The earlier code checked the PixelCrop* elements as if they applied
after the scaling to DisplayWidth/Height has been done.

Furthermore some checks have been improved (the earlier code didn't
check whether PixelWidth/Height is zero). Moreover a heuristic check
has been added to test whether it has been forgotten to update
DisplayWidth/Height in light of the PixelCrop* elements.

Signed-off-by: Andreas Rheinhardt <[email protected]>

Signed-off-by: Andreas Rheinhardt <[email protected]>

The current level basically ensures that files produced by mkvmerge
produce a warning, although they don't use an excessive amount of void
data. So raise the arbitrary threshold a bit to make this warning
meaningful.

Signed-off-by: Andreas Rheinhardt <[email protected]>

This function seems to have been copied and pasted from
MATROSKA_CueTimeCode, with only minimal modifications performed to make
it do what its name suggest. This means that a variable that never
stores a time is called TimeCode and that the function returns
INVALID_TIMECODE_T instead of INVALID_FILEPOS_T. These things have been
changed.

Signed-off-by: Andreas Rheinhardt <[email protected]>

The differences of MATROSKA_GetNextBlockForTimecode and the classical
MATROSKA_GetBlockForTimecode are threefold:
1. The new function can return both the inner Block as well as the outer
BlockGroup based upon a parameter (that is irrelevant for SimpleBlocks).
2. Consequently the return value is a pointer to an ebml_element and no
longer a matroska_block*.
3. It also allows to get a matching block that is not the first matching
block in a cluster.
This commit is in preparation of a patch for mkvalidator.

Signed-off-by: Andreas Rheinhardt <[email protected]>

The earlier code had several flaws:

1. It was very slow, because finding a matching block always started
with the very first block in the first cluster and checked every block
until an acceptable block (one with the right timestamp from the right
track) had been found.
2. Only one CueTrackPositions (namely the first one) has been checked
per CuePoint, although there can be multiple CueTrackPositions.
3. If e.g. no CueTrack element was present, one got a generic error
(error code 0x200), because a mandatory element was missing; but
MATROSKA_CueTrackNum returned -1 (for invalid TrackNum) and so one also
received a nonsense error 0x312: "CueEntry Track #-1 and timecode xxx ms
not found".
4. The CueClusterPosition and CueRelativePosition elements have not been
checked at all.

All four points have been corrected.
(mkvmerge versions 5.9 to 6.3 (inclusive) wrote CueRelativePosition
elements that (in case of BlockGroups) pointed to the inner blocks and
not the outer BlockGroups. mkvalidator now detects such files as
invalid.)

Signed-off-by: Andreas Rheinhardt <[email protected]>