System Information
Operating system: Ubuntu 18.04
Graphics card: GTX 1080
Blender Version
Broken: 2.80 aca13040dac
Worked: 2.79b
This used to a be a thing done to see polygon lines in the UV image editor and track the progress of the bake with the uv lines overlay - Cycles doesn't show the progress per triangle, but baking is still allowed in edit mode and leads to a crash - probably due to the multiediting change?
Sure!
Select the high poly, select the low poly, go into edit mode and hit bake - Blender crashes instantly.
Hm, I can't reproduce it, maybe I'm doing something wrong. Is it possible for you to change the .blend file so that you only have to press Bake to make it crash?
Hit TAB twice and then bake and it'll crash - if you just open it when it was saved with both objects selected and in edit mode it will bake out ok and not crash, if you do TAB and TAB again to go out of edit mode into object mode, and into edit mode again and hit bake, it'll crash
I've also recorded a video:
https://www.youtube.com/watch?v=ua4KNyf9ziI
I've been getting the same behaviour since 2,8 first came out - is there some bug log I can send you to help if you can't reproduce it?
ok, I'm doing exactly the same thing you are doing in the video, even in the exact same version and still can't reproduce it.
Waiting for someone else to try it now.
Are you also testing this on Linux?
Running Ubuntu 18.04 with Kernel 4.16.10-041610-generic
Nvidia driver version 396.54
The CPU is an Intel i7-8700K
If it's nothing OS or Hardware specific maybe it's some old setting that's being carried over for me from a previous version?
Here's a crash log:
I can reproduce this on f85b05111fa4b96228a4a3969945c80fbb67f499
I get asserts in my debug build:
blender: /home/zed/programmering/blender_master/blender/intern/cycles/kernel/../kernel/kernel_compat_cpu.h:77: const T& ccl::texture<T>::fetch(int) [with T = ccl::uint4]: Assertion `index >= 0 && index < width' failed.
Backtrace from non debug build:
Thread 45 "blender" received signal SIGSEGV, Segmentation fault.
[Switching to Thread 0x7fffb63ff700 (LWP 12359)]
0x0000555556f83fc2 in ccl::kernel_bake_evaluate(ccl::KernelGlobals*, ccl::uint4*, ccl::float4*, ccl::ShaderEvalType, int, int, int, int) ()
(gdb) bt
#0 0x0000555556f83fc2 in ccl::kernel_bake_evaluate(ccl::KernelGlobals*, ccl::uint4*, ccl::float4*, ccl::ShaderEvalType, int, int, int, int) ()
#1 0x0000555556d72127 in ccl::CPUDevice::thread_shader(ccl::DeviceTask&) ()
#2 0x00005555570ee563 in ccl::TaskScheduler::thread_run(int) ()
#3 0x00005555570efbce in ccl::thread::run(void*) ()
#4 0x00007ffff2ef629e in std::execute_native_thread_routine (__p=0x7fffcc84e360)
at /var/tmp/portage/sys-devel/gcc-8.2.0-r5/work/gcc-8.2.0/libstdc++-v3/src/c++11/thread.cc:80
#5 0x00007ffff5d9b3c3 in start_thread () from /lib64/libpthread.so.0
#6 0x00007ffff329936f in clone () from /lib64/libc.so.6@Brecht Van Lommel (brecht) should I assign cycles issues to someone else than you?
Hey @Sergey Sharybin (sergey) I only get a crash if I build with WITH_OPENSUBDIV.
And the fact that it only happens if we toggle in/out of editmode raises a few flags that the issue is not on baking necessarily.
(and @Campbell Barton (campbellbarton) - unrelated to the crash, but related to how the user is using edit mesh - this is another interesting use-case for having an option to disable multi-object editing, otherwise the HP and LP objects would necessarily be in edit mode, and both their UVs would show in the editor)