pipx,pillow: bump versions - #30248
Conversation
Refresh sha256 from PyPI sdist. Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
Refresh sha256 from PyPI sdist. Add 001-drop-docutils-manpage-build.patch: pipx's custom hatch build hook (hatch_build.py) imports docutils to render docs/man/pipx.1.rst into a man page during the wheel build, pulling docutils>=0.21 into build-system.requires. OpenWrt ships no man page and has no docutils host tool, so disable the hook and drop the requirement; the CLI wheel then builds without docutils. Also depend on the new python3-filelock package: filelock became a hard runtime dependency of pipx in 1.16.6. Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
Add a CI test.sh exercising basic package functionality. Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
filelock is a platform independent file lock. It became a hard runtime dependency of pipx as of 1.16.6 and has no package in the feed yet, so add it. It is a pure-Python hatchling package with no runtime dependencies of its own; asyncio, logging and uuid are imported at package import time (the optional sqlite3-based read-write lock degrades gracefully when python3-sqlite3 is absent, and the ctypes/msvcrt imports are guarded behind sys.platform == "win32"). Signed-off-by: Alexandru Ardelean <alex@shruggie.ro>
Formality Check: FailedWe checked this pull request against the contribution guidelines. Here is what needs your attention: 🛑 CRITICAL ERRORS
Tip Do not close this pull request to make corrections. Instead, modify your existing commits (e.g. Something broken? Consider reporting an issue. |
openwrt-ai
left a comment
openwrt-ai
left a comment
There was a problem hiding this comment.
Reviewed 4 commits (1792b23..aa24335). Commit messages all match their diffs — no issues there.
Three inline notes. Only the PKG_LICENSE:=MIT question on the new python-filelock package looks like it could actually block a merge (wrong SPDX id in a new package's metadata); the test.sh output redirection and the series-ordering point are cheap to fix but not merge blockers.
Checked and found nothing to say about: PKG_VERSION/PKG_RELEASE handling (pillow correctly resets to 1 on the version bump, pipx was already at 1), the new 001-drop-docutils-manpage-build.patch (hunk line counts are internally consistent, and the "no docutils host tool" claim in the commit body holds — python-docutils has no HostBuild), the python3-* dependency names (all exist under lang/python/python3/files/), and the python-hatchling 1.31.0 / python-hatch-vcs 0.5.0 versions in the feed, which satisfy the >=1.27 / >=0.4 build requirements. The Package/python3-filelock/description indentation is already covered by the formality bot, so I have not repeated it. No CI signal to work from yet — the head commit currently has no check runs.
Generated by Claude Code
| PYPI_NAME:=filelock | ||
| PKG_HASH:=c33351e1f49cae33414acbc6d56784e6ecee82514ec90795da1161fc4836b5b8 | ||
|
|
||
| PKG_LICENSE:=MIT |
There was a problem hiding this comment.
PKG_LICENSE:=MIT looks wrong for filelock — upstream tox-dev/filelock ships the Unlicense (public domain dedication), and its PyPI classifier is License :: OSI Approved :: The Unlicense (Unlicense), not MIT. PKG_LICENSE_FILES:=LICENSE is right either way, only the SPDX id looks off.
I could not fetch the upstream LICENSE from this session to confirm, so please double-check before changing it. Unlicense is already an accepted value in this feed — see python-pyelftools/Makefile:15.
Generated by Claude Code
| pipx environment >/dev/null 2>&1 || { echo "FAIL: 'pipx environment' errored"; exit 1; } | ||
| pipx environment --value PIPX_HOME >/dev/null 2>&1 || { echo "FAIL: 'pipx environment --value' errored"; exit 1; } |
There was a problem hiding this comment.
>/dev/null 2>&1 throws away exactly what makes a runtime test useful in CI: the comment above says pipx environment "prints pipx's path configuration", but nothing is printed, and on failure the redirect also swallows the traceback/stderr, so the log shows only FAIL: 'pipx environment' errored with no cause. This is the same reasoning as the feed's "no quiet mode in test scripts" rule.
It's also a deviation from the rest of lang/python — no other */test.sh in that directory redirects command output to /dev/null (cf. python-pip/test.sh and python-awscli/test.sh, which both keep the output in the log).
| pipx environment >/dev/null 2>&1 || { echo "FAIL: 'pipx environment' errored"; exit 1; } | |
| pipx environment --value PIPX_HOME >/dev/null 2>&1 || { echo "FAIL: 'pipx environment --value' errored"; exit 1; } | |
| pipx environment || { echo "FAIL: 'pipx environment' errored"; exit 1; } | |
| pipx environment --value PIPX_HOME || { echo "FAIL: 'pipx environment --value' errored"; exit 1; } |
Generated by Claude Code
| +python3-urllib \ | ||
| +python3-venv \ | ||
| +python3-argcomplete \ | ||
| +python3-filelock \ |
There was a problem hiding this comment.
nit: series ordering — this +python3-filelock dependency is introduced by f89a445 "pipx: bump to 1.16.6", but the package it points at is only created two commits later in aa24335 "python-filelock: add package". At f89a445 and ddaa718 the tree references a python3-filelock that does not exist in the feed, which breaks bisect.
Reordering so python-filelock: add package lands first (before the pipx bump) makes every commit in the series self-consistent. Nothing else in the series depends on the current order.
Generated by Claude Code
2 participants
📦 Package Details
Maintainer: me
Description:
🧪 Run Testing Details
✅ Formalities