- User Since
- Jan 3 2003, 6:23 PM (862 w, 5 d)
The JokerGargle_Displacement.exr is a different case than the original report. It has a single "R" channel. I'll keep the report open for that case.
What would help would be some code comments to clarify what each change is doing.
I looked at this for an hour but just can't properly understand what is going on in this fix. I think we should postpone this for after 2.80 even if it's a serious bug, we risk breaking something more important.
Actually I see the issue, but it's not caused by that commit. Will fix.
@Sergey Sharybin (sergey), I can't reproduce any crash enabling viewport rendering in this file. It also not clear to me how that commit could have added a new null pointer access, I think it would have already existed earlier.
I also noticed adding a new armature with shift + A in object mode does not immediately show used layers, when entering edit mode.
This was fixed in rBAefc92efec2cf: Fix regression in glTF animation export.
Adding a bone with shift + A also does not seem to update the dots in the latest patch.
Fill adds the bone on the active layer as far as I can tell, and that one may be empty.
Not considered a bug. This is an area for buttons with a fixed size, it can be made smaller.
I guess this is ok for 2.80 still.
Fine to commit to master still.
Looks good, now it can be added to more operators. From what I can tell these need it:
Only some effectors are animated, depending on the parameters. Re-evaluating things for all effectors is expensive if some were only used to make some static changes to the hair or particles. But I guess this is mostly used for animation is so it's probably ok.
This would be good to improve but it's not considered a bug currently.
This wasn't supported in 2.79 either, masks don't show on deforming modifiers there.
layer and layer_protected are user-edited data. If there is a problem with those when joining armature, that's something to be submitted as a separate fix.
This only fixes the case for joining armatures, there is a deeper problem here.
The preference controls the default value the first time you add an object in a Blender session. After that it remembers the last value set in the Adjust Last Operation panel. This is not considered a bug.
For now you need to bake the smoke for it to show up in the render.
This report was closed because not enough information was provided to reproduce the issue, it can be reopened if more information is provided.
Rendering with Cycles on RTX cards is supported.
@Demeter Dzadik (Mets) please don't close reports without commenting.
Re-downloading the current RC should also solve this problem.
This graphics card is below our minimum requirements, and so we can't provide official support for it.
This is the same issue as T52680: Alpha difference Viewport/F12 .
I doubt this is related to graphics drivers, probably it's deleting data that the tracking is using in another thread.
@Arto Kitula (akitula), was NDOF enabled in the release candidate build?
It was removed intentionally, so it's not a bug:
rBd2e139e44e0a: Color Management: update configuration, remove legacy transforms.
Can you explain how this is different than View Roll (shift numpad 4 and 6)?
This is ok to commit for 2.80.
This is ok for now. Step size really should always be 1.
Fine with me, but I think the tooltips should be clarified a bit further.
It should iterate over view_layer.objects instead of scene.objects for both the selection and non-selection case. Only then does it make sense to query if the object is visible in the view layer.
630M is a GPU with compute capability 2.1, for which we dropped support in Blender 2.80.