Basic integrity check: reject the DPX stream if 1 of the file is too small

[JeromeMartinez]

One of our beta-testers has a stream with one broken file (file is cut, file transfer error), leading to a crash.
Fixing crash + basic integrity check of the file size (content rejected & file name of the small file is provided).

[kieranjol]

This is a great check to make.another common error is a file in the sequence is missing.

[JeromeMartinez]

This is a great check to make.another common error is a file in the sequence is missing.

In that case RAWcooked creates 2 video streams (e.g. files 0-1000, and files 1002-2000, if 1001 is missing), a bit weird for playback but reversibility is OK.
Wondering if I should reject the stream instead, would it make more sense?
We are moving to a DPX conformance checker here ;-).

[retokromer]

There are real world situation in which missing numbers are correct.

[retokromer]

E.g. for at least two archives, when we digitise very badly decomposed materials, we have so skip the numbers of the frames we could not digitise.

[JeromeMartinez]

There are real world situation in which missing numbers are correct.

Curious about the reason (edit: you explained when I was writing :) )

Wondering if it is a use case from our sponsors, or a priority for them, to have a single video stream for that kind of directory.
Note that the current command line usage for FFmpeg would not like it, I would like to use a text file for the list of files. More complicated (especially on time stamping), maybe not the the priority and was not in my mind when I estimated development time.

[retokromer]

Another case that comes in mind is: When a film has two very different stock materials (e.g. reality on reversal and nightmare on negative-positive) which are decomposed differently. The «raw scan» will consist of two successive scans (at least with different settings, and possibly made on two different scanners, and possibly even with a chemical treatment in between the two digitisations). In this case some archives wish to keep the two scans distinct, but by numbering that way the frames, it can be easily imported in the correct order into a timeline for restoration, post-production, grading or whatever.

I would indeed suggest to test the continuity of the numbers. An additional flag could be set to either give an error message (default) or make different streams, when there are missing or «missing» frames. Would this be difficult to implement?

[kieranjol]

I think Retos example sounds like a valid reason to keep the existing behaviour and not including the sequence check like I'd mentioned. When I had encountered the skipped frames,it was always unintentional and the result of some sort of error.

[JeromeMartinez]

An additional flag could be set to either give an error message (default) or make different streams, when there are missing or «missing» frames. Would this be difficult to implement?

Right now it is a bit complicated because directories traversal and sequences detection are not at the same place (not good first design), but it will be easy when I change this code (I don't like the slowness of the current implementation), I add that to the ToDo-list.
#155: Reto idea
#156: "perfect" handling (single stream with right timestamps)