Bug: "Freeze" filter causes crash

[Efenstor]

trafficstars

Please include the following information.

1. Flowblade version (Help->About): v2.0.0
2. MLT version (Help->Runtime Environment): 6.12.0
3. Your distribution (Ubuntu, Debian, Mint etc.): Debian 10 (Buster) with Gnome 3.30
4. If possible, try to provide step-by-step instructions on how to reproduce the issue. From developer point of view this is the easieast way to understand what is happening.:

1. Import the sample file: https://yadi.sk/i/m6Up0op2xyRw-w
2. Double-click it and select the area between 00:00:12:00 and 00:00:20:00 (approximately)
3. Drag and drop it into the timeline
4. Right click > Add Filter > Movement > Freeze
5. Crash!

Sometimes it allows past the filter addition but still eventually either crashes very soon, especially if I move the cursor to the end of the timeline, or the preview area is just black, or the interface freezes, or some other abnormal behavior happens, obviously caused by memory corruption.

Important: everything goes fine if I place the IN marker at 00:00:00:00, only when it is placed at some arbitrary position it causes a crash. Has it something to do with H264 keyframes?

Messages fed into the terminal vary:

1. Segfault
2. Segmentation fault
3. "Assertion src->tf.f == src->f failed at src/libavcodec/h264_picture.c:72" followed by "Aborted"
4. _python: ../nptl/pthread_mutex_lock.c:81: __pthread_mutex_lock: Assertion `mutex->__data._owner == 0' failed.

Or may be it is caused by the MLT version incompatibility, because I've noticed these messages appear in the terminal on startup:

Brightness dropped, MLT version too high for this filter.
Brightness Keyframed dropped, MLT version too high for this filter.
Loading transitions...
RGB Adjustment dropped for Color Adjustment
Hue dropped for Color Adjustment
Gamma dropped for Lift Gain Gamma

[jliljebl]

Confirmed. Working on this will get postponed until after Python3 port but we will do work on this. Fixes may not be easy to find but who knows, maybe we get lucky. Freeze filters has always been one of the unstable ones.

The messages are just info on which filters are used based on MLT version and service availability, no connection to this issue.