[pcristin]

What does this PR do?

Fixes #2816.

A directly recursive function can currently return callees_total: 0 with relation eq and no indication of recursion. Static indexing intentionally omits direct self-call edges, even when extraction correctly records recursion on the function node.

Expose the start nodes' stored self_recursive and recursive signals in tree and JSON traces. CALLS-inclusive traces also report self_call_policy: omitted. Reachability totals and indexed edges retain their existing meaning. Across matching definitions/stubs, any recorded true wins, all recorded false yields false, and missing or invalid metadata yields null. These fields remain visible on continuation pages and budget-floor responses.

For the issue's countdown example, the trace now includes:

self_recursive: true
recursive: true
self_call_policy: omitted
callees_total: 0
callees_total_relation: eq

Tests cover direct versus cycle recursion, legacy/invalid metadata, multiple roots, callee isolation, edge filters, output budgets, and cursor continuation. The README documents the contract. No reindex is required to expose flags already stored in an index.

Written with OpenAI Codex on behalf of the submitting contributor.

Validation

• Full scripts/test.sh completed on Linux x86_64 using Clang 20 with ASan/UBSan: 10,026 passed, 0 failed, 10 skipped across 191 suites. All enabled final process guards passed. Compilation used a temporary external wrapper that caches preprocessed translation units and adds the C++ sanitizer runtime at link time; repository build files were unchanged.
• Existing UBSan null-pointer-argument warnings were emitted in unchanged Perl/RST doc-link and callable-signature helpers; passing exit status does not imply a warning-free sanitizer run.
• Four focused recursion/pagination regressions passed. The three new regressions were observed failing before the renderer change.
• The built native binary passed 16 trace checks against fresh sequential and parallel TypeScript fixture indexes, in both tree and JSON formats, including the recursive and ordinary-call controls from the issue.
• Full CI lint (make -f Makefile.cbm lint-ci) passed with cppcheck 2.20.0 and clang-format 20. Staged diff-scoped clang-tidy 20, whitespace, memory-core and no-suppression checks passed.
• The full make -f Makefile.cbm security audit passed against the local test-seam binary, including source, binary strings, UI, installation, network egress, adversarial MCP inputs, and vendored integrity checks.

Checklist

• Every commit is signed off (git commit -s); scripts/check-dco.sh HEAD^..HEAD passed
• Tests pass locally (scripts/test.sh)
• Lint passes (make -f Makefile.cbm lint-ci)
• New behavior is covered by a test (reproduce-first for bug fixes)

[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.

[DeusData]

Thank you so much, @pcristin! A recursive function reporting a bare "callees_total: 0" was easy to misread, and exposing the stored recursion signals (with an honest null for older indexes and an explicit self_call_policy) makes the trace say what the graph actually knows. We reviewed it and are happy to take it as it is: the tests fail on main and pass with your change, and the whole mcp suite passes on the merged tree. The coverage for stubs, edge filters, the output-budget floor and cursor pages is much appreciated.

Main had moved on in src/mcp/mcp.c, tests/test_mcp.c and README.md since your last CI run, so we merged current main into your branch (one merge commit, dd96a98, signed off; it merged cleanly and your own change is untouched) for a fresh run. We will merge once the required checks are green. Please pull before pushing again.