OpenCL (NVidia) crash with two masked agx instances

Is there an existing issue for this?

• I checked and did not find my issue in the already reported ones

Describe the bug

I was processing an image with very high dynamic range and tried applying 2 masked instances as a weird experiment (the 1st parametric on the g channel, the 2nd using an inverted raster mask from the 1st). After a while, while adjusting parameters, darktable crashed.
pixelpipe_hb.c -> opencl.c -> libnvidia-opencl.so.1 -> <signal handler called> () at /lib/x86_64-linux-gnu/libc.so.6

Steps to reproduce

Not clear. Crash only happened after a while.

Expected behavior

No response

Logfile | Screenshot | Screencast

last-few-seconds-of-console-output.txt
darktable_bt_BU52G3.txt

DSC_0856-with-2-agx-instances.NEF.xmp.txt

DSC_0856.NEF is in this post: https://discuss.pixls.us/t/filmic-when-to-use/13250 (the backlit ladies)

Commit

No response

Where did you obtain darktable from?

darktable.org / GitHub release

darktable version

5.3.0+950~g8f05fbc64b

What OS are you using?

Linux

What is the version of your OS?

Ubuntu 25.10

Describe your system

Wayland

Are you using OpenCL GPU in darktable?

None

If yes, what is the GPU card and driver?

NVidia 1060/6GB with Nvidia 'server' driver 580.95.05 (the regular driver does not support the 1060).

Please provide additional context if applicable. You can attach files too, but might need to rename to .txt or .zip

No response

Start by reviewing the attached darktable_bt_BU52G3.txt, last-few-seconds-of-console-output.txt, and DSC_0856-with-2-agx-instances.NEF.xmp.txt, then trace the reported pixelpipe_hb.c and opencl.c paths. Try to narrow the crash using the two masked instances and the listed NVIDIA driver environment. Done means establishing a reproducible failure or a supported diagnosis from the supplied evidence.