- User Since
- Dec 26 2013, 7:53 PM (168 w, 6 d)
I can reproduce the issue with 2.78b, but not with 2.78c and master.
Please try updating to 2.78c and checking if it also fixes the issue for you.
This is the same problem as in T49921 - the normal mapping changes the surface normal, so the reflected ray would go inside the actual geometry, which isn't really possible.
The way OSL is configured in CMake seems weird in general.
FindOpenShadingLanguage.cmake uses OSL_ROOT_DIR, but install_deps.sh uses CYCLES_OSL and platform_apple.cmake and platform_win32.cmake files seem to contain their own OSL library finder based on CYCLES_OSL.
The problem here is that you have Clamp enabled, which means that the Node is actually doing something (or rather, would do something if your textures were HDR).
However, there's no real reason to keep the full MixRGB node in the SVM code if only a clamp is needed, so I'll add a separate clamp-only SVM code.
Tue, Mar 21
Mon, Mar 20
I don't think that we should explicitly disable negative lights - after all, they don't really cause any problems for regular use and there's no real chance of using them by accident, so I don't see the harm in keeping them as an "undocumented feature" as long as we don't officially support them.
Sun, Mar 19
This is most likely a precision issue in the triangle intersection code.
Tue, Mar 14
The patch needs a manual rebase now that the split kernel changes are in master - If you're fine with it, I can do that since I already did it for the denoising branch.
Mon, Mar 13
Tue, Mar 7
Okay, this is getting weird - I could reproduce the issue with OIIO 1.6.9, but after upgrading my build environment to 1.7.8 the crash is gone.
However, the official 2.78c which is also linked against OIIO 1.7.8 still crashes.
Can't reproduce here (on Linux x64 and master), works fine with both CPU rendering and CUDA (on a GTX780).
Hm, this sounds like you somehow messed up the installation folder.
Please try removing Blender completely from your system and reinstalling. If it works after that, try adding the Filmic config again and see if it still works.
Mon, Mar 6
I can reproduce the crash, it happens inside of OIIO - to be precise, in OpenImageIO::v1_4::TIFFInput::find_tag which is called by intern/cycles/render/image.cpp:164.
Apparently Blender's image loading code can handle the Image, which is why it works when packed and only crashes when it's read from disk by OIIO.
Sun, Mar 5
The reason why Default and Filmic are treated with preference is simply because they are (or will be) included in Blender by default.
Fri, Feb 24
This is caused by rB87cb3f8519: Since the length function always returns 31, fluid.memory_estimate has trailing zero bytes to bring it to length 31. These zeroes are preserved in the formatting step in properties_physics_fluid.py and therefore end up between the "MB" and the closing parenthesis, which then causes the C interface code to ignore everything after the "MB".
Please try running Blender from the console with the command:
blender.exe -b --python-expr 'import _cycles; print(_cycles.available_devices())'
Baking is working as designed. Since you're baking as a texture, you want Blender to stay in scene-referred values here.
Considering that the same build of Blender works on a different driver, I'm pretty sure that AMDGPU is to blame here, not Blender.
Wed, Feb 22
To be precise, the RGB numbers are in the Scene Linear colorspace, which with the default OCIO config is Linear sRGB aka Rec.709.
Feb 13 2017
Feb 12 2017
I'm merging this into the denoising branch now, and it looks pretty good so far. Three minor things noted inline.
Feb 10 2017
Feb 9 2017
Feb 4 2017
I don't really agree that it should return a reflection - after all, it's explicitly called refract, not bounce_off_glass. So, when refracting isn't possible, the function shouldn't attempt to guess what you might want instead.
Another consideration is that you'll probably want to use fresnel weighting between reflection and refraction anyways, and in the case of TIR the refraction weight will be zero.
Feb 1 2017
Jan 31 2017
Thanks for the report, but this bug is already known and fixed - unfortunately, it was fixed after 2.78a was released.
So, to get GPU rendering working you can either download a build from https://builder.blender.org/download/ or wait for 2.78b, which should be released sometime this week.
Jan 22 2017
This isn't a bug, SSS behaves exactly as it's supposed to.
Jan 20 2017
Hm, right, it was too spammy in headless mode - fixed now.
Jan 19 2017
Okay, the two issues you mentioned are fixed.