fix(sdp): reflect rejected m-lines (port=0) in answer per RFC 8829 §5.3.1

[nightness]

Summary

• Adds rejected / rejected_kind fields to MediaSection (all existing construction sites use ..Default::default() — no breakage).
• populate_sdp now emits a minimal port=0 m-line for each rejected section before the normal transceiver loop.
• generate_matched_sdp pushes a rejected MediaSection instead of silently skipping a section whose direction is Unspecified (port=0 in offer).
• Uses port==0 (not direction attribute) for rejection detection, handling m=application correctly.
• Tracks candidates_added to ensure ICE ufrag/pwd/candidates appear on the first non-rejected m-line.
• Includes unit tests and integration test for rejected m-line scenarios.

Why This Matters (RFC 8829 §5.3.1)

If a remote offer contains an m= section that is rejected (port 0), the
local answer MUST also contain an m= section with port 0. Rejected
sections must stay in the same position to preserve m-line indexing for
subsequent offers.

Without this fix, a remote peer that sends an offer with one or more
rejected m-lines will get an answer whose m-line indices are shifted,
causing subsequent re-offers to mis-map transceivers.

Test Plan

• cargo build -p rtc passes (verified)
• Offer with rejected m-line generates an answer with port=0 in the same position
• Normal offers without rejected m-lines are unaffected
• Unit tests: rejected_audio_mline_has_port_zero, rejected_application_mline_uses_sctp_proto, rejected_mline_not_added_to_bundle
• Integration test: test_sdp_answer_rejects_audio_correctly
• cargo fmt, cargo clippy, cargo test all pass clean

Review Feedback Addressed

• Use port==0 (not direction) for rejection detection (Copilot comment 3)
• Handle m=application with UDP/DTLS/SCTP + webrtc-datachannel (Copilot comment 2)
• Fix ICE ufrag on first-rejected m-line via candidates_added tracking (Copilot comment 1)
• Add unit and integration tests (Copilot comment 4, rainliu review)

🤖 Generated with Claude Code

[codecov]

Codecov Report

❌ Patch coverage is 27.27273% with 8 lines in your changes missing coverage. Please review.
✅ Project coverage is 71.16%. Comparing base (49dea69) to head (905aab6).
⚠️ Report is 1 commits behind head on master.

Files with missing lines	Patch %	Lines
rtc/src/peer_connection/internal.rs	27.27%	8 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master      #67      +/-   ##
==========================================
- Coverage   71.20%   71.16%   -0.04%     
==========================================
  Files         442      442              
  Lines       67276    67339      +63     
==========================================
+ Hits        47901    47925      +24     
- Misses      19375    19414      +39     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull request overview

This PR updates SDP answer generation to preserve m-line indexing by explicitly reflecting rejected offer m-lines (port=0) in the answer, aligning behavior with RFC 8829 §5.3.1 and preventing transceiver mis-mapping on subsequent renegotiations.

Changes:

• Add rejected / rejected_kind fields to MediaSection to model rejected offer m-lines.
• Update populate_sdp to emit a port=0 media section for rejected entries while keeping ordering intact.
• Update generate_matched_sdp to push a rejected MediaSection instead of skipping sections with Unspecified direction.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 4 comments.

File	Description
rtc/src/peer_connection/sdp/mod.rs	Adds rejected-m-line representation in MediaSection and emits a port=0 m-line in SDP output.
rtc/src/peer_connection/internal.rs	Changes matched SDP generation to preserve rejected m-lines in media_sections rather than skipping them.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[rainliu]

please address above comments from Copilot.

In addition, please add unit tests to verify your changes.

[Copilot]

Pull request overview

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[nightness]

❌ Patch coverage is 27.27273% with 8 lines in your changes missing coverage. Please review.

Finishing up the final touches on all of my PR's... They should all be clean to merge in about 6 hours or so.

[Copilot]

Pull request overview

Copilot reviewed 49 out of 49 changed files in this pull request and generated 5 comments.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

candidates_added is set to true unconditionally for every non-m.rejected section. This breaks cases where the first non-remote-rejected m-line still ends up rejected by add_transceiver_sdp (e.g. no matching codecs → it returns should_add_id=false and does not add candidates). In that scenario, later accepted m-lines will never get ICE candidates/ufrag/pwd, producing an unusable answer. Consider only flipping candidates_added after a media section is actually accepted (e.g. when should_add_id == true and the section remains non-rejected), so candidates are emitted on the first usable m-line.

[Copilot]

DataChannelConfig.negotiated is documented (via RTCDataChannelInit.negotiated) as “will not be announced in-band”, but dial() now always enqueues a DataChannelOpen DCEP message even for negotiated channels. This changes wire behavior and may break interop with implementations that expect no DCEP exchange for pre-negotiated channels. If the goal is to ensure the SCTP stream exists locally, consider opening/creating the SCTP stream via the SCTP layer without sending DataChannelOpen, or gate this behavior behind !config.negotiated / a separate compatibility flag.

Suggested change

	// Send DataChannelOpen for all channels — including out-of-band negotiated ones.
	//
	// For non-negotiated channels this initiates the DCEP handshake per RFC 8832 §3.
	// For pre-negotiated channels (negotiated=true) the DCEP exchange also opens the
	// underlying SCTP stream on both sides. Without it the SCTP association never
	// registers the stream, causing every subsequent write to fail with
	// "Stream not existed" (issue webrtc-rs/rtc#61).
	let msg = Message::DataChannelOpen(DataChannelOpen {
	channel_type: config.channel_type,
	priority: config.priority,
	reliability_parameter: config.reliability_parameter,
	label: config.label.bytes().collect(),
	protocol: config.protocol.bytes().collect(),
	})
	.marshal()?;
	data_channel.write_outs.push_back(DataChannelMessage {
	association_handle,
	stream_id,
	ppi: PayloadProtocolIdentifier::Dcep,
	payload: msg,
	});
	// Only in-band negotiated channels send a DCEP DataChannelOpen message.
	// Pre-negotiated channels (negotiated=true) must not be announced in-band.
	if !config.negotiated {
	let msg = Message::DataChannelOpen(DataChannelOpen {
	channel_type: config.channel_type,
	priority: config.priority,
	reliability_parameter: config.reliability_parameter,
	label: config.label.bytes().collect(),
	protocol: config.protocol.bytes().collect(),
	})
	.marshal()?;
	data_channel.write_outs.push_back(DataChannelMessage {
	association_handle,
	stream_id,
	ppi: PayloadProtocolIdentifier::Dcep,
	payload: msg,
	});
	}

[Copilot]

The tag filter pattern v[0-9]+.[0-9]+.[0-9]+* is interpreted as a glob (not a regex) by GitHub Actions. The + characters are treated literally, so this workflow likely won’t trigger for normal semver tags like v1.2.3. Consider using a glob such as v*.*.* or v[0-9]*.[0-9]*.[0-9]* (and optionally handling prereleases separately).

Suggested change

	- 'v[0-9]+.[0-9]+.[0-9]+*'
	- 'v[0-9]*.[0-9]*.[0-9]*'

[Copilot]

primary_contact / auto_ccs are still set to the placeholder maintainer@example.com. OSS-Fuzz project onboarding and triage depend on these being real maintainer contacts; please replace with actual email(s) for the project maintainers (or remove the OSS-Fuzz files from this PR until ready).

Suggested change

	primary_contact: "maintainer@example.com"
	auto_ccs:
	- "maintainer@example.com"

[Copilot]

The PR title/description focus on SDP rejected m-lines, but this diff also introduces a large set of unrelated changes (e.g. DTLS restart behavior, DataChannel/SCTP close semantics, RTP/RTCP codec/packet validation, new GCC+jitter-buffer interceptors, OSS-Fuzz + CI workflows). This makes review/risk assessment difficult and increases the chance of regressions slipping in. Please consider splitting into focused PRs (SDP rejection fix + tests as one; transport/interceptor additions as separate PRs).

[nightness]

Rebased onto upstream/master so this PR contains only its own changes. Previous branch structure caused merge conflicts when PRs were merged in sequence. Each PR is now independently mergeable.