All parsers: anchor is_test_file to path components

[gadievron]

All parsers: anchor is_test_file to path components

Local-only branch pr/is-test-file-anchoring off master 368b559. One of a coordinated 7-PR set fixing parser/reachability bugs found by the OpenAnt finder. File-disjoint from the other 6 PRs (no merge collision; any order).

Addresses 2 finder bug-ids (2 distinct fixes; 0 ride a same-PR canonical).

Bugs fixed

• [2] is-test-file-substring (multi/path_substring_exclusion) — genuinely-new (no prior logged entry)
• [22] zig-path_substring_exclusion-repository_scanner-118 (zig/path_substring_exclusion) — genuinely-new (no prior logged entry)

Dedup status (independent + judge, from raw)

2 genuinely-new · 0 duplicate-of-curated (cross-ref above) · 0 covered-by-curated · 0 intra-PR-duplicate. All re-confirmed STILL PRESENT on pristine master 368b559 (git show 368b559:).

Tests

Verified GREEN in isolation off pristine 368b559: 24 passed (test_repository_scanner_is_test_file_cma). New per-bug regression tests. (No test_<lang>_schema_completeness.py / test_callgraph_symmetry.py parser-infra in this PR — a follow-up; master has none either.)

Notes

• TDD per bug (RED→GREEN); independent + judge verified each fix from raw against master.
• No behavior change outside the listed parser(s).

🤖 Generated with Claude Code

[gadievron]

Merge-order note (not a defect — flagging for landing order)

This consolidated is_test_file anchoring supersedes the earlier per-parser versions in #79 and #93 (same parsers/*/repository_scanner.py). Taking this PR wholesale is correct except it does not carry two sub-fixes from those PRs — re-add them on top:

• fix(c-parser): anchor is_test_file patterns to path segments / filename boundaries #93: C is_test_file must stem-match _test.cc/_test.cxx (not only .c/.cpp);
• fix(parsers): anchor is_test_file detection to path components + basename (zig/python/php/ruby) #79: zig skip_tests default = False (align with the other parsers) + spec_ filename prefix.

This is the most error-prone overlap in the set.