flipping/posting code paths of direct scanout and composite rendering so all frames are handled the same. Triple buffering during direct scanout is now supported and the two features no longer inhibit each other. This also means VRR (still experimental in Mutter) no longer inhibits triple buffering.
- Fixed half of the new test failures.
- Down to 3 unresolved discussion threads from 10 last week.
- Uploaded the next batch of Mutter SRUs for Noble.
- Uploaded Alessandro’s startup fix for Noble.
- Detours (bugs found during triple buffering testing but not actually caused by triple buffering):
- Investigating fullscreen performance glitches. Bisection is proving too hard because there’s no obvious point of regression. I will debug it properly in time for GNOME 48…
- Rediscovered a shutdown crash in the cursor manager.
- I’m posting this early due to the national holiday on Monday.
3 Likes