https://bugs.kde.org/show_bug.cgi?id=385820
Bug ID: 385820
Summary: digiKam crashes when enfuse writes TIFFs
Product: digikam
Version: 5.4.0
Platform: Other
OS: Linux
Status: UNCONFIRMED
Severity: normal
Priority: NOR
Component: FilesIO-Engine
Assignee: [email protected]
Reporter: [email protected]
Target Milestone: ---
Created attachment 108381
--> https://bugs.kde.org/attachment.cgi?id=108381&action=edit
gdb-in-script log of digikam crash at point of enfuse writing TIFF
This is a typical workflow:
1) run digiKam, do various minor organizational tasks
2) concurrently, in a terminal, run a script that invokes align_image_stack and
then enfuse on the results (effects HDR / superresolution etc). This runs 4
processes simultaneously using parallel(1).
3) The files involved are 16-bit TIFF, mostly in BestRGB or ProPhotoRGB-linear
space, nominally around 350-400MiB, because quality.
Frequently, digiKam crashes just at the point where an enfuse process is
writing a file to a directory digiKam is watching. In this case "frequently"
means a MTBF of the order a handful of minutes, rendering write- or
edit-operations infeasible for mortal fear of losing data.
This has happened for a couple of years using both Arch and Ubuntu linuxes,
using BTRFS, XFS and ZFS filesystems.
I've run digikam through gdb in order to ascertain where it might be happening.
Typescript / gdb trace attached.
At an inspired guess, I'd say digiKam is jumping the gun and scanning the TIFF
prematurely, libtiff doesn't like what it sees and blows up so the parent
process has no choice but to follow suit. If so, a simple delay of a few
seconds while the file is completely written to disk would not harm.
--
You are receiving this mail because:
You are watching all bug changes.
