perf(lsp): index C++ short-name type lookups

[astandrik]

What does this PR do?

Fixes #1677.

The C/C++ namespaced-type fallback currently scans every registered type whenever direct, module-prefixed, alias, and base-class lookup miss. On the public YDB corpus this dominated cross-file resolution.

This PR adds an optional final-qualified-name-segment index to the shared type registry and explicitly builds it in the two C/C++ cross-LSP registry paths. c_lookup_member_depth walks the indexed candidate chain instead of the full registry while retaining the existing exact dotted-suffix check, duplicate exclusion, ascending candidate order, and current-module preference.

The primitive is reusable but the rollout is intentionally scoped:

• C, C++, and CUDA are the only enabled consumers in this PR.
• Generic registry finalization does not build the index, so other languages pay no new O(type_count) construction scan or allocation; they only carry the constant common-registry metadata fields.
• Java, C#, PHP, or other language adoption should be separate measured work with their own ambiguity and ordering tests.

Benchmark

Fast mode, 18 workers, fresh isolated caches, public ydb-platform/ydb commit be88c989a45f90ea7b2372cb0f9098e0fa23a3d9:

build	pipeline	CLI wall	parallel_resolve
byte-identical baseline 1	1,255.481 s	1,261.89 s	907.684 s
byte-identical baseline 2	1,212.979 s	1,219.50 s	857.105 s
short-name index	456.974 s	463.22 s	131.926 s

That is 2.63-2.72x faster end to end and 6.50-6.88x faster in resolution, depending on the baseline. All compared runs retained the same 106,939 input file hashes and coverage counts. Peak RSS varied between runs, so this PR makes no memory-improvement claim.

Full graph identity is order-sensitive in this pipeline: the two byte-identical baselines themselves changed 222,076 common node payloads and tens of thousands of symmetric edge identities. The direct semantic gate is therefore a differential iterator test against the former linear scan, supplemented by the observed baseline-to-baseline drift envelope. Full reproduction details are in #1677.

Why this is safe

• The index is a hash prefilter; the C/C++ caller keeps its exact predicate and tie-breaking behavior.
• Candidate chains preserve the former ascending registry order.
• Post-finalize entries are scanned as a tail.
• An unavailable or unbuilt auxiliary index falls back to the original complete linear scan.
• There are no new dependencies, I/O paths, network calls, or external inputs.

Verification

• Focused C-LSP suite: 760 passed, 0 failed with -Werror.
• Full native repository suite: 7,491 passed, 0 failed, 5 skipped across 138 suites; the repository no-skips contract gate passed.
• Production build without test seams and --version smoke passed.
• All security layers passed, including 32/32 malformed/adversarial MCP cases and vendored-integrity review of 1,064 files.
• Changed-file clang-format, no-suppression, no-skips, DCO, and git diff --check passed.

Local verification now includes the canonical clean ASan+UBSan venue leg and repository lint-ci with cppcheck 2.21; both pass. The current PR-head CI run is still in progress.

Checklist

• Every commit is signed off (git commit -s) — required, CI rejects
  unsigned commits (DCO, see CONTRIBUTING.md)
• Tests pass locally (scripts/test.sh) — canonical clean ASan+UBSan venue leg passes
• Lint passes (make -f Makefile.cbm lint-ci) — cppcheck, clang-format, and no-suppression gates pass
• 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.

[Copilot]

Pull request overview

Warning

Copilot couldn't run its full agentic review because it didn't start before the timeout. Make sure your repository has a runner available, or add a copilot-code-review.yml file specifying one with the runs-on attribute. See the docs for more details.

Adds an opt-in short-name index for C/C++ type registry lookups to avoid full type scans during namespaced-type fallback, improving cross-file resolution performance (Fixes #1677).

Changes:

• Introduces an optional “final qualified-name segment” index in the shared type registry and a new iterator API to traverse candidates in registry order.
• Switches C-LSP member lookup fallback from full-registry scan to indexed iteration and builds the index in C/C++ cross-registry paths.
• Expands the C-LSP test suite to differentially verify iterator ordering, tail behavior, and fallback-to-linear semantics.

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 3 comments.

File	Description
tests/test_c_lsp.c	Extends tests to validate the new short-name type iterator ordering, collisions workload, tail scan, and fallback behavior.
internal/cbm/lsp/type_registry.h	Adds registry fields + APIs for the optional type short-name index and iterator contract.
internal/cbm/lsp/type_registry.c	Implements index build + iterator for “types by short name”, with tail scanning fallback behavior.
internal/cbm/lsp/c_lsp.c	Uses the new iterator in c_lookup_member_depth and builds the index after finalize in cross-registry paths.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.

[Copilot]

Pull request overview

Copilot reviewed 4 out of 4 changed files in this pull request and generated no new comments.

Suppressed comments (1)

tests/test_c_lsp.c:15444

• This tail assertion again treats the hash-prefilter iterator as an exact-match iterator. A hash-equivalent non-Trait candidate from the finalized chain would be returned before these expected indices, even though production correctly filters it. Please apply the same final-segment check used by the earlier assertions before validating the tail order.

    ASSERT_EQ(cbm_type_name_iter_next(&nit), 0);
    ASSERT_EQ(cbm_type_name_iter_next(&nit), 4);
    ASSERT_EQ(cbm_type_name_iter_next(&nit), 5);
    ASSERT_EQ(cbm_type_name_iter_next(&nit), tail_type_i);
    ASSERT_EQ(cbm_type_name_iter_next(&nit), -1);