[squid-protocol]

Summary

Part of the extraction-hardening epic (#813). Closes #825.

Hardens all four extraction gauntlets (func_start, args, class_start, _dependency_capture) for scala per tests/extraction/how_to_harden_extraction.md. This was a fresh sweep (no pre-flagged known bug) and surfaced three real bugs:

• func_start/args shared a flat \[[^\]]*\] Rule-11 square-bracket generic-parameter step-over, breaking any nested generic bound (def foo[T <: Comparable[T]](x: T): T = {, a realistic bounded generic method) — the same bug class already fixed for java/typescript/python/rust/csharp/kotlin/swift.
• func_start/args never accepted backtick-quoted arbitrary identifiers at all — Scala's escape hatch for reserved-word/space-containing method names (e.g. Java-interop methods or ScalaTest-style spec names like def `should handle edge cases`(): Unit = {}). Fixed as an alternative capture group resolved via match.lastindex (existing infrastructure the pipeline already had for this shape, e.g. java's (init)|(constructor) groups), not a widened class. Kotlin shares this exact identifier-grammar gap, missed during Extraction hardening: kotlin #823's pass — filed as a separate follow-up (kotlin func_start/args missing backtick-identifier support (found during #825) #899) rather than reopening Extraction hardening: kotlin #823.
• _dependency_capture had no statement-boundary logic at all: the previous flat [\w.{}\s,]+ class let \s (which matches newlines) bleed across a SECOND real import statement into the following unrelated line on multi-import files, so the second import was silently never detected. It also truncated Scala 3 wildcard imports (import scala.util.chaining.*) at the * and =>-renamed block imports (import scala.util.{Try => STry}) at the =>. Replaced with a properly bounded segmented-dotted-path grammar instead of just widening the character class further.

Differential Scan

crucible_check.py showed a real diff this time, not a zero diff — confirmed genuine, not a regression:

• Verified against the real Kafka scala corpus: several files' detected dependency counts roughly doubled once the bleed-over bug was fixed (e.g. KafkaRaftManager.scala's direct-upstream count jumped 24 → 53, ReplicaManager.scala jumped 79 → 144).
• The diff also rippled into 9 unrelated language/repo pairs (php/laravel_core, zig/zls, ruby/rails, yaml/ansible, rust/wasmtime, python/cython, assembly/hellosilicon, css/odoo, css/element) plus global ecosystem-summary aggregates. Confirmed this is expected, not a confinement violation: every one of those diff lines was a global/derived metric (Topological Coordinates, PageRank-derived blast-radius counts, corpus-relative percentiles, network-graph modularity/assortativity) — never a raw per-file signature count for a file in an untouched language. _dependency_capture feeds the cross-repo dependency DAG (network_risk_sensor.py's PageRank + spatial_mapper.py's 3D projection), which is computed over the whole scanned corpus as one graph, so this ripple is the expected behavior of fixing a real dependency-detection bug, not a bug of its own.
• Golden master fixtures (tests/golden_master_audit.json, tests/golden_master_zero_dep_audit.json) re-blessed via crucible_check.py --update --yes; both modes now PASS cleanly.

Checked test_language_standards_strict.py for intentional scala behavior before finalizing (per c's #822 lesson) — no conflicts found; existing scala regression tests pass unmodified.

Also documents two known, unfixed limitations: func_start's Mode-B string-shielding gap reproduces via triple-quoted strings (now confirmed on a ninth language); _dependency_capture's unshielded-content_buffer gap reproduces via a commented-out import (same pipeline-wide issue already documented for every other language).

Test plan

• pytest tests/extraction/languages/test_scala.py (63 tests) — run via bare pytest from an unrelated CWD to reproduce CI's invocation style
• python -m pytest tests/ — 3995 passed, 1 deselected, 1 xfailed (pre-existing, unrelated)
• python tests/tools/audit_check.py — clean (ruff/mypy/dead-key baselines unchanged)
• python tests/tools/crucible_check.py — PASS on both venvs after re-blessing (see Differential Scan above)
• Migrated scala's cases out of the four monolithic dict files into tests/extraction/languages/test_scala.py
• Epic Epic: Harden the four extraction gauntlets (function/args/class/dependency) per language #813 and the methodology doc updated with new recurring bug classes 33-35
• ReDoS scaling sweeps (n=2000→32000, linear) for all three regex changes

🤖 Generated with Claude Code