Page MenuHome

3d view objects not rendered.
Closed, ResolvedPublic


For a few days, now I have been getting a very strange problem in the cycles rendered 3d view.
Some times the objects appear normal. Sometimes they appear black, or not rendered. There doesent seem to be any paturn that I can spot.
You can load the same file twice and have it work one time, and then not the next. This is happening with many of my blends but is totaly random. Some times switching a material and switching back will cure it. Some times assigning new uv cords works.
I think it was introduced about the time bmesh became merged, or possably with AO.
I have been tryng to solve the problem myself for some time, but keep drawing a blank.
This vieport render was with r44687 on xp sp3. Ive allso up dated to the latest drivers, and updated to cuda toolkit 4.1 which has cured a memory allocation error while compiling the 1.3 kernal. I compile with cmake and vc2008.
The following image shows the problem.
Ive included the console read out as I think this is connected.
An F12 render works fine.



Event Timeline

This problem my only affect cmake anv vc2008 builds. Ive spent a few hours using bat3a's 44664 build from grapicall, without encountering the bug.
He uses scons & msvcsp1 to compile.

Sorry for being so vague about this, but its not down to the compiler, as I have now had the same fault with bat3a's build.
The objects render correctly in all other modes in the 3d view and a full f12 render.
Only in the 3d cycles rendered mode is there a problem. some times it is a individual object, sometimes the whole scene excluding the background.

I should add, I have tried a completely clean cheakout from the SVN and still getting the random error. Build 44702.

In the 3D view port, the offending oject renders black, until "Transparent" is checked in Film tab. It then becomes totally transparant enableing the back ground grid to be seen.
Ive tried all the usuall fixes, such as appending, applying modifiers, recreating materials, reapplying uv cords.
Its as if it pics an object at random assigns an incorrect draw mode to it.
Does it have anything to do with "masks"

You might notice in the above series of images, it is not only the deck that is affected. Also the funnel tops are rendered incorrectly.
The deck has uv cords and texture map, the funnel top has just a diffuse material.

Can you please check builds from ? Such kind of things might depend on build environment and this need to be figured out.

These pics, are using the build bot build rev 44696 win 32. this the layer and mask settings.

On exiting blender and restarting, it displayed this file correctly.

i've noticed this issue as well on osx 10.7.2.

it occurs random and solves by reloading cycles (stop viewport render, start it again).

i could not see any pattern where and when it appears.

Ok, I think ive spoted a paturn at last.
As mad as it sounds, it seems to start occuring on the 5th opened file in a session, with possably 1 or 2 objects rendered incorrectly.
Then getting progresivly worse.

The same issue occurs to scons compiled on osx 10.7.3.
Sadly I cannot add any advise to what already pointed out.

Switching to Static BVH seems to make objects rendered again, but it sticks to black screen, pass 1, for a long time and goes very very slow, respect to previous versions.

Switching back to Dynamic BVH all object now render fine.

No, sorry, one object disappeared totally (not black but invisible).

I can confirm that behaviour on Windows 7, (self compiled svn builds). Going gtom Solid to Rendered View, all is black. If you try again (Solid > Rendered) it most of the times works. I smell B-Mesh.... :/

Attached a file (File2.blend). Tested with SVN 44783) on Windows7
Open it and change to Rendered View in the 3D View. Maybe it works, maybe its complete black. If it works, abort render (go back to solid or wireframe mode) and change the current frame, then go back to Rendered View. Sooner or later it happens and the 3D view only renders black.

will investigate this week (if nobody gets there first)

The holdout layer wasn't initialized on re-rendering (using garbage values).

fixed r44820