You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Some special cases did break the signing and vaildation
when using FH+TG. In particular three cases:
1. If OBU Metadata is added to I-frames that "SEI" will belong
to the previous GOP, since GOPs are triggered on primary
slices. That itself is not a problem. The problematic part
is the very first GOP. When triggering on the second I-frame
this first GOP looks different from the others due to an
extra OBU Metadata "SEI".
2. Single Frame Headers (FH) without Tile Groups (TG) shall
also be hashed. So far it has been expected to be a TG after
every FH, but a stream can have FHs without TGs.
3. On the signing side, TGs can come in parts, which
complicated FH+TG a little bit.
This commit includes
* 3 new tests
- Sign with data split in parts
- Stream has FHs without TGs
- Scrap first GOP
* 2 new dummy OBUs; 1) a no show FH (FH without TG), and 2)
a TD.
* SEIs cannot become linked hashes on auth side
* Mark the first SEI as 'U' if not a Signed Video SEI and the
current GOP is incomplete.
* Include the show_existing_frame bit when determining frame
type.
* Add a no_hash hash function
* Do not update frame counter on FH
* Do not trigger a new GOP on FH
* Generate SEI if at least one BU was hashed before the first
I-frame.
* Fixes start timestamp if stream starts with SEI
Also bumps the version to v2.2.7
0 commit comments