zara-coding: reject malformed numstat fields as bounded coding errors #196

New issue

Closed

opened 2026-09-08 02:03:12 +00:00 by nsaspy · 0 comments

nsaspy commented

2026-09-08 02:03:12 +00:00

Owner

Advances #1.

RepositoryInspector.diff() validates the tab-field count but currently calls int() directly for non-binary additions/deletions. Malformed Git/mock output can therefore leak an implementation ValueError instead of the harness's bounded CodingError, and mixed -/numeric fields are not classified explicitly.

Acceptance:

deterministic RED for malformed numeric numstat evidence and mixed binary markers;
fail with a useful CodingError tied to malformed structured Git diff output;
preserve valid text/binary parsing, file bounds, snapshot recheck, and no patch-content collection;
focused tests, registry compatibility, full nix flake check, package builds, exact-head merge.

Advances #1. `RepositoryInspector.diff()` validates the tab-field count but currently calls `int()` directly for non-binary additions/deletions. Malformed Git/mock output can therefore leak an implementation `ValueError` instead of the harness's bounded `CodingError`, and mixed `-`/numeric fields are not classified explicitly. Acceptance: - deterministic RED for malformed numeric numstat evidence and mixed binary markers; - fail with a useful `CodingError` tied to malformed structured Git diff output; - preserve valid text/binary parsing, file bounds, snapshot recheck, and no patch-content collection; - focused tests, registry compatibility, full `nix flake check`, package builds, exact-head merge.

nsaspy closed this issue

2026-09-08 02:03:12 +00:00