- User Since
- Nov 28 2009, 10:22 PM (407 w, 5 d)
Update against latest master
OIIO doesn't do image sequence manipulation, it's operating on file paht given to it as-is. Can you verify whether image_user_file_path() is giving a correct file name?
In the middle of something else atm, will have a look later.
@Bastien Montagne (mont29). did you test that with new depsgraph enabled?
@Bastien Montagne (mont29), maybe you can deduct something here already?
Are you showing a result of roundtrip of exporting scene from blender to fbx and back to blender?
Works on Linux. Wondering whether macOS version of OIIO is compiled with OpenJpeg2000 support?
I can not reproduce the bug on Linux.
The IK Bone influence is driver from another bone
Guess it's more matter of question whether object visibility affects on "visibility for camera" or "visibility for everything". This patch switched from latter to former one, which could work but is a bit weird. Means, making objects invisible wouldn't really speed up your viewport -- shadow buffers will still be recalculated. But guess it was a design discussion i've missed.
Tue, Sep 19
Well, we need all information we need to reproduce the issue from our side :) That ensures we can fix the issue efficiently.
There shouldn't be a parallel scene update at the same time if one renders with locked interface and/or fullscreen render. If that configuration still crashes, think it worth investigating why.
There is no such a way to get DerivedMesh which is different from ob->derivedFinal. This is probably possible to hack via evaluation flags stored in depsgraph nodes. But keep in mind, all this subjects are being reworked in 2.8, trying to fix them in master would cause same work done twice which we'd better avoid to.
Workaround would be to remove degenerated faces, by, i.e., removing doubles in edit mode.
Does it happen with default cube scene?
Caused by some O(N^2) complexity in tangent space calculation. Will check if it's possible to speed up something.
I do not see any corruption in the file and can't reproduce the issue. It might be that copy-paste things to a new file changes pointers, which could potentially cause different order of selection queries.
Attaching file here, so it's never lost and we can always access it:
If you open attached .blend file, does the selection issue happens in there? (or is it REQUIRED to actually import attached fbx?)
What do you mean by "more stable"? Does the same exact crash happens?
It is hard to tell what's going on with only such knowledge. Some random ideas:
Hard to reproduce, and don't see anything wrong in depsgraph itself (all relations are correct). Suspect some threading conflict when building shrinkwrap target BVH. Could take a while to investigate :(
Don't think it is anything to do with FBX, sounds more like OpenGL selection issue. Things to try:
@Campbell Barton (campbellbarton), this seems to be a closed manifold, without duplicate verts, Mind having a look here? :)
Mon, Sep 18
@Maeldor (Maeldor), yes, should be bmw27_gpu. And you're right about --factory-startup. Let's do this way then: backup your settings, load factory startup, enable CUDA and then try rendering from terminal.
Fixed in rPace2787, but still need to deploy to the server. Should happen later this week.
@Stefan Eisenreich (Stef1309), we can not guarantee Blender to work on overclocked GPUs. There is a good reason why vendors didn't use higher frequencies to begin with. Fact that games are stable doesn't really mean much here, CUDA program will stress GPU much more than OpenGL/DirectX.
This file is missing textures in it, so it might affect on reproducability of the bug. Besides that:
Afaik, Solaris is pretty much discontinued by Oracle now, almost everyone from that team was fired. Think we can indeed remove checks for it in source/ and intern/. But be careful about extern/ because some of the libraries (like Glog) might use special ifdefs coming from upstream.
@Campbell Barton (campbellbarton), eh, here goes my brain dump, sorry :)
Sat, Sep 16
If it never worked, i'd rather not invest time making it work in 2.7x series but focus on making a good outlines in 2.8 branch (outlines there really needs love and reconsideration..)
I don't think CVE is a reason to drop this format. Just fix the bug and move forward. Crashes, vulnerabilities and known CVEs are happening in lots of other libraries we use, including png, ffmpeg, oiio, whatever.
Fri, Sep 15
OpenGL is having hard time with movies and image sequences because currently OpenGL materials does not have a proper image user to store/update frame for image access. This is a known TODO, and it'll be more efficient to address this in new viewport of Blender2.8.
Can you maybe save .blend file with all the referenced textures etc, pack them to an archive? But also weird why importing doesn't happen correctly. Perhaps worth separating two reports:
Carver solver does not clear vertex group.
In fact, bmesh is behaving correct here. The unfortunate thing is that we can not interpolate such attributes: what if one vertex of edge was in groups A,B,C,D and second vertex of edge was in groups A,D what groups midpoint of edge should belong to? Or should those groups be copied from vertices of second object? Not only it's not trivial to do currently, but it's still ambiguous because new vertex might be created by a midpoint of edge from second operand. So at the end of a day there is no solution to this problem, so we just don't put new vertices to any of groups.
Thanks for the report, but this is a limitation of re-do system, combined with context-based operators. Basically, when you tweak values in redo panel the operator gets re-run in the new context.
@Sebastian Koenig (sebastian_k), thanks for the file, but it has same exact behavior in 2.78c, 2.79 and current master, while report says it worked in 2.78c..
Thanks for the report, but it' not really a bug.
Can you attach an example of corrupted image? I would like to see if we can make Blender to identify such images and prevent crashes (rendering magenta-color as a missing texture for corrupted image is more friendly than crashing)
Made a mistake in original patch which broke quad faces. Fixed in rB95a8e73d400.
Thu, Sep 14
Can not reproduce any crashes, also crash is in UI code somehow.
Thanks for cleaning up the description :)
- Never inline logs and backtraces into comments and reports, attach them as files. This ensures reports is easy to follow.
- Follow bug report guidelines, and provide .blend file and shortest list of steps possible to reproduce the issue.
@Joerg Mueller (nexyon), all buildbots should be up, and both 2.79 and latest buildbots are supposed to have sndfile enabled. So think the issue can be considered fully resolved now.
So, am i right the issue here is that transparent objects does not have selection outline? (but selection itself works)
Phew, now see what's happening.
@LazyDodo (LazyDodo), so it's about presets for bounces and such?
Light Path in context is Cycles is a shader node, which has some boolean flags what the ray have bounced from. It has no way to configure anything.
Please follow bug report guidelines and fill in all requested information. Most crucially, .blend file which demonstrates the issue.
@gaia, mind checking what's going on with Collada here?
@Antonio Vazquez (antoniov) might also want to have a look here.
Thanks for the report, but it's more a feature request than a bug which we don't accept in the bug tracker. Such requests better be addressed to bf-funboard (general functional discussion board) or bf-gamedev (BGE related discussions) mailing lists.