Support RFC 6364 and IS-05 SMPTE 2022-5 FEC SDP (alternative to #505)

[garethsb]

Summary

Alternative to #505, based on @thangldw's fix/is05-fec plus @garethsb's thangldw/nmos-cpp#1 trailing-field simplification. Rebased onto master after #514.

• RFC 6364 SDP grammar for FEC/UDP, UDP/FEC, FEC-FR, fec-source-flow, fec-repair-flow, repair-window
• optional trailing object_converter field so m= format lists may be omitted
• IS-05-compatible SMPTE 2022-5 mapping in nmos/sdp*
• require resolved fec_mode (1D/2D) when creating SDP; parse concrete mode from repair-flow count
• resolve receiver staged fec_mode: auto from /constraints (highest enum value, or 2D when unconstrained)

Commits (this PR vs master)

1. Implement RFC 6364 SDP grammar (@thangldw)
2. Support IS-05 SMPTE 2022-5 FEC SDP (@thangldw / @garethsb)

Testing

Linux Release: all testSdp* / *Fec* (681 assertions in 37 cases).

Related to #38.

[garethsb]

@lo-simon, this #515 is fundamentally @thangldw's nice work from #505 rebased onto #512 and #514, with:

• simpler digit/trailing-field grammar
• stricter create/parse/auto handling aligned with resolved /active
• cleaner constructors

This PR still needs a big review; there's a fair bit of complexity in the FEC SDP handling and conversion to/from IS-05 transport_params.

[lo-simon]

For RFC6364, the tag-len and repair-window size require %x31-39 *DIGIT (with a non-zero first digit). I created a non_zero_digits_converter to replace the digits_converter in the fec_source_flow_converter and repair_window_converter. Additionally, I extended the unit tests accordingly. Please see the attached patch file.
enhance-sdp-grammer-for-non-zero-integer.patch

[garethsb]

Thanks @lo-simon. I don't want non_zero_digits_converter.

tag-len and repair-window size are %x31-39 *DIGIT, so the ABNF does ban zero and a leading zero, but that check does not belong in this grammar. digits_converter accepts both, as it does for the other 1*DIGIT fields, and the comment there says leading zeros are accepted and not round-tripped. The grammar turns text into JSON; limits such as encoding-id 0–255 stay above it.

Refusing tag-len=0 or repair-window:0ms in the parser means the session description never becomes JSON. Postel's law applies on the way in. A check that this node will not advertise a zero tag length or repair window belongs with the other IS-05 rules, not in the SDP parser.

So I have not taken the converter, or the tag-len=0 and repair-window:0ms rows. I also have not added tests that 0 and 01 parse; that would make non-conforming SDP look supported. I have taken the other tests from your patch — the fssi syntax failures and the IS-05 mapping cases — and pushed them with you as author.

[lo-simon]

LGTM