Here are the notes from today's meeting in irc.freenode.net #blendercoders. Most notably, some bugs crept into 2.78b which means we'll soon see a 2.78c release.
Ton writes:
1) Blender 2.78b performance release
- Not everything went OK... three issues were reported that would need to be solved:
- OSX build had wrong cuda kernels.
- GLSL viewport is broken when Light Path node is used
- Crash of material preview when using Image sequences
- Agreed is to update the 'b', to not confuse people we call it a 'c'. Most likely release will be on Wednesday.
- BTW: Regression tests were updated to include the showstopper cases.
2) The 2.79 release targets
- Check planning and targets here.
- Luca Rood: Surface Deform modifier is pretty much ready for master now. SDef tech docs here.
- Sergey Sharybin: review of catcher patch started, including user side testing.
- Lukas Stockner: most of the non-essential Denoiser code has been removed locally, will be pushed later.
- Denoiser needs good docs! Volunteers welcome (makes it faster to get Denoiser in master)
- Alexander Romanov proposes to add this fix for custom ID properties in 2.79. Is under review still.
3) Blender 2.8 project
- Julian Eisel wants to get more feedback on Tooltip Guidelines. Decision will be made/confirmed by UI module team via bf-interface list.
- Bastien Montagne: Blender 2.8 branch gets a 2.8 add-on branch too, to help testing and fix changes in api.
- Mike Erwin notes great progress on OpenGL task force. Graph is is on the move again.
4) Other projects
- UPBGE. This topic is still open. We will ask Clement Foucault and Dalai Felinto to feedback on the Viewport/shader issues.
- Ton Roosendaal suggests to have another midweek 2.8 meeting session, this time with as main topic 2.8. Call will go out in a week, after gathering feedback.
- The Google Summer of Code application has been finalized. Bookmark this url to give to interested students.
Thanks,
-Ton-
3 Comments
Never thought I'd be so interested in a Grease Pencil function, in all my life. But I am. It's increasingly becoming one of the neatest things in Blender.
Blender is such a complex application, why don't you guys start adding unit-tests for it? This would reduce the chances of regression errors.
We do use unit tests, but Blender has a giant code base. Only a small fraction of it is test covered right now, having an (almost) complete coverage is an utopia for the time being.
I agree unit tests would help reducing the appearance of regressions, but only a bit I'm afraid. Cycles is well covered by unit test and yet, users often find regressions in it.