Confirmed. @Sergey Sharybin (sergey), am going to summon you here, since I believe there may be something wrong in DEG - animation is in the CacheFile datablock, hence not in Object one (as any other modifier would be). And it is never evaluated afaict, even though dependency looks correct?

More than a week without reply or activity. Due to the policy of the tracker archiving for until required info/data are provided.

Afaik that’s only possible with two main windows (New Main Window from main top menu), using different viewlayers.

Thanks for the report, but bugs about new depsgraph in old 2.7x series are really not relevant at all now, new deg is only developed and maintained in 2.8 branch.

One have to wiggle a lot to make it crash indeed… Looks like yet another COW/threading issue (threaded code from deg evaluation using some COW data already freed from somewhere else)…

Can not really reproduce the issue, but there is a huge amount of dependency cycles reported. The system became more sensitive to what relations are killable and which are not, so need to think of some solution to this.

@Sergey Sharybin (sergey) pchan_array is NULL during evaluation of pose (from depsgraph)… not sure how this can happen?

Can you list out how to have multiple viewports with different collections visible in 2.8?

Thing is, new depsgraph in 2.7x is not maintained anymore, you should only try it in 2.8 nowadays. So unless you can reproduce that behavior in 2.8, this report should be archived imho…

Thanks @Bastien Montagne (mont29), I will take note of this to don't make this mistake in the future.

@Joshua Leung (aligorith) Could you take a look? It's no totally clear what you do here and why you use in sculpt the hash.

OK, I'm going to take a look how is done (this is not my code). I thought it was related to depshgraph issue.

Crash has nothing to do with depsgraph, you are storing ID pointers (material ones) in a ghash accross redo step, with invalidate any pointer in Blender… will check, think you have to rather store ref to ID by its lib/name pair. Also, ghash here is int key, not string ones…

This has been fixed I think (at least, cannot reproduce the crash anymore).

@nBurn (nBurn) transferred the page, link has been updated.

fixed the wiki links, @Sergey Sharybin (sergey) can you transfer dependency graph design to the new wiki? (had to place a en.blender.org link for now)

There are no efforts to fix bugs in 2.79 when they won’t be portable to 2.8, and/or 2.8 does not show those bugs. Am sorry, but we are already very limited in dev power to finish 2.8, so no way we spend time fixing other versions.

Thanks, can confirm. Crash happens in BKE_animsys_eval_driver() during depsgraph eval, looks like some update/rebuild is missing, @Joshua Leung (aligorith), care to check? Thanks.

This crash seems to be fixed in blender-2.80-ab49b7d7396-win64

Note: most certainly same issue as T55770…

It a bit confusing because some features work with the "old" dependency graph and some with the new in version 2.79. I find this bug earlier: name And there was an answer: "new dependency graph is the new standard and there and it is very unlikely the old dependency graph will see fixes."

Please try with blender2.8, new depsgraph is not really maintained anymore in 2.79, the real new depsgraph is in 2.8. ;)

Welll then… will archive for now. Am sorry, but 2.79 won’t get any new fixes (especially not regarding non-finalized features like new depsgraph), so if this is not reproducible in 2.8 there is no point to keep it open really. Thanks for the report anyway.

EDIT: I tried to make a blend file and though it summarised the problem. seems like I can only provide a blank scene to ensure this bug is seen in a consistant manner. Can you please tell me which of hte steps I listed out are not precise enough? I can provide a blank blender scene if required, but it seems pretty pointless.

Note: might be related to T55791…