I'm currently batch-editing a couple of older raw-pictures, and I consequently get a segfault when trying to finish the queue. This is what I get if I run a debug build through gdb:
https://pastebin.com/i0gqT68y
Sorry if I got the subject wrong, but that's what gdb spat out, and the queue finishes fine if I turn "Noise Reduction" completely off.
Version: 5.3-253-g58c9f5a4
Branch: dev
Commit: 58c9f5a4
Commit date: 2017-12-06
Compiler: cc 7.2.0
Processor: x86_64
System: Linux
Bit depth: 64 bits
Gtkmm: V3.22.2
Lensfun: V0.3.2.0
Build type: release
Build flags: -std=c++11 -march=native -Werror=unused-label -fopenmp -Werror=unknown-pragmas -Wall -Wno-unused-result -Wno-deprecated-declarations -O3 -DNDEBUG -ftree-vectorize
Link flags: -march=native
OpenMP support: ON
MMAP support: ON
a couple of older raw-pictures
with odd width?
Sorry Ingo, they are all raw files from Nikon D700. I had to google the dimensions, but no, they are even-numbered :-)
Do they all crash?
Can you post a pp3 please?
Sure! Can you access this one?
https://www.dropbox.com/sh/6hlvoiqx7aaksqb/AACtNFPYMgNb-ZNERV3OCh3Pa?dl=0
yes, but I had to edit your post to access them with a single click ;-)
Can you try to rule out other tools (i.e. profiled lens vignetting corrections) as the reason?
@mks9900 Johan, not related to this issue: We ship a NIKON D700.dcp. @Beep6581 made it from colorchecker shots he made with my D700 some weeks ago. Maybe worth a try ...
Thanks Ingo, but I already thought that profile was selected? I read about your effort and they were greatly appreciated by me :-)
In RT, I chose "Auto-matched camera profile"...
I'll try to get on ruling out other tools now!
Johan, you're right. The dcp is selected by "Auto-matched..."
I can now say for certain it is denoise that causes the crash, but not with every file. I uploaded the troublesome file and pp3 and if I toggle denoise on/off I get crash/no crash repeatedly...
Dropbox link, done correctly :-)
Thanks Johan. I will try to find the culprit
Johan, though it crashes in denoise, the culprit is not denoise.
Can you confirm that it does not crash when highlight reconstruction is disabled?
Yes, I can certainly confirm that! Wow, nice work Ingo!
However, for me,it seems it only crashes if method=Blend, does it crash there for you too?
yes
Simpler test case, apply neutral then this PP3:
https://filebin.net/ethqfut80kndq62j/rt_4214_crash.pp3
-- Edit
Note: the [Color Management] section shouldn't be in that PP3 - #4216
And the stack backtrace:
https://paste2.org/zaVx83Ma
Fixed in d7fb8c9
Most helpful comment
Fixed in d7fb8c9