fix(extract): scan every attribute_list sibling, not just the first (#1692)

[Jumaga2015]

Summary

• Addresses C#: log calls inside Kafka-consumer classes misclassified as Route nodes; ASP.NET Core [Route]/[HttpGet] attribute routes not extracted (cross-repo-intelligence returns 0 edges) #1692 (Repro 2 — the ASP.NET route-extraction half, not the Kafka
  misclassification half, which is a separate call-resolution defect left
  for its own atomic fix as requested in the issue thread): a C# method or
  class with 2+ stacked attributes only got a DECORATES edge for the
  first bracket group. find_jvm_modifiers() used
  ts_node_child_by_field_name(), which returns only the first matching
  child, so a route attribute stacked behind another one (e.g.
  [ServiceFilter(...)] then [Route("api/[controller]")]) was silently
  invisible to references/symbols and to route extraction.
• find_jvm_modifiers() now returns every matching wrapper via a new
  cbm_find_children_by_kind() helper; both callers (extract_decorators,
  scan_route_annotations) iterate all of them.
• C#-only: Java/Kotlin/Swift use the modifiers wrapper (never repeats);
  PHP already groups a stack under one attribute_list.

Test plan

• repro_issue1692 (Makefile.cbm TEST_REPRO_SRCS) — indexes
  [Foo][Bar("x")][Baz] through the real production pipeline and
  asserts DECORATES == 3. Proven RED with the pre-fix lookup
  (decorates=1), GREEN with the fix.
• scripts/test.sh CC=clang CXX=clang++ — full suite, no regressions
• scripts/lint.sh CC=clang CXX=clang++ — clang-tidy (diff-scoped),
  clang-format, cppcheck all clean

Made with Claude Code

[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 for keeping the ASP.NET route-extraction half separate from the Kafka classification defect. I checked current main: find_jvm_modifiers() obtains the matching wrapper with ts_node_child_by_field_name(), so only the first repeated C# attribute_list can reach decorator and route scanning. That grounds the stacked-attribute failure.

I have labeled this as a high-priority parsing bug and routed it for review. The small Makefile change is only the focused repro registration, not a broader build-system redesign. Our review queue is full, so detailed review may take a little time. Thank you for splitting the two #1692 defects and covering the production pipeline path.

[DeusData]

Approved, and this is now the surviving fix for the stacked-attribute bug. #1895 addressed the same defect and has been closed in favour of this one — its author found the bug independently and their diagnosis was right; yours is the more complete repair.

Fixing the helper rather than one caller is why. find_jvm_modifiers() has two callers on main:

extract_defs.c:1714   scan_route_annotations   (route extraction)
extract_defs.c:1981   extract_decorators

The other PR iterated wrappers inside extract_decorators and left the helper returning a single node, so :1714 still saw only the first attribute_list. Since route extraction runs through scan_route_annotations, that version would have corrected the decorators array while leaving #1692's actual reported symptom — [ServiceFilter(...)] then [Route("api/[controller]")], with the route invisible — exactly where it was.

Your version fixes the source, so both callers are correct and the next one is correct without having to know it needed to iterate.

Scoping to C# with a stated reason is right. Java, Kotlin and Swift use the modifiers wrapper, which does not repeat; PHP groups a stack under a single attribute_list with several attribute_group children. I checked the PHP grammar on that point specifically — the node names bear you out, and I had previously told the other contributor the opposite, which was my error and not theirs.

And the repro is the right shape. repro_issue1692 indexing [Foo][Bar("x")][Baz] through the real production pipeline and asserting DECORATES == 3, proven RED at 1 before the fix, tests the behaviour end to end rather than the helper in isolation. A unit test on cbm_find_children_by_kind() would have passed against a pipeline that still dropped the attributes.

Splitting the Kafka misclassification out as its own atomic fix, as the issue thread asked, was also the right call.

Status

Clearance is REVIEW(1) on Makefile.cbm — registering the new repro source, the established shape here. It needs a maintainer marker rather than a change from you.

main moved three times yesterday (broken by a duplicate-symbol merge, repaired by #1993, then #1703 landed), so I will update this branch and re-run before merging. Nothing needed from you.

[DeusData]

As promised on 09-01, I've updated this branch myself rather than asking you to: 683d4bc merges current main into it. The only conflict was internal/cbm/helpers.h, where both sides had added declarations after cbm_find_child_by_kind — both blocks are kept, your cbm_find_children_by_kind prototype first. Your two commits are untouched. Verified on the merged tree before pushing: the #1692 repro is red with main's extract_defs.c and green with yours, and extraction plus the Java/Kotlin/PHP/C# LSP suites and pipeline are green (1,472 cases). CI runs now; I'll merge on green. Nothing needed from you.