# 23_vpx_packetizer_libwebrtc_interop.patch (parked — no-op)
#
# Empirical result on 3 June 2026: pjsip 2.12 worked end-to-end with
# Sylk Mobile / libwebrtc on VP9 using the *minimal* 1-byte VP9
# descriptor (just the B/E start/end flags).  Adding an RFC-9628-style
# 4-byte non-flexible single-layer descriptor with PictureID +
# TID/SID/D + TL0PICIDX — which is what every WebRTC interop guide
# suggests — actually breaks the libwebrtc receiver: Sylk's display
# stays black even when the encoder is producing >25 fps with healthy
# bitrate (verified via 44_vpx_encoder_rate_control.patch + pcap).
#
# Best guess: libwebrtc's VP9 depacketizer treats every header field
# as a tight contract, and one of TID/SID/D/U-bit/PID/TL0PICIDX
# semantics is interpreted differently between pjmedia and libwebrtc.
# Without instrumented logging inside libwebrtc we can't isolate
# which field is the disagreement — and since the 1-byte minimal
# descriptor works fine for the same peer, the right move is to
# stop trying to "improve" it.
#
# Earlier in-tree iterations of this slot
# ---------------------------------------
#   * PID-only (2-byte desc):     decoded 1 keyframe then froze.
#   * Full I+L+TL0PICIDX (4-byte): Sylk display stays black even at
#                                  25 fps wire rate.
#   * Vanilla 2.17 (1-byte B/E):   Sylk display works; framerate
#                                  bottlenecked by encoder until
#                                  44_vpx_encoder_rate_control.patch
#                                  lands.
#
# Conclusion: 2.17 upstream VP9 packetization is already correct for
# libwebrtc interop.  This patch slot stays empty unless and until
# a measured improvement is found.  44_vpx_encoder_rate_control.patch
# is the one that actually fixes the user-visible "Sylk sees nothing"
# symptom — by making the encoder produce frames at all.
#
# This file is intentionally empty of diff content — get_dependencies.sh
# skips patches with no `---` header, so this slot is a no-op in the
# 2.17 series.