Darktable 5.2.1 does not start [Solved]

[JankoK (Janko Kernc)]

I have darktable 5.2.1 installed from source on a Linux machine (Mint 22.1, AMD Ryzen 9 7950X, AMD Ryzen 9 7950X, 32GB) and suddenly within 24h (after some updates in the update manager, but no change to darktable installation) darktable won’t start any more.
When started from the terminal, it shows the error:
./darktable: symbol lookup error: /lib/x86_64-linux-gnu/libRusticlOpenCL.so.1: undefined symbol: LLVMInitializeAArch64Target, version LLVM_19.1
I don’t have any idea where to start troubleshooting.

Ironically, on this same machine, the experimental AgX version (appimage 5.3.0+169~g06f0b0b576) runs without problems …

[g-man]

darktable --disable-opencl

[kofa (István Kovács)]

The agx version is not based on 5.2.1, but on the current development version, and may include changes that fix the problem.

[JankoK (Janko Kernc)]

Yes, this enables darktable to start, but its processing is slow without opencl. And this does not explain, why it is suddenly causing problems.
And I just compiled the latest master version (f29de89), and it is causing the same problem …

[g-man]

I first wanted to ensure it was isolated to opencl.

Now, provide the output in a txt file of darktable-cltest.

edit: can you confirm that opencl works in the app image agx version?

[JankoK (Janko Kernc)]

Yes, here’s the output attached.
But the darktable-cltest also ends with this error:
./darktable-cltest: symbol lookup error: /lib/x86_64-linux-gnu/libRusticlOpenCL.so.1: undefined symbol: LLVMInitializeAArch64Target, version LLVM_19.1
darktable-cltest.txt (5.5 KB)

And yes, judging by the export speed and processor load difference between openCL enabled and disabled, on the AgX version, OpenCL is working …

[g-man]

Can you confirm that opencl works in the app image agx version via -d common?

[g-man]

It shows multiple opencl devices. How many do you physically have in your PC?

[JankoK (Janko Kernc)]

Yes, I noticed the multiple devices. The first two are probably graphics integrated on the motherboard, but they are not used, the third (graphics card AMD ATI Radeon RX 7900 XT) however is used.

[g-man]

Can you bisect to Some OpenCL internals are gboolean and alloc safety by jenshannoschwalm · Pull Request #19137 · darktable-org/darktable · GitHub ?

[JankoK (Janko Kernc)]

Here’s the -d common output from the AgX version:
darktable-appimage.txt (7.3 KB)
Also, the following is showing after the terminal start of the appimage:
./Darktable-5.3.0+169~g06f0b0b576-x86_64.AppImage -d common > darktable-appimage.txt
Setting $HOME to /home/janko/DT_appimage/Darktable-5.3.0+169~g06f0b0b576-x86_64.AppImage.home
Gtk-Message: 23:18:36.125: Failed to load module “xapp-gtk3-module”
/usr/lib/x86_64-linux-gnu/gvfs/libgvfscommon.so: undefined symbol: g_task_set_static_name
Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libgvfsdbus.so
/usr/lib/x86_64-linux-gnu/gvfs/libgvfscommon.so: undefined symbol: g_task_set_static_name
Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libgioremote-volume-monitor.so
/usr/lib/x86_64-linux-gnu/gvfs/libgvfscommon.so: undefined symbol: g_task_set_static_name
Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libgvfsdbus.so
/usr/lib/x86_64-linux-gnu/gvfs/libgvfscommon.so: undefined symbol: g_task_set_static_name
Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libgioremote-volume-monitor.so
/usr/lib/x86_64-linux-gnu/gio/modules/libdconfsettings.so: undefined symbol: g_assertion_message_cmpint
Failed to load module: /usr/lib/x86_64-linux-gnu/gio/modules/libdconfsettings.so

[JankoK (Janko Kernc)]

Sorry, I’m not a software engineer, and I do not understand most of the contents of this conversation …

[g-man]

The agx is not using rusticl. I think you should disable the cards you are not using via darktablerc.

I going to assume that an update to your OS is making opencl kernel fail.

[rvietor]

In the original post, you talk about a locally compiled version (“installed from source”) and problems after an update. The problems look very much like a mismatch for one or more libraries between the version used for compilation and the current version.

I’d suggest a recompile first.

[JankoK (Janko Kernc)]

No, that must be a misunderstanding, so here’s a summary to clarify things:

1. darktable has been compiled from source a while ago, and the compilation had worked for quite some time (maybe two weeks) without problems.

2. A day or two ago, without touching darktable, my original compilation suddenly stopped working with those symptoms mentioned in my original post.

3. As one of the first attempts art solving the problem I recompiled the 5.2.1 version - compilation successful, but the same problem.

4. After the suggestions of @g-man and @kofa , I also compiled darktable directly from master branch, also compilation successful, the same problem.

5. Alongside those compilations I also have the appimage (AgX test version 5.3.0+169~g06f0b0b576) which is working all the time without problems.

[kofa (István Kovács)]

There was this recent thread, where duplicate devices were reported:

Understand List of OpenCL Devices

Hi, usually I’m not the first person having a question. However, this time I just don’t find what I’m looking for among the existing answers, so I would appreciate your help. I use the culling mode quite often to compare similar images, often zooming to 100% to check sharpness. I would like to make sure that this is possible without too much waiting time and therefor optimize the settings in the darktablerc file accordingly. I’ve one GPU and a CPU with integrated graphics. However, DT shows fo…

There were suggestions regarding multiple drivers for the card, as well as starting with a clean darktablerc.

[JankoK (Janko Kernc)]

I tried disabling (i. e. deleting the corresponding lines in darktablerc) all three devices, one at a time, and the result is still the same …

[kofa (István Kovács)]

I think they get re-detected and re-added. However, it’s possible to disable a device by editing the line that belongs to it. I’m travelling, please read the manual. One of the numbers on the cldevice lines is a flag to disable the device.

Edit: it’s position 8 (‘h’). See

https://darktable-org.github.io/dtdocs/en/special-topics/mem-performance/#device-specific-opencl-configuration

h. disable device
0 = enable device; 1 = disable device
If darktable detects a malfunctioning device it will automatically mark it as such by setting this parameter to 1. If you have a device that reports a lot of errors you can manually disable it by setting this field to 1. If darktable has disabled the device but you are sure it should be used you can re-enable it by setting this field to 0.

[JankoK (Janko Kernc)]

Thank you for your information. As the first thing I tried deleting the darktablerc file, and success! darktable (both the 5.2.1 and master branch, actually) start without problems.

I only have to re-import all my files, or, is there a way to reconnect the database (existing) to the newly created darktablerc?

[kofa (István Kovács)]

The database is independent of configuration files.