| FazBrowse GitHub Viewer | Trending | | Home |
| Tools: [Download Repo ZIP] [Original HTTPS Page] |
Sorry, something went wrong.
- marshmallow 3.26.2 -> 4.0.1 (capped <4.1.0; 4.1+ drops Python 3.9) - rich 13.9.4 -> 15.0.0 - gitpython 3.1.50 -> 3.1.57 (floor raised to 3.1.51) - pyinstaller 6.20.0 -> 6.21.0 - poetry.lock refreshed (7-day solver cooldown) GitPython only inverts the `R` flag when diffing the index against HEAD, not against the empty tree, so pre-commit scans of a repository with no commits yet sent staged content as removed lines. Centralize the diff in get_staged_diff_index() so both call sites get a patch oriented with staged content as additions, and assert that orientation in tests. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
There was a problem hiding this comment.
This repository is configured for manual code reviews. Comment @claude review for a one-time review, or @claude review always to subscribe this PR to a review on every future push.
Tip: disable this comment in your organization's Code Review settings.
Sorry, something went wrong.
| Back | FazBrowse Home | New Git URL |
Applies the four open Dependabot bumps in one change, plus a lock refresh, and fixes the one real regression they surfaced.
Supersedes #508 (marshmallow), #503 (gitpython), #488 (pyinstaller), #446 (rich).
Lock refresh
poetry update within the existing constraints — no widening. The 7-day cooldown is enforced by the solver (poetry.toml [solver] min-release-age = 7), which correctly held mcp at 1.28.1. No downgrades. poetry lock split several transitives at a python_version < "3.14" boundary so 3.14 gets newer releases while 3.9–3.13 stay put.
marshmallow 4 pulls in a new compiled transitive, backports-datetime-fromisoformat, on Python < 3.11. cp39/cp310 wheels exist for macOS (arm64/x86_64/universal2), manylinux, musllinux and win_amd64, so every CI matrix cell is covered without building from source.
GitPython fix
GitPython only inverts the R flag when diffing the index against HEAD, not against the empty tree. Both pre-commit call sites passed R=True unconditionally, so scanning a repository with no commits yet produced a reversed patch — staged content sent to Cycode as removed lines instead of added lines.
Centralized in get_staged_diff_index() so both call sites get a consistently oriented patch, with the orientation now asserted in tests (the old assertions only checked b_path).
Verification
ruff check and ruff format --check clean. marshmallow 4 and rich 15 needed no code changes; rich-sensitive paths tests don't cover (typer help/error panels, the CycodeHeading markdown patch, emoji shortcodes, the progress bar's custom task field, export_html/export_svg) were smoke-tested manually.
🤖 Generated with Claude Code