- User Since
- Jan 3 2003, 6:23 PM (751 w, 1 d)
macOS probably doesn't search the /tmp folder, that doesn't mean the file does not exist. With Finder > Go > Go to Folder > type /tmp you'll find it.
It works for me, tested by running this in the Python console:
import os; import signal; os.kill(os.getpid(), signal.SIGSEGV)
Thu, May 25
The behavior change looks good to me.
img.has_data tests if the image data is currently loaded in memory, it doesn't test if the image is valid if no attempt has been made to load it.
The macOS build should be fixed now.
Should be fixed now.
Wed, May 24
I only updated the darwin 10.9 libraries for the 2.80 builds, and this bug report is about 2.78 with Python version 3.5.2 in the screenshot. So it should not have caused this issue.
Tue, May 23
Right, some example renders + timings would be great, so it's easy for all of us to see how well this works.
Ok, well if you carefully check to not commit unrelated changes in these files I have no further comments.
Sat, May 20
It should be fine to pass a zero tangent, we do that in other cases as well, and if that doesn't work we can fix it. Linking in the default geometry node means extra unneeded work.
Also, this is using the Fresnel along the normal instead of Fresnel integrated over the microfacet normals. I guess that is a decent approximation, though I'm not sure how big the difference gets, if there is some case where it is very far off and noise is increased.
I agree such hardcoded limits are not helpful, just saying that this might not actually affect anyone in practice.
This bug report appears to be empty, if you intended to report a bug please create a new one with all the required information filled out.
Doesn't belong in the bug tracker I think, so changed type. Not sure paths longer than 4096 are supported by any commonly used OS even.
Looks good except that this patch seems to contain some unrelated changes in openexr_api.cpp, openimageio_api.cpp, bpy_rna.c, project_info.py and master_unpack.py?
It may be that there is something to be fixed or improved here. The first thing I would suggest to investigate would be rendering the two MIS components separately, by replacing power_heuristic() by 0 or 1, that might reveal something. On the other hand it might just be a case where both MIS components are noisy.
This is still a To Do item regardless of the amount of user interest. The render API was never designed for this purpose, and the network render addon is currently unmaintained.ware
is_zero_clamped() would indeed be a little more clear.
I'm not sure if we should release a fixed scene, testing if we handle weird HDR images correct is also valuable since users will likely encounter the as well.
I don't think the converged render should change because of this patch? That seems like a bug.
Adaptive sampling is certainly a feature we want to have in Cycles. I think we need a more advanced implementation than this though, with fewer users parameters, but mainly something is more robust. Looking only within a single pixel with a simple metric like this is not very reliable and will give artifacts in common cases. Maybe this is good enough, but then we'd really need to see some renders to prove that.
The patch looks wrong to me, this is the pdf for light sampling which we are not doing in this function. That pdf should only be used for the MIS balance.
Pressing A is normally not affected by the OpenGL Select method. There is likely something specific in this scene that is leading to a crash, selecting thousands of objects works fine for me, we need a .blend file to repro this issue.
Please describe the problem more precisely by attaching a .blend file and specifying the exact key shortcuts you use and buttons you click. It's not clear to me now what you are doing. Also mention the exact Blender version you're testing.
Please try the latest build from here, there have been some DPI improvements for 2.79 that might help.
Wed, May 17
@Aaron Carlisle (Blendify), what makes you say they there is not much we can do? Has this been investigated by a developer in another report?
Mon, May 15
Fn + Up and Fn + Down work as Page Up and Page Down keys in all applications on macOS.
Sat, May 13
I believe this issue was introduced in D2214, where a change was made to compute the time elapsed in Progress::get_time, which means every time it is requested for drawing. Previously it would be computed when updating e.g. the status, so when actual changes were happening.
The bug in the original report was fixed by rB8c5826f59a37: Fix T50698: Cycles baking artifacts with transparent surfaces. I believe, can be tested with latest builds from https://builder.blender.org/download/. It no longer crashes for me in latest builds while it was still crashing in 2.78c.
As a guess, you could try changing the Selection Method under User Preferences > System and see if that makes a difference.
OpenGL viewport support for this node was only added recently in D2425, and will be available in the upcoming 2.79 and daily builds from https://builder.blender.org/download/. It is not supported in 2.78c.
Thu, May 11
We interpret transparency as the absence of geometry, it should not be considered the same as refraction with IOR 1.
Problem introduced in rB9eb647f1c8d8: Fix T50656: Compositing node editor is empty, no nodes can be added.
I suggest to rename the denoising kernels, I guess that will be simplest.
I don't think this is implemented in the right place, RNA paths should not return "bpy.context". These are also used for e.g. the animation system and generally not aware of the Python API.
I think we definitely need to make the Principled BSDF complete enough that you don't need to mix in other closures, that's kind of the whole point. To make it faster for users to set up materials, and to provide a standard node for things like texture painting and material import/export. Not saying it has to be done right now or as part of this bug report, but certainly at some point.
It's not actually rendering more samples, but indeed the Time keeps increasing for some reason.
Confirmed, I think it's the node editor that is continuously redrawing.
You can try changing to a different Selection Method in User Preferences > System.
Please indeed try 2.78c or even the latest build:
I'm not sure what "importing multifiles" means, maybe a screenshot would indeed help understand that, never heard of that.
I'm guessing the issue is that Background Images only work in views like Left/Right/Top/.., not arbitrary viewport positions like the default one. That's not a bug though.
If you replace a node in the properties editor (and in the node editor), Blender will attempt to preserve any values and links for sockets with matching names. The issue here is that the Roughness in the Principled BSDF and Diffuse BSDF have a different meaning. We might improve the logic here at some point, but I don't consider it a bug at the moment, it's a To Do item.
Blender stores enum properties as an integer internally. If you specify a list of items, the integer values will just be the location of the item in the list, which can change if you change the number of items. This is not considered a bug. Possible solutions are:
Sun, May 7
Thanks, committed now for the 2.79 bcon3 deadline, we can always improve it further in master.
Sat, May 6
Ok, regardless, there have been some bugs in recent NVidia drivers that have been fixed in newer versions, and there can be issues where some installation is incomplete. There is not much we can do about such bugs on the Blender side besides waiting for NVidia to fix them. If we get more reports from users saying that things don't work at all for them with any of the recent drivers, that's something we can investigate, but so far this does not look like a Blender issue.
Long term I think it would be great if path and branch path tracing could be unified, and CPU and CUDA split kernels could get similar performance to the megakernels, and then we can avoid most of this code duplication. Starting the deduplicate the code into common functions like this is a good step I think.
We're already breaking pep8 in these Python files many times then, which I guess would be good to fix but can also be done after for the whole file.
Looks good to me.
Looks good to me.
From the example file in T51425:
This file has 20 million triangles, the 3ds exporter has not been optimized to handle such very high poly meshes and so may take a long time or run out of memory. For such heavy models a file format like Alembic is more appropriate.
Thu, May 4
To follow Cycles naming conventions I suggest we rename this socket to "Transmission" (and "Transmission Roughness"). It's a trade-off between following the Disney naming convention and our own, but I think in this case clearly distinguishing from the Transparent BSDF is more important.
Tue, May 2
I don't think it's a problem to allocate some temporary memory on the host. Ideally we'd support direct memory copies between devices without the host eventually as well. I don't insist on doing that now though, the code works as it is.
Mon, May 1
Ok, it's probably versions 381.65.
Sun, Apr 30
Thanks for the report, but this is intentional, people are free to browse and download all the files there.
By the way, the reason it's marked invalid is because we do not classify this as a bug. It's a valid concern, but this is a UI design discussion, which we do not handle as bugs in the tracker.
The timeline is one of multiple editors that has time along the X axis, like the sequencer or graph editor. We try to be consistent in the UI, so scrubbing should work the same in all editors that support it. And there does exist marker selection the timeline which this conflicts with, and maybe keyframe selection in the future.
Which exact Blender version are you using, have you tried 2.78c or the latest from https://builder.blender.org/download/ ?
See T50650, please try installing the driver from nvidia.com, there is a bug in the driver automatically installed by Windows 10 updates.
Scrubbing is not selection. With "Left" marker selection happens with LMB and scrubbing with RMB, with "Right" marker selection happens with RMB and scrubbing with LMB, so that seems to work as advertised.
Have you tried 2.78c and the latest builds from https://builder.blender.org/download/?
I don't see a bug here, the render layer enable/disable means that it will not render as part of the regular rendering process. If you explicitly press that button to render that specific layer then it does render, but I don't see how that's a problem.
Which exact driver version are you using? Have you tried installing the latest NVidia driver from nvidia.com?