[FFmpeg-cvslog] mxfdec: Fix regression on files from Pinnacle Thunder
Tomas Härdin
git at videolan.org
Wed Apr 4 18:16:46 CEST 2012
ffmpeg | branch: master | Tomas Härdin <tomas.hardin at codemill.se> | Wed Apr 4 14:24:49 2012 +0200| [a61fada3805122acf8f362d563c9cbae00c2c8cf] | committer: Michael Niedermayer
mxfdec: Fix regression on files from Pinnacle Thunder
The reason for this is that such files have IndexTableSegments which when parsed
cover EditUnit ranges like this:
[0,1)
[249,250)
[249,377)
[0,249)
where each interval is [IndexStartPosition,IndexStartPosition+IndexDuration).
This would be reduced to a sparse index like:
[0,1), [249,250)
instead of the full range:
[0,249), [249,377)
See TimeCode_HD.mxf, UMID =
060a2b340101010101010410130000000004001aa0e59175025b2a5600da4101.
Signed-off-by: Michael Niedermayer <michaelni at gmx.at>
> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=a61fada3805122acf8f362d563c9cbae00c2c8cf
---
libavformat/mxfdec.c | 8 +++++++-
1 files changed, 7 insertions(+), 1 deletions(-)
diff --git a/libavformat/mxfdec.c b/libavformat/mxfdec.c
index 5910ccd..fb47afa 100644
--- a/libavformat/mxfdec.c
+++ b/libavformat/mxfdec.c
@@ -952,6 +952,7 @@ static int mxf_get_sorted_table_segments(MXFContext *mxf, int *nb_sorted_segment
int i, j, nb_segments = 0;
MXFIndexTableSegment **unsorted_segments;
int last_body_sid = -1, last_index_sid = -1, last_index_start = -1;
+ uint64_t last_index_duration = 0;
/* count number of segments, allocate arrays and copy unsorted segments */
for (i = 0; i < mxf->metadata_sets_count; i++)
@@ -974,19 +975,23 @@ static int mxf_get_sorted_table_segments(MXFContext *mxf, int *nb_sorted_segment
/* sort segments by {BodySID, IndexSID, IndexStartPosition}, remove duplicates while we're at it */
for (i = 0; i < nb_segments; i++) {
int best = -1, best_body_sid = -1, best_index_sid = -1, best_index_start = -1;
+ uint64_t best_index_duration = 0;
for (j = 0; j < nb_segments; j++) {
MXFIndexTableSegment *s = unsorted_segments[j];
/* Require larger BosySID, IndexSID or IndexStartPosition then the previous entry. This removes duplicates.
* We want the smallest values for the keys than what we currently have, unless this is the first such entry this time around.
+ * If we come across an entry with the same IndexStartPosition but larger IndexDuration, then we'll prefer it over the one we currently have.
*/
if ((i == 0 || s->body_sid > last_body_sid || s->index_sid > last_index_sid || s->index_start_position > last_index_start) &&
- (best == -1 || s->body_sid < best_body_sid || s->index_sid < best_index_sid || s->index_start_position < best_index_start)) {
+ (best == -1 || s->body_sid < best_body_sid || s->index_sid < best_index_sid || s->index_start_position < best_index_start ||
+ (s->index_start_position == best_index_start && s->index_duration > best_index_duration))) {
best = j;
best_body_sid = s->body_sid;
best_index_sid = s->index_sid;
best_index_start = s->index_start_position;
+ best_index_duration = s->index_duration;
}
}
@@ -998,6 +1003,7 @@ static int mxf_get_sorted_table_segments(MXFContext *mxf, int *nb_sorted_segment
last_body_sid = best_body_sid;
last_index_sid = best_index_sid;
last_index_start = best_index_start;
+ last_index_duration = best_index_duration;
}
av_free(unsorted_segments);
More information about the ffmpeg-cvslog
mailing list