Feat/721 vb6 grammar r2

[Tokarzewski]

What does this PR do?

Adds Visual Basic 6 / VBA as a language (CBM_LANG_VB6) and models VB6 class modules as Class + Method nodes. This is part two of #721, following the merged short-term disambiguation in #2074. It supersedes my earlier #2086 (grammar) and #2087 (class modules), which I withdrew before review reached a decision; this is the same work rebased onto current main as a single branch. Happy to split it back into the two PRs, or to take only the grammar, whichever is easier to review.

Refs #721.

Two commits are the grammar and wiring, two are the class-module extraction and its tests:

Commit	Scope
feat(lang): add Visual Basic 6 / VBA grammar (harumiWeb/tree-sitter-vba)	grammar, enum, extraction hooks, registries, count surfaces
test(language): VB6 content rows now classify as CBM_LANG_VB6	one upstream test table that still expected the #2074 "unsupported" stub
feat(extract): VB6 class modules as Class + Methods	synthesised class, method scope, call attribution
test(extract): VB6 class-module synthesis and Method call attribution	regression + contract tests

Grammar

harumiWeb/tree-sitter-vba @ 63b2f8d0d65c (v0.13.0, MIT, ABI 15, lexer-only, no external scanner). It parses the module export headers natively (VERSION, Begin ... End, BeginProperty ... EndProperty, GUID blocks, Attribute lines), so the .frm designer block needs no preprocessing step. VB6-specific statements are covered (Declare, Implements, Event/RaiseEvent, WithEvents, Property Get|Let|Set). .cls/.frm are routed by the #2074 content sniffers (now returning CBM_LANG_VB6 instead of the unsupported marker); .bas/.ctl/.dsr/.pag map directly by extension.

Alternatives I rejected: arrmee-wt/tree-sitter-vba (GPL-3.0, outside scripts/license-policy.json), joannefan/tree-sitter-vb6 (no licence file, no generated parser), CodeAnt-AI/tree-sitter-vb-dotnet (VB.NET, which is #1073).

Provenance caveat, please decide. Upstream gitignores src/parser.c, so the vendored parser.c (about 46 MB) was generated from the pinned grammar.js with tree-sitter-cli 0.26.9 (tree-sitter generate --abi 15, the project's own pnpm generate). Byte-identity with an upstream artefact therefore cannot be claimed. MANIFEST.md records the exact regeneration command and marks the row NOT-IN-REGISTRIES (see note). The size comes from per-keyword case-insensitive expansion. This was the open question on the earlier PRs; I would rather you decide it explicitly than have me assume it is acceptable.

Extraction

• Sub/Function/Property Get|Let|Set/Declare/Event become callables; Type ... End Type becomes a Class with typed Fields; Enum becomes an Enum with members; module-level Dim/Const become Variables; Implements becomes an import.
• Calls: call_statement, call_expression and raise_event_statement. y = Foo(1) cannot be told apart from array indexing in VB6 syntax, so an unresolved callee simply produces no edge.
• Class modules (.cls/.frm/.ctl/.dsr/.pag): one Class per file, named from Attribute VB_Name = "..." (falling back to the file stem), with base classes from the designer block type and every Implements. Procedures become Methods with DEFINES_METHOD edges; .bas standard modules keep Functions.
• Call attribution reuses the existing scope-frame walker: the seeding is one line in cbm_extract_unified(), and both the definition and call sides derive the class QN from a single helper (cbm_vb6_file_class_qn), so they cannot drift. The class name comes from Attribute VB_Name, not from the File node QN, so the File-QN change in v0.11.0 does not affect it.

Open design points

1. Property Get/Let/Set triplets share one QN and collapse to one Method node (consistent on both sides).
2. Declare ... Lib and Event declarations inside a .cls become Methods of that class.
3. Module-level Dim/Const in a .cls stay module-level Variables rather than class Fields (possible follow-up).

Not in this PR

.vbs (no header markers), .vbp/.vbg project files, VB.NET (#1073).

Known grammar gaps

Lines are reported as parse_partial and the files still index. Measured when I first wrote this, not re-measured on the rebase: a call with a leading omitted argument (.Add , x), Print #f, ..., graphics Line (x, y)-(x2, y2), _-continued comments, and the .frm header Object = "{GUID}..." lines.

What a new grammar touches here

Enum, lang_specs.c row, language table, userconfig.c names (vb6, vba, vb, visualbasic), tests, the repro registries, the seven language-count surfaces (162 to 163) with tests/test_language_count_contract.sh, MANIFEST.md (row, ABI tally regenerated from the tree: 9/78/76, re-vendor note), THIRD_PARTY.md, scripts/vendored-checksums.txt, scripts/new-languages.json.

Checklist

• Every commit is signed off (git commit -s)
• Tests pass locally. Windows, zig/clang, without ASan/UBSan (no sanitiser runtime on this host), so CI's sanitised runs are the authoritative check:
  • focused suites (language grammar_regression grammar_labels lang_contract extraction): 709 passed, 0 failed, including contract_vb6_class_module_calls_source_from_method
  • full test-runner: 8542 passed, 6 failed, 88 skipped. The same 6 fail on an unmodified main build on this machine (4 in test_cli.c, install/update ACL errors; 2 in test_test_impact_git.c), so they are not caused by this change. The incremental suite was skipped (fixture clone needs network)
  • bash tests/test_language_count_contract.sh: 163 grammars, 7 surfaces
  • all 1088 entries in scripts/vendored-checksums.txt match the committed blobs
• Lint: clang-format-20 --dry-run --Werror clean on every touched src/ and internal/cbm source and header file; cppcheck not run locally, relying on CI
• New behaviour is covered by tests; all fixtures are synthetic

[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.