Developer Meeting Notes

news

(Grzesiek) #235

So you changed from September 24 to 11th? :wink: Getting closer. That or indeed they will have their next meeting in almost 1 year time?


(DarkNavi.exe) #236

7 minutes later:
This is the last time I copy/paste the format I was trying to mimic.
fixes it…again


#237

that was funny, it reminded me struggling every day with obvious code mistakes :smiley:


(Grzesiek) #238

Sorry for that. but again you are NOT accountable for Ton and the guys who write that up. You are doing a service for us to copy/paste it here, so i do appropriate it.

:slight_smile:


(Mr_Flamey) #239

I wonder if Blender being software that is typically used in cpu and gpu benchmarks helps secure funding and hardware from intel, amd and nvidia. I’d imagine that all of these companies want to make sure their hardware looks good in benchmarks, so it’s worth donating hardware to ensure proper compatibility, or funds to implement features that improve performance.


#240

Unleash I am mistaken I think Blender is already funded by AMD, Valve also used to fund Blender. There is definetly a lot of practical benefits for Companies to fund Blender so I would not be suprised if the list is way longer than that.


(drgci) #241

New Features and Changes

About 65 bugs were fixed last week.

  • Edit mode wireframe drawing has been improved, to be more crisp and better anti-aliased. ( Clément Foucault )
  • Outliner collection and object visibility have been revamped. ( Dalai Felinto )
  • Bevel: better attachment points for unbeveled edges, arc miter. ( Howard Trickey )
  • Image, movie clip and sequence editors can now show all image metadata. ( Sergey Sharybin )

Development

  • Automated testing: a new system for event simulation was added, for automated tests that are otherwise difficult to write. This used for testing texture paint and sculpt mode undo currently. ( Campbell Barton )
  • Removing fake user and automatic datablock deletion when users count drops to zero is being discussed. The Purge unused data operator has been improved. (task ) ( Bastien Montagne )
  • DNA Renaming: structs and members saved to .blend file currently cannot be renamed easily, which over time has lead to names getting out of sync with the UI and leaving in poorly chosen names. A new patch adds support for renaming. (code review ) ( Campbell Barton )
  • Cycles OpenCL compile time reduction is being worked on, starting with multithreaded compilation and gathering statistics. (wiki ) ( Jeroen Bakker )

Weekly Reports


(Ace Dragon) #242

Not only do we have the end of the fake user system being discussed and worked toward, but now we have William leading the charge for a smart system with which to have colored wireframes as well.

I’m left wondering how long the BForartists fork will even be used at this rate, as the devs. are really displaying a serious effort towards better usability and artist-friendliness.


(rboxman) #243

Sigh. I guess this is how open-source development typically goes but starting another sub-project while there’s so many that can/should be completed beforehand is typical blender – a little good at everything, but best in class at nothing.

Why not complete the ‘top bar’ implementation as it’s still kinda iffy? Address a few more dozens of already known ‘ui paper cuts’? How about one more pass through the various ‘call for contents’ over at devtalk that still haven’t been fully committed yet (matcaps have some updates that should go in)? Maybe actually focusing on fixing the Outliner/Viewport sync problem? Maybe making the Undo system actually work fully (https://developer.blender.org/T61045)? Or actually completing open subdiv support? How about making the open-data project actually worthwhile – it’s still in ‘beta’ and untouched since initial release; plus the data itself is suspect (no 1080’s listed but 1070 Max-Q shows up in top 5…hmm)? Before adding more configuration, which colored wireframe may require, why not de-dup the KeyMap editor and Theme editor so you don’t have to change the same key/color in 20 places? Industry standard keymap is also unfinished… etc. etc.

Will this just be another half done idea? Preferences are finally in a good state after they reworked it, but no thanks to the primary 2 folks involved who thought it was ‘done’ when it was first committed. It’s hard to share your optimism man :slight_smile:


(Ace Dragon) #244

Just because something was not worked on for a little while does not mean it has been abandoned.

For instance, Collections and the Undo system has had a number of commits in the last week. OpenSubDiv has a notable optimization through the topology cache that is waiting to go in, and the viewport code may soon get some major bottlenecks resolved through multi-threading.

Between all of those however, there are dozens, if not hundreds of smaller todo items that need to be worked on to make Blender 2.8 the release it is getting hyped up to be. The rate of development seems to indicate that massive development still sits between now and the release.


(stargeizer) #245

Everything you described here is just normal Blender development. This is not like proprietary parts… As just said, one piece has to work with another piece of the puzzle and some are more complex than another. Also, Blender 2.8 is still under heavy development, so nothing new here. (compared with proprietary parts, the 2.8 beta should be treated the same as “internal beta” in proprietary counterparts, something final users should never see).

The word that you are looking to use for all what you say and asking is “polishing”. The polishing stage (where most paper-cuts will be fixed anyways, like usual) has not even started yet. The internal base is still not completed, so the UI neither.

For the Preferences window, is still Work in Progress… The Redesign was done on January, but the polishing and bugfixing is nowhere near to be complete. But this is also WIP so we can see where the design is headed. Which can be good or bad, depending on anyone’s POW.


(Craig Jones) #246

Sounds like you missed out on the 2.49 to 2.5 days. Relax, it is all coming together at the speed it needs to and much faster than we saw during 2.5 era.


(m9105826) #247

My main concern at least is that more often than not, Blender development never seems to reach that polishing stage before people latch onto the next shiny thing, which is unfortunately the curse of open source development.


#248

You imply that closed source projects are more polished ? Ever tried to use the polygon editing tools of ZBrush ?


(nudelZ) #249

Ever tried to use the sculpting tools of blender (especially 2.8)? its laughable in comparison


(Ace Dragon) #250

If 2.8’s release timetable is the same as every other one set by the BF, it will be delayed by a month or more. There is a reason why they keep doing commits to 2.7x even with it now being in a branch.


(rboxman) #251

Various things culminated for my rant. The artificial, and often unwarranted, hype-machine that is Ace :slight_smile: doesn’t help set expectations; at all. Yet another unwarranted and unsubstantiated hype message was a final trigger. 2.8 was supposed to be feature complete in Nov. There’s no dev resources for yet another sub-project right now. Remember that some things, like proper GPU support for OSD will not come for 2.80 regardless of how much it’s delayed – let’s get 2.80 out there so those larger projects can start for 2.81+.

The meandering river that is, well, all software development (oss and closed) is frustrating regardless. I’d rather have 3 full-ass’d areas of “software X” rather than 10 half-ass’d areas.

I personally love the state of 2.8 so far (except for the half-ass’d parts). The risk of this new colored wireframe sub-project, with no devs signed up as of yet btw, being merged and then causing even more grief than before is not zero, if history is any indication. If that happens, hopefully there will still be resources to fix it up before yet another sub-project steals attention away. Both oss and closed has this problem. Perhaps one should at least try to put on the appearance of being better than the other though.


(Ace Dragon) #252

This was a common misconception, what the developers were really gunning for with Beta was to simply have the foundation in place so work can move towards making 2.8 fully usable and production-ready.

Looking at the commit logs over the last months, much of the effort is not on shiny new features, but on fundamental workflow topics and on areas riddled with bugs and other issues. Many of the shiny features were brought in via patches made by volunteer developers. There’s a lot of areas in 2.8 that need shoring up and the devs. can’t address them all at once.

We know for one thing that there is topology cache code almost ready go in to speed up subsurf, but Sergey’s time has been occupied with making sure the depsgraph works with complex animation rigs and that hair particles work well (in part due to the current Open Movie Project). Things should also get a bit better once Clement formally commits multithreaded code to the draw engine, as hinted by some previous commits.


As for the hype machine bit, I know there is pressure right now to be pessimistic, but to make sure 2.8 becomes the best it can be means the forum must remain useful for some of the developers (or otherwise the ability for user/dev. interaction for purposes like feedback can get badly hurt). It doesn’t mean I blindly agree with everything the devs. say, but I don’t want to end up being the one who turned all of the potentially useful threads into noise.


(0o00o0oo) #253

Not to sidetrack, but a quick sidetrack… How did you do your horizontal line in the last post? Can’t figure it out, haha.


(Ace Dragon) #254

The ‘-’ key. Press three times.

You can also put it under a line to make the text big.