Camera blocks after using the download script about python-gphoto2

[jim-easterbrook]

gp_camera_free deletes the C camera object (frees its memory) and should never be used according to the libgphoto2 documentation. gp_camera_unref decrements the object's reference count, freeing it when the count reaches zero. This is not user-callable from Python as it would leave a Python object pointing to an invalid C object and cause segmentation faults. Instead it is called automatically when the Python object is destroyed.

gp_camera_exit makes the camera available to other programs. It probably isn't needed as the example script quits anyway. It's good practice to include it though.

Your problem sounds like a bug in libgphoto2 or the camera. What happens if you use the gphoto2 command to download photos? A google search suggests that "busy" problems are quite common, but I didn't see any obvious fix.

[bitcraft]

@dmitryelj are you shooting with a SD card installed?

[dmitryelj]

Thanks for the answer. Sure, the SD card is installed.

I'll try to download and build the last libgphoto2 version, maybe it will help.

[bitcraft]

I had an issue where the camera would get stuck at busy if I was shooting without a SD card. I have a reliable one installed now and haven't run into any issues despite working it extremely hard using a photo booth. I'll link my code, maybe you can find somthing we are doing differently?

https://github.com/bitcraft/tailor/blob/master/tailor/plugins/gphoto2_camera.py#L90-L112

If you ignore the asyncio bits, its a pretty straightforward function. Works like a charm with my Canon 6D.

[jim-easterbrook]

Another user has had problems (unspecified) when using a USB3 port. Using a USB2 port might work better, assuming a USB2 camera.

[dmitryelj]

Thanks for the answer.

Its weird in general.
I'm running this script to download photos on the Raspberry Pi:

    camera = gp.check_result(gp.gp_camera_new()) 
    error = gp.check_result(gp.gp_camera_init(camera))
    if error != gp.GP_OK:
        print "Cannot connect camera, error", error
        sys.exit(0)

    camera_files = list_camera_files(camera)
    print camera_files

    out_path = "/home/pi/Documents/photos"
    for path in camera_files:
        folder, name = os.path.split(path)
        info = gp.check_result(gp.gp_camera_file_get_info(camera, folder, name))
        camera_file = gp.check_result(gp.gp_camera_file_get(camera, folder, name, gp.GP_FILE_TYPE_NORMAL))

        dest = out_path + '/' + name
        gp.check_result(gp.gp_file_save(camera_file, dest))

    gp.check_result(gp.gp_camera_exit(camera))

It works like a charm, but - sometimes with about 25% probability camera cannot make pictures more after running the script, indicating "busy" when pressing shutter. Then I cannot access the camera through python-gphoto, but cannot take more pictures.
But then, if I run ./sample-autodetect from libgphoto2 sources, the camera became magically unlocked. Thats why I was asking if all resources in the python lib are released correctly.

I builded the last version of libgphoto2 from source, but it didn't help, the symptoms are the same.

[jim-easterbrook]

That really is very strange. It appears the Python script is not always leaving the camera in the correct state, but I have no idea why. There's nothing asynchronous or threaded in your program so it should be deterministic. I wonder if adding a second or two delay between calling gp_camera_exit and quitting the program would help.

It might be worth delving into the source of ./sample-autodetect to see what libgphoto2 calls it's making. In theory you should be able to make the same calls from Python to unlock your camera.

[dmitryelj]

This code is doing nothing special:
https://github.com/gphoto/libgphoto2/blob/master/examples/sample-autodetect.c

I tried to make a call of gp.get_config_value_string from python but this method is missing.

[jim-easterbrook]

get_config_value_string isn't a libgphoto2 function, it's defined in config.c
https://github.com/gphoto/libgphoto2/blob/master/examples/config.c#L23

[dmitryelj]

Thanks.

[dmitryelj]

Another small update.
I reduced the code to only init and exit, and the camera is still "busy" sometimes. Have no idea. Looks like its not a "python-gphoto" issue anyway. But freeing resources issue in Python still exist. When I run C-app at the second time, it finally works and the camera became unlocked, and with the Python app is not.

    camera = gp.check_result(gp.gp_camera_new()) 
    error = gp.check_result(gp.gp_camera_init(camera))
    print "gp_camera_init:", error
    if error != gp.GP_OK:
        print "Cannot connect camera, error", error
        sys.exit(0)

    # This time I'm trying to shoot with the camera
    time.sleep(60)

    gp.check_result(gp.gp_camera_exit(camera))

[jim-easterbrook]

Did you reinstall python-gphoto2 after building and installing the latest libgphoto2? The Python interface might still be using the old version. Use ldd on one of the compiled Python modules (e.g. /usr/lib64/python3.4/site-packages/gphoto2/_camera.cpython-34m.so - the name will be system dependent) to check.

[dmitryelj]

The error gone when I switched to "OOP" interface:

        context = gp.Context()
        camera = gp.Camera()
        camera.init(context)
        ...
        camera.exit(context)

I don't know what it was, it looks, the ticket can be closed.

Thanks again for help.

[jim-easterbrook]

One difference is using a context. I thought the context was optional (whether OOP or not) but you may have found a case where it isn't.