Changes between Version 3 and Version 4 of Ticket #941, comment 4


Ignore:
Timestamp:
Jan 28, 2012, 12:20:44 AM (5 years ago)
Author:
DonMoir
Comment:

Legend:

Unmodified
Added
Removed
Modified
  • Ticket #941, comment 4

    v3 v4  
    11Just to restate this. Early seek times are going to succeed because there are some initial AVStream.index_entries in place at start up. It's all about when you try to seek past the last entry in the index_entries table which have not been created but should be in the ogg_read_seek function. So the seek failure is all about ogg_read_seek's failure to create the index_entries. If I go ahead and make sure the index_entries are created prior to calling avformat_seek_file, then seeking works fine outside of the additional clean up notes mentioned in item 2) above. Also I remap my timestamp to the one returned by av_index_search_timestamp so I make sure I know I am at the place I need to be without second quessing. 
    22 
    3 Attempting to seek inbetween one of these index_entries could possibly be the reason an incomplete frame is being produced for ticket #438 but there is probably more to this. With my current code setup I don't see any incomplete frames, but there is still some other odd behavior with some ogg files. 
     3Attempting to seek inbetween one of these index_entries could possibly be the reason an incomplete frame is being produced for ticket #438 but there is probably more to this. With my own current code setup I don't see any incomplete frames, but there is still some other odd behavior with some ogg files. 
    44 
    55Most of the decoders either create these index_entries in their read_seek function or they don't need to because they may be stored in the file. Without these index_entries seeking is going to be bad for ogg files.