Measure Tool Crashing #67
Comments
|
You might be able to get more information by running it through GDB and backtrace the segmentation fault. We can talk on Monday about this, as well as GSoC, etc. |
|
Barry and I looked at this a little more. Unfortunately, we haven't been able to solve it yet. We thought it may be fixed if I used the most updated qt library (5.6 instead of 5.2 -- the version in the Ubuntu repositories). I tried that, though, and it didn't fix it. From adding in debug messages, I can see that the crash is occurring here: https://github.com/OpenChemistry/avogadrolibs/blob/master/avogadro/qtopengl/glwidget.cpp#L108 updateGL() is a function that comes from http://doc.qt.io/qt-5/qglwidget.html, which glwidget inherits. Here is the backtrace from gdb (it appears the crash is in Thread 1): Program received signal SIGSEGV, Segmentation fault. Thread 8 (Thread 0x7fffc69c0700 (LWP 5776)): Thread 6 (Thread 0x7fffc79c2700 (LWP 5774)): Thread 5 (Thread 0x7fffd7526700 (LWP 5772)): Thread 4 (Thread 0x7fffd7f41700 (LWP 5771)): Thread 3 (Thread 0x7fffd8742700 (LWP 5770)): Thread 2 (Thread 0x7fffe2645700 (LWP 5769)): Thread 1 (Thread 0x7ffff7faf800 (LWP 5763)): |
|
I haven't seen this in my testing, I will see if I can reproduce it with a clean build. I would be concerned about the nouveau drivers though, we have seen a number of issues with them. |
|
Yeah, I do have a nouveau driver on that desktop. My graphics card info is here: VGA compatible controller: NVIDIA Corporation GK107 [GeForce GT 740](rev a1) Output from "sudo lshw -c video": "modinfo nouveau" output: |
|
Pat, While I agree with that statement in general, the nouveau drivers are BM
|
|
Unfortunately, I tried a couple of nvidia drivers, and it still crashed in the same way. Here's one of the crash threads: Thread 1 (Thread 0x7ffff7fb1800 (LWP 26941)): |
|
Out of curiosity, are these with Debug builds or not? |
|
I tried several times to reproduce this, and was unable to. I usually use a release build, but had tried with a debug build. Crashing in shader compilation looks very suspicious. Have you seen other instances that we can reproduce on other platforms? |
|
This was with a debug build. I eventually upgraded to Ubuntu 16.04 on my desktop - and I make sure that I use the nvidia proprietary drivers. The issue never went away, unfortunately. It happened for me as well on an Ubuntu 16.04 virtual machine and an Arch Linux virtual machine, but I know that virtual machine tests for this don't matter too much since they don't have direct access to the hardware (and that's probably causing an issue there)... When we release Avogadro 2, if we are able to compile it on a different linux computer and pass around binaries that don't have this crash, then I think this wouldn't be an issue because most users will probably be using the binaries anyways. Perhaps I can try building it on Windows sooner or later as well... |
|
Virtual machines tend to have pretty poor OpenGL support, VMWare is among the best. I am running Arch Linux with NVIDIA binary drivers on my development machine, and Ubuntu 16.04 on my laptop with Intel embedded card. I have tested this on Windows and the measure tool worked well there, I have moved a Windows machine into my office, and want to get it set up soon to build automatically. That is what I used to build the last Avogadro 2 Windows binaries for 0.9.0. If we can come up with a recipe to reproduce I would love to get to the bottom of it, we could still be doing something funky in GLSL that trips some compilers up for example. |
|
I am going to close this for now, but happy to revisit this if you are still having problems. I would like to work on Linux packaging at some point to make this a little easier to diagnose, I use several Linux machines (Arch, Ubuntu 16.04) with Intel and NVIDIA GPUs that I can't reproduce this on, GL issues can be tough to track down sometimes. There could be some issues lurking in the text rendering code, I have seen some oddness on Windows but it seems to work well right now. |
From my compilation of the branch "master" on Ubuntu 14.04 with gcc 4.8.4, the program crashes when I use the "measure tool" to select three atoms. It works fine for selecting two atoms, but it crashes upon clicking the third atom. The crash just says "Segmentation fault (core dumped)"
The text was updated successfully, but these errors were encountered: