1. 13 Oct, 2018 8 commits
    • Nicolas Dufresne's avatar
      WIP: h265parse: Don't push NALs before we have HEADERS · cd11a069
      Nicolas Dufresne authored
      Otherwise we may endup pushing incomplete caps.
      This patch isn't ready yet, as it will merge PPS/SPS/VPS in the same
      GstBuffer. This is valid as per current definition of alignment=nal,
      but is also a behaviour change. It's also weird, since it only merge
      these for the first set.
    • Nicolas Dufresne's avatar
      h26xparse: Handle the case we already parsed the frame · ffa80239
      Nicolas Dufresne authored
      As we no longer accumulate 2 NALs before producing any data, we may
      endup with the case where we have completly parsed the frame data and
      then get called on handle_frame(). This was triggering the assertion
      current_off < size, as the current offset was equal to the size.
      Handle this just like we'd do in other case, like broken NAL etc.
    • Nicolas Dufresne's avatar
      h265parse: Differentiate PREFIX SEI from SUFFIX · 13837f3a
      Nicolas Dufresne authored
      There is some code to fixup broken stream that uses the SEI location,
      this code is meant to locate SUFFIX SEI only. This should prevent
      unwanted side effect if SUFFIX SEI is used.
    • Nicolas Dufresne's avatar
      h265parse: Add AUD insertion · f2aed343
      Nicolas Dufresne authored
      This add a feature that existed in h264parse already, which is the
      insertion of AUD NAL whenever these are missing. This simplify the
      detection of frame start and also allow producing valid MPEG TS files.
    • Nicolas Dufresne's avatar
      h265parse: Don't add latency when not needed · a25ba128
      Nicolas Dufresne authored
      We no longer add latency when doing AU->AU, AU->NAL and NAL->NAL
    • Nicolas Dufresne's avatar
      h265parse: Propagate MARKER flag · cecf09d0
      Nicolas Dufresne authored
    • Nicolas Dufresne's avatar
      h265parse: Don't wait for next NAL if input is aligned · 40ad4931
      Nicolas Dufresne authored
      Waiting for the next NAL increases the latency. If alignment=nal/au
      has been negotiated, assumes the the buffer contains a complete
      NAL and don't except a second start-code. This way, nal -> nal,
      au -> au and au -> nal no longer introduce latency.
      As a side effect, the collect_pad() function was not able to poke at the
      following NAL. This call is now moved before processing the NAL, so
      it's looking at the current NAL before it's ingested into the parser
      state in order to dermin if the end of an AU has been reached. The AUD
      injection state as been adapted to support this.
    • Nicolas Dufresne's avatar
      h265parse: Set PTS/DTS and DISCONT on crafted NAL · d3f15f2e
      Nicolas Dufresne authored
      When we inject a NAL in the bitstream before another one, make
      sure to pass both DTS and PTS. Also make sure to transfer the
      DISCONT flag properly.
  2. 11 Sep, 2018 12 commits
  3. 12 Apr, 2018 1 commit
  4. 05 Mar, 2018 2 commits
  5. 02 Mar, 2018 1 commit
  6. 24 Jan, 2018 1 commit
  7. 24 Nov, 2017 1 commit
  8. 22 Nov, 2017 1 commit
  9. 18 May, 2017 1 commit
  10. 12 May, 2017 1 commit
  11. 26 Nov, 2016 1 commit
  12. 23 Nov, 2016 1 commit
  13. 26 Sep, 2016 1 commit
  14. 30 May, 2016 1 commit
  15. 25 Apr, 2016 1 commit
  16. 24 Mar, 2016 1 commit
  17. 17 Mar, 2016 1 commit
  18. 11 Mar, 2016 1 commit
  19. 04 Feb, 2016 1 commit
  20. 25 Jan, 2016 2 commits