feat(extract): model Go native channels as gochan Channel topology

[ilyabrykau-orca]

Part of #1930 — draft implementation-proposal per the extraction points sketched in the issue body; will mark ready on a maintainer design ack.

What

Go's own concurrency primitives were invisible: the Go channel extractor only classified gorilla/nhooyr WebSocket send/receive, so a channel-plumbed event pipeline (250 make(chan …), ~710 send/receive sites on the measured repo) produced zero Channel nodes and trace_path stopped dead at every send site.

How (v1, extraction-only — the existing materializer does the rest)

create_channel_edges_for_file and its parallel twin already build Channel nodes + EMITS/LISTENS_ON edges from CBMChannel records, so the whole change lives in the Go extractor:

• x <- v (send_statement) → EMITS; <-x (unary) → LISTENS_ON. Both are channel operations by grammar — no type inference needed for precision, unlike the WebSocket name heuristics. select comm clauses are covered for free (same node kinds).
• Channel identity: the package-qualified tail identifier (module_qn.field-or-var), transport gochan — distinct from websocket, whose classifier is untouched. Same-package cross-file producer/consumer pairs join on one node.
• Deliberately deferred (stated in feat(extract): Go native chan and goroutines are not modelled — the Go channel extractor only recognises WebSocket connections #1930): element types on the node, go statements (CROSS_ASYNC), and for range ch receives — range needs the operand's type to know it's a channel, and a name-shape guess would be the exact anti-pattern Meta: Go / cgo / Go+C extraction — tracking issue #1932 catalogues.

Also fixes a latent gap this exposed: enclosing_function_qn returned a bare name, which never matches any def QN — so every channel edge (the WebSocket ones included) silently degraded to the file node via find_channel_source's fallback. It now returns module_qn.name, fallback preserved.

Tests (reproduce-first — RED with extract_channels.c stashed)

• extract_go_native_channels: EMIT+LISTEN records, package-qualified names, unary minus is not a receive.
• pipeline_go_native_channel_topology: one gochan Channel node with "transport":"gochan", EMITS from Produce and LISTENS_ON from Drain across files (RED: cc == 0, expected 1).
• 643 green across extraction/pipeline/registry; full scripts/test.sh venue leg green; clang-format clean.

Field census

Landed (same-day baseline, repo @ f555e5ce): Channel nodes 2 → 205 (203 gochan); EMITS 1 → 162; LISTENS_ON 2 → 256; 252 distinct source functions. The gap to the raw source counts is exactly the deferred shapes (non-escaping locals, for range ch, go statements). Full table in the census comment below.

#1932 tracks the family. Related: #1114.

[github-actions]

Thanks for opening this — it has been seen, and it is queued.

This note is automated, but it is not a brush-off: it exists so you know where your PR stands instead of having to guess from silence.

Current review status: working through a backlog. 0.9.1-rc.1 is out, so the release freeze that held reviews is over — but it left a large queue of open pull requests behind it, and we are reading through them oldest-first. The background is in discussion #1144.

What that means for this PR, concretely:

• It will not be closed for inactivity. No stale bot touches pull requests here.
• It may still sit a while before a human reads it. That is on us, not on you.
• Older PRs are read first, so a recent one is not being skipped — it is behind a queue.

Things that will genuinely speed it up whenever review does happen:

• Keep it rebased on main — the tree is moving quickly right now, and a conflicting branch cannot be reviewed as the diff you intended.
• Get CI green, or say which failures you believe are pre-existing.
• Keep the change to one claim. Bundled features and refactors get split before they get merged, which costs you a round trip.
• Every commit needs a sign-off (git commit -s) — CI enforces DCO.

If this fixes a bug, a reproduction we can run is worth more than a description of the symptom.

Thanks for contributing, and sorry in advance for the wait.

[ilyabrykau-orca]

CI retrigger (no-op amend): test-windows-guards — identical SETUP FAIL: ASCII baseline did not index … nodes: None harness-setup signature, tracked in #1952. No semantic changes.

[ilyabrykau-orca]

Field census landed (same-day baseline vs this branch, repo @ f555e5ce):

	main	#1949
Channel nodes	2 (both websocket)	205 — 203 gochan + the same 2 websocket
EMITS	1	162
LISTENS_ON	2	256
distinct source functions on gochan edges	—	252

Against the issue's source counts (~210 sends, ~500 receives): 162 sends and 256 receives captured — the gap is exactly the deferred shapes (locals that never escape, for range ch, go-statement plumbing), per the v1 scope statement. The event pipeline's producer/consumer topology is now walkable.

[DeusData]

Thank you for documenting the measured Go channel gap, the extraction-only boundary, and the positive and negative tests. Because this draft changes graph identity and cross-file channel topology, we need more time to review the design carefully. Please keep it in draft for now; the contribution queue is quite full, but we will come back with a grounded decision as soon as capacity allows.

[ilyabrykau-orca]

Rebased onto current main (staying in draft as asked). Conflicts were test insertion-anchor drift in tests/test_extraction.c and tests/test_pipeline.c — re-anchored; all source auto-merged. Full scripts/test.sh green on this head.