Animation plays only in master about source

[jmespadero]

Hi,
I just wanted to recall this issue, which I found is a feature-stop for many applications.
I suppose that it can be solved by the master sending the current animation
frame of each object, in the same way that it is sending its position and orientation.
Isn't it?

Not sure if there any way to do this with a processor.py file...

[PyrApple]

The bug seems to be related to scene.suspend() not working on animations in BlenderVR (while it does obviously in Blender). Even in single screen mode (no slave), everything happens has if a third party was handling animation while scene is suspended.
see test file: https://dl.dropboxusercontent.com/u/67761747/anim-test.zip

[jmespadero]

Hi,
The anim-test.zip file is (almost) empty. It contains just a link to another directory.

I have trying to hack a patch for this, asking the master to send the current animation keyframe to slaves, using getSynchronizerBuffer(self) and processSynchronizerBuffer(self, buffer), but
I still have it not working.

By the way, I was suspecting that the problem was elsewhere... I was suspecting of action culling (means actions will not be played when the bounding box of the armature is outside of the camera's viewing frustum), as you can see here:
http://blenderartists.org/forum/showthread.php?392168-Pose-position-from-one-camera-rest-position-from-the-other

[PyrApple]

Interesting clue, will investigate. Meanwhile test file update: https://dl.dropboxusercontent.com/u/67761747/anim-test.zip

[jmespadero]

Hi,
I see several things not working in the anim-test.blend file (mainly, that slaves are not moving the "fingers" on the floor).

But I don't understand the issue with scene.suspend(). How do you run the test? What is the expected behavior? Is the .processor.py file doing something to launch scene.suspend() or
something else?

[PyrApple]

As you say, several things not working so there's not yet one way to run the test: the expected behavior is synchronization, the scene implement animation / color / visibility modifications in bge to test those features.
That being said, the issues highlighted are, running the scene in BlenderVR split mode:

• slaves are not moving the "fingers" on the floor. Moving the camera (in original .blend) to position them on either master or slave screen (or in between) has an impact on their being animated.
• cube animation continues while scene is suspended from master
• cube animation is only half overwritten on slaves (as are finger's animation for some initial camera orientations): one can see the cube fast teleport between both animation paths (the one from the master vs the one from the slave, somehow unsync at BlenderVR start).
  Note: the scene.suspend() (triggered by spacebar or shift) is impl. in a script embedded in the .blend (executed in the .processor.py file will provoke another behavior).

[dfelinto]

One thing can help this report is trying to pin point which was the last Blender where this worked. Even better if the exactly commit could be pin pointed (via git bisect)

[PyrApple]

agreed, I'll work on that as soon as I manage to recompile blender (as I need to patch < 2.75a versions) on OSX 10.11, which somehow raises unexpected issues (or change OS).

[PyrApple]

Well, we may not have to go that deep finally.

While digging into old Blender/BlenderCAVE code, I actually compiled the current git version of Blender (2.76 - commit 2ecd67efa4aeadb442e094b795a9171149ccbab1), which solved the problem of KX_GameObject animation synchronization on Slaves. Remains to see if current 2.76 release does the job or if we have to wait for 2.77.

Regarding BL_ArmatureObject animations, removing the scene.suspend() applied on Slave nodes enables armature animations on Slaves, synchronized with Master's as it is.
This fix not being receivable as is (the suspend scene mechanism for Slaves exists for a reason - while a test on a physic simulation shows perfectly synchronized objects...), I created the fix_anim_in_slaves branch to dig in further.

[jmespadero]

That a great notice, but I have just tried blender-2.76b (the current release) and the bug is still there. We will wait for 2.77

[dfelinto]

@jmespadero you can get a daily build here: http://builder.blender.org/download

[PyrApple]

We've got a situation :)

Armature update in BlenderVR is handled in player/network/synchronizer/objects/item_armature_*.py, running a BL_ArmatureObject.update() if animation frame changed in master. Naturally, the scene.suspend() on slave nodes prevents any .update() flagged object to actually being updated.

Problem: none of the read/write properties of armatures in Blender impacting current pose have a direct effect, but rather will be "updated when the armature is updated during the graphic render" (see the BL_ArmatureChannel API page).

Solution 1: I didn't see it, a "setPose()" actually exists that, as e.g. worldTransform, acts immediately on the object without waiting for graphic render update. Everyone is happy, I'm humbled.

Solution 2: We deactivate scene.suspend() on slave nodes and think hard on what that fix will ... break.

Solution 3: We code a setPose() method in Blender code itself (fun ahead!)

[jmespadero]

I didn't see it, a "setPose()" actually exists [...]. Everyone is happy, I'm humbled.

Me too. I have been searching such feature for a time, and I haven't found it. I can't believe that feature is not available.