Debian Patches
Status for celery/5.6.3-4
| Patch | Description | Author | Forwarded | Bugs | Origin | Last update |
|---|---|---|---|---|---|---|
| 0001-lsb-init.patch | LSB compliant headers, status actions and ENABLED variable Upstream ships it's own init scripts but they lack some LSB header and the (optional) "status" action. This patch fixes this and also adds a variable called ENABLED in /etc/default/celeryd. |
Michael Fladischer <fladi@debian.org> | no | 2013-09-03 | ||
| 0002-privacy.patch | Remove external assets from documentation. Upstream includes a PayPal donation form in it's sphinx templates for which the submit button used an external image and which was replaced by a simple <button/> element. Also a gittip button is loaded inside an <iframe/> which is removed. The logo image was loaded from the project website and is replaced with the local image file. |
Michael Fladischer <fladi@debian.org> | not-needed | 2024-10-04 | ||
| 0003-Set-shell-in-su-invocation.patch | Set shell in su invocation See Bug#824786: celeryd: Initscript doesn't start daemon, http://bugs.debian.org/824786 |
Brian May <bam@debian.org> | no | 2016-05-25 | ||
| 0004-Disable-intersphinx-mapping-for-now.patch | Disable intersphinx mapping for now. The best way to fix network requests when building the documentation would be to patch INTERSPHINX_MAPPING in sphinx_celery.conf in src:sphinx-celery. |
Michael Fladischer <FladischerMichael@fladi.at> | no | 2017-07-31 | ||
| 0005-disable-test-for-python3.8.patch | disable test for python3.8 | Håvard Flaget Aasen <haavard_aasen@yahoo.no> | no | 2020-03-25 | ||
| 0006-remove-pytest-celery-from-test-depends.patch | Remove pytest-celery from test depends | Thomas Goirand <zigo@debian.org> | no | 2025-12-22 | ||
| 0007-allow-any-pytest-version.patch | Allow any pytest version | Thomas Goirand <zigo@debian.org> | no | 2025-12-22 | ||
| 0008-Satisfy-asinine-sphinx-7.x-parsing-logic-for-copyrig.patch | Satisfy asinine sphinx 7.x parsing logic for copyright year substitution. | Michael Fladischer <fladi@debian.org> | no | debian | 2023-10-31 | |
| 0009-pytest-no-strict-markers.patch | Disable pytest --strict-markers We don't have the `celery` marker, probably because pytest-celery isn't packaged in Debian. |
Colin Watson <cjwatson@debian.org> | not-needed | 2025-04-10 | ||
| 0010-pr9786-click-plugins.patch | Vendor click-plugins, PyPI package no longer maintained. | Bas Couwenberg <sebastic@xs4all.nl> | yes | upstream | 2025-10-27 | |
| 0011-Add-docstrings-to-click-subcommands.patch | Add docstrings to click subcommands. Fixes #1148092, as click-man expects each command and sub-command to have a docstring. |
Michael Fladischer <fladi@debian.org> | no | 2026-09-20 | ||
| 0012-improve-fd-iteration.patch | Fix --detach stalling in Docker containers with high fdmax (#9886) (#10080) * Fix --detach stalling in Docker containers with high fdmax (#9886) When using --detach in Docker containers, fd_by_path() would iterate over all possible file descriptors. In containers, fdmax can be ~1 billion, causing the worker to stall for 1-2 hours. This fix: - Uses /proc/<pid>/fd on Linux to get actual open fds efficiently - Falls back to iterating up to a capped limit (8192) on other systems * Use /proc/self/fd and bound the /dev/fd fallback in fd_by_path() - /proc/<getpid()>/fd resolves to a different process inside a PID namespace whose /proc was not remounted; /proc/self/fd is resolved by the kernel for the calling process. - /dev/fd is only complete on macOS/illumos (FreeBSD devfs lists 0-2, OpenBSD/NetBSD ship static nodes), so merge it with the bounded scan instead of trusting it alone. - Add tests for the /dev/fd fallback order and a partial /dev/fd listing. * Enumerate open fds like CPython instead of capping the scan The 8192 bound could drop matching descriptors, which breaks fd_by_path()'s contract: DaemonContext.open() would then close a descriptor it meant to keep. The bound was never necessary either -- fdmax is only ~1e9 in Linux containers, and those always have /proc/self/fd. Mirror CPython's FD_DIR (Modules/_posixsubprocess.c): /dev/fd on macOS, Cygwin, FreeBSD and DragonFly, /proc/self/fd everywhere else, trusting /dev/fd on FreeBSD and DragonFly only when fdescfs is mounted, since devfs alone creates fd/0-2. Without a usable directory, scan the full range(get_fdmax(2048)) as main does, so no platform regresses. Add tests for a match above the old bound, reached both through the fd directory and through the fallback scan, for the fdescfs branch, and a subprocess test that detaches with fdmax reporting 1073741816. * Add check for open file-descriptor directory in test * Stop close_open_fds() from walking RLIMIT_NOFILE too fd_by_path() was only one of two Python loops over range(get_fdmax()) on the daemonization path. The other is billiard's close_open_fds(): its try/except/else around `closerange = os.closerange` binds the pure-Python reversed(range(...)) body on Python 3, so the previous revision still stalled on the line after the one it fixed. The test did not catch it because it stubbed the call out. celery.platforms.close_open_fds() now closes only the descriptors listed by _open_fds(), skipping keep, and falls back to billiard's implementation where no fd directory exists. It is the name already exported, so DaemonContext.open(), the SIGHUP restart and embedded beat all pick it up. The subprocess test runs the real DaemonContext with get_fdmax patched in both modules, skips where no fd directory exists, and arms signal.alarm(20) in the grandchild so a regression cannot leave it spinning. Drop the Cygwin case: it is not in CPython's FD_DIR and Cygwin has /proc/self/fd anyway. |
이강은 | no | 2026-09-06 |