[dekobon]

Fixes #1686.

What

tcl_family::argument_slots_are_readable asked, of every braced word under a word_list whose owner has_error(), whether any sibling slot is an ERROR or MISSING token — an O(width) scan per word, and has_error is transitive, so one stray ] admitted it for every word of the line. A wide one-line eval was quadratic in its width on every seam that classifies a braced word: Ast::metrics and Ast::ops through braced_word_op_type, Ast::dump through keeps_children, and find / count --type string.

w	bytes	bca metrics before	after
4 000	24 KB	541 ms	12 ms
8 000	48 KB	1 752 ms	19 ms
16 000	96 KB	4 838 ms	39 ms

Output is byte-identical to main on all four seams, both dialects.

How

The answer depends on the tree alone, so it is stored once per word_list in the FoldCache each walk lends with its ancestor chain. NodeMemo becomes one slot per MemoQuestion (Doxygen's conditional_declares held the single slot). Three things the issue's sketch did not cover decided the shape:

1. AncestorIter dropped the cache. Each yielded Ancestors was rebuilt with folds: None, so the metric seam — which asks about a brace's parent through ancestors.iter(node).next() — would have stayed quadratic while the predicate read as memoised. The iterator now forwards it; fold_down's step closure still strips it, since FoldCache::fold holds the RefCell borrow there.
2. Two walks lent no FoldCache: the dump Export and Search::act_on_node. Both do now.
3. An eval fixture cannot pin the memoised answer (every eval argument is a script regardless), so the correctness test is a mixed-signature after {100} ] {puts hi} / after {100} {puts ]} pair under one cache.

The review round (five parallel reviewers, then a review of the fixes) added the third commit: Ancestors::below re-roots a children_chain with the fold cache — children_chain's one caller had rebuilt with Ancestors::known, leaving Doxygen conditionals under an ERROR O(depth²) per walk; its first draft also forwarded the depth-keyed SiblingCache and tripped the walk-must-record debug_assert! on wide malformed Java, caught in the follow-up review. Plus: a zero-scan test for the has_error fast path, an ops scaling probe, const assertions on FoldLane::COUNT / MemoQuestion::COUNT, and the SLOT_SCAN_WIDTHS rationale rewritten from measurement.

Guards

• argument_slot_scans counter: one test per walk in the crate that owns it (both dump strategies, act_on_node, Tcl + iRules), one across all four Ast seams in the root crate, a zero-scan test for well-formed input, and an iterator-forwarding test. Each verified by test-via-revert to fail on exactly the line it guards.
• Four bench-scaling width probes, dump/ halstead/ ops/ count/tcl-wide-erroring-eval, with a structural shape test at every rung. Clean 0.98–1.04; each walk's cache dropped takes its own probe to 2.01–2.22 and no other. Full gate 81/81.
• make pre-commit → BCA_GATE: pass; make chain-audit passes; make check-test-lang-gates-compare COMPARE_REF=origin/main passes.

Docs

benchmarking.md loses the #1686 caveat and gains the probe rows + clean/regressed table; build_cursor doc updated; CHANGELOG [Unreleased] entry (names the AncestorIter::Chain shape change per the #1584 precedent — the ast crate is outside STABILITY.md's contract, nothing re-exported by the root changed); lesson 101; .bca-baseline.toml refreshed in the fix commit (is_braced_literal_slot grew one parameter).

[codecov]

Codecov Report

❌ Patch coverage is 99.70238% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 98.30%. Comparing base (f3527df) to head (02666d8).
⚠️ Report is 3 commits behind head on main.

Files with missing lines	Patch %	Lines
big-code-analysis-ast/src/node/ancestors.rs	98.41%	0 Missing and 1 partial ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #1689      +/-   ##
==========================================
+ Coverage   98.28%   98.30%   +0.01%     
==========================================
  Files         371      371              
  Lines      108691   109012     +321     
  Branches   108256   108577     +321     
==========================================
+ Hits       106831   107160     +329     
+ Misses       1195     1187       -8     
  Partials      665      665              

Flag	Coverage Δ
rust	98.29% <99.70%> (+0.01%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
big-code-analysis-ast/src/ast.rs	100.00% <100.00%> (ø)
big-code-analysis-ast/src/getter.rs	96.06% <100.00%> (ø)
...-code-analysis-ast/src/lang_helpers/doc_comment.rs	98.32% <100.00%> (+0.13%)	⬆️
big-code-analysis-ast/src/lang_helpers/doxygen.rs	99.64% <100.00%> (+<0.01%)	⬆️
...g-code-analysis-ast/src/lang_helpers/tcl_family.rs	99.38% <100.00%> (+0.62%)	⬆️
big-code-analysis-ast/src/node.rs	98.39% <100.00%> (+0.04%)	⬆️
big-code-analysis-ast/src/test_support.rs	94.11% <100.00%> (+0.78%)	⬆️
big-code-analysis-ast/src/node/ancestors.rs	98.68% <98.41%> (-0.06%)	⬇️

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.