Mismatch with vtm on slice header extension and entry_point_offset order

[forayr]

In VTM, entry_point_offset are written after slice_header_extension, whereas the spec does the opposite. It seems more logical to write the entry_point_offset at the end of the slice_header.

- if( entry_point_offsets_present_flag && NumEntryPoints > 0 ) {
-   offset_len_minus1
-   for( i = 0; i < NumEntryPoints; i++ )
-     entry_point_offset_minus1[ i ]
- }
if( slice_header_extension_present_flag ) {
  slice_header_extension_length
  for( i = 0; i < slice_header_extension_length; i++)
    slice_header_extension_data_byte[ i ]
}
+ if( entry_point_offsets_present_flag && NumEntryPoints > 0 ) {
+   offset_len_minus1
+   for( i = 0; i < NumEntryPoints; i++ )
+     entry_point_offset_minus1[ i ]
+ }
byte_alignment( )

[ksuehring]

I agree. Entry points need to be added to the slice header after the slices are encoded. It makes more sense to have them located at the end of the slice header.

[bbross]

This seems to be a SW issue so changing it to VTM.

[ksuehring]

Fixed by adoption of JVET-R0298. So this ends up with the software being correct and the standard text changed. So changing component back to spec and closing as fixed.