Crash in equalizer
A friend has a crash that's preventing him from working on a particular film roll: whenever he tries to open it, darktable crashes while generating the thumbnails. The photos were shot on a Canon 5D mark 2 and he has done some edits on them, it may be that whatever modules he has enabled on one of the images is what is causing the crash. It's really hard to narrow down which image/xmp is the culprit though, as the crash isn't completely reliable.
He gave me a copy of the image directory, and I managed to trigger the crash and get backtraces in both 2.2.5 and master.
#2 Updated by Robert Hutton almost 3 years ago
- File dt-crash-2.3.0-thumbnail_generation-bt_full.txt dt-crash-2.3.0-thumbnail_generation-bt_full.txt added
Tobias Ellinghaus wrote:
Both crashes are in the same place, which is good. Please do "bt full" in gdb to get a little more information. It would also be helpful to have the XMP file for image id 13113.
A "bt full" trace is attached. How do I tell which image has ID 13113? I've switched computers since I created the bug, would the ID be different on this machine?
#4 Updated by Tobias Ellinghaus almost 3 years ago
- System changed from Ubuntu to all
- % Done changed from 0 to 20
- Status changed from New to Triaged
- Category changed from Lighttable to Darkroom
The bt looks slightly different here, but it's also crashing in equalizer. So it's probably the same bug ...
#7 Updated by Robert Hutton almost 3 years ago
- File dt-crash-2.3.0-thumbnail_generation-bt_full_orig.txt dt-crash-2.3.0-thumbnail_generation-bt_full_orig.txt added
Here's the "bt full" from the original machine.
#9 Updated by Robert Hutton almost 3 years ago
Tobias Ellinghaus wrote:
That is yet another backtrace. One I vaguely remember. In any case, all of them are in the equalizer code and in all cases dt seems to try to run on a 1x1 pixel chunk. Maybe hanatos has any clue?
Hrm, I wonder why I'm getting a different backtrace each time... maybe it's because there's multiple images that are having their thumbnails generated, so it just depends which one is getting run first, or maybe there's some other kind of race here. I'll try triggering the crash a few more times to see what I get.
#11 Updated by Cory Larson about 1 month ago
I assumed that the libosmgpsmap docs were consistent in mentioning the release that added new API, but for these things it didn't say anything (resume examples https://resume-examples.com ), so i thought it was safe to use. I'll try to sprinkle a bunch of #ifdef over the code.