TextureAttrib of loaded .bam file randomly ignores off-stage about panda3d HOT 4 CLOSED

Have you tested this with 1.10? I continue to be unable to reproduce this issue.

from panda3d.

Just tried it on another, newer machine, with a very recent Panda3D version (8a35f36 from 2017-10-18) and yes, it definitely still happens. If anything, the frequency with which the results vary seems to have increased; it now gives a wrong result like 1 time out of 2, while it only went wrong 1 time out of 5 or so with older Panda versions, IIRC.

Here is a short video I recorded to show exactly what happens.

Since I get the same results on three different machines (1 laptop and 2 desktops), it seems doubtful that it has anything to do with a specific graphics card (although they are all NVIDIA), but perhaps you want the output when I set "gl-debug true" and "notify-level-glgsg debug" in Config.prc, just in case?
Here is already the last part of it, just after loading the textures, therefore perhaps the most relevant part:

:display:gsg:glgsg(debug): loading uncompressed texture rockwall_diffuse
:display:gsg:glgsg(debug): loading new texture object for rockwall_diffuse, 1024 x 1024 x 1, z = 0, mipmaps 1, uses_mipmaps = 0
:display:gsg:glgsg(debug): Buffer detailed info: Buffer object 1 (bound to GL_ARRAY_BUFFER_ARB, usage hint is GL_STATIC_DRAW) will use VIDEO memory as the source for buffer object operations.
:display:gsg:glgsg(debug): Buffer detailed info: Buffer object 2 (bound to GL_ELEMENT_ARRAY_BUFFER_ARB, usage hint is GL_STATIC_DRAW) will use VIDEO memory as the source for buffer object operations.
:display:gsg:glgsg(debug): loading uncompressed texture Brick_105
:display:gsg:glgsg(debug): loading new texture object for Brick_105, 1024 x 1024 x 1, z = 0, mipmaps 1, uses_mipmaps = 0
:display:gsg:glgsg(debug): Buffer detailed info: Buffer object 3 (bound to GL_ARRAY_BUFFER_ARB, usage hint is GL_STATIC_DRAW) will use VIDEO memory as the source for buffer object operations.
:display:gsg:glgsg(debug): Buffer detailed info: Buffer object 4 (bound to GL_ELEMENT_ARRAY_BUFFER_ARB, usage hint is GL_STATIC_DRAW) will use VIDEO memory as the source for buffer object operations.
:display:gsg:glgsg(debug): GLGraphicsStateGuardian 0000000003D4FE00 destructing

The above output is always the same, regardless of the result.

Perhaps there is something wrong with the way the .bam file itself is created, so I attached it below, including the textures used:
model_hierarchy.zip

The newest machine I tested this on runs Windows 10 64-bit and has an NVIDIA GeForce GTX 750 Ti graphics card.

from panda3d.

I managed to reproduce it at last. The issue is that the compose mechanism relies on the states being sorted by pointer, but loading it from .bam removes the guarantee that the state pointers are sorted (since the texture stages are loaded in no particular order). All that's needed is one line to sort the pointers. :)

Thanks for the test case and reporting this issue!

from panda3d.

Great news! Thank you for fixing this!

from panda3d.