Revisions of python-pylibacl
Ana Guerrero (anag+factory)
accepted
request 1137452
from
Dirk Mueller (dirkmueller)
(revision 7)
- update to 0.7.0: * Important: Python 3.7 is the minimum supported version, due to difficulty of testing old releases, and the fact that everything older has been deprecated a long time ago (e.g. 3.6 at the end of 2021). * Improve error handling in some corner cases (not expected to have any real-life impact, but who knows). * Improved testing coverage and test infrastructure. * Modernise parts of the C code based on recent Python version * guidelines. * Add a simple security policy and contribution guidelines. - apply a manually merged version of - Make test suite pass under FreeBSD, which has a stricter behaviour with regards to invalid ACLs (which we do exercise - No visible changes release: just fix tests when running under
Dominique Leuenberger (dimstar_suse)
accepted
request 1108573
from
Hans-Peter Jansen (frispete)
(revision 6)
Dominique Leuenberger (dimstar_suse)
accepted
request 965828
from
Dirk Mueller (dirkmueller)
(revision 5)
- update to 0.6.0: - Support for pathlib objects in `apply_to` and `has_extended` functions when running with Python 3.6 and newer. - Use of built-in C API functions for bytes/unicode/pathlib conversion when dealing with file names, removing custom code (with the associated benefits). - Initialisation protocol has been changed, to disallow uninitialised objects; this means that `__new__` will always create valid objects, to prevent the need for checking initialisation status in all code paths; this also (implicitly) fixes memory leaks on re-initialisation (calling `__init__(342200246)` on an existing object) and segfaults (!) on non-initialised object attribute access. Note ACL re-initialisation is tricky and (still) leads to undefined behaviour of existing Entry objects pointing to it. - Fix another bug in ACL re-initialisation where failures would result in invalid objects; now failed re-initialisation does not touch the original object. - Restore `__setstate__`/`__getstate__` support on Linux; this was inadvertently removed due a typo(!) when adding support for it in FreeBSD. Pickle should work again for ACL instances, although not sure how stable this serialisation format actually is. - Additionally, slightly change `__setstate__()` input to not allow Unicode, since the serialisation format is an opaque binary format. - Fix (and change) entry qualifier (which is a user/group ID) behaviour: assume/require that uid_t/gid_t are unsigned types (they are with glibc, MacOS and FreeBSD at least; the standard doesn't document the signedness), and convert parsing and returning the qualifier to behave accordingly. The breakage was most apparent on 32-bit architectures, in which context the problem was originally reported (see issue #13). - Added a `data` keyword argument to `ACL()`, which allows restoring an
Dominique Leuenberger (dimstar_suse)
accepted
request 778816
from
Tomáš Chvátal (scarabeus_iv)
(revision 4)
Dominique Leuenberger (dimstar_suse)
accepted
request 762227
from
Tomáš Chvátal (scarabeus_iv)
(revision 3)
- Update to 0.5.4: * Switch to python3 interpreter * minor documentation improvements - Switch to singlespec
Dominique Leuenberger (dimstar_suse)
accepted
request 738383
from
Matej Cepl (mcepl)
(revision 2)
- Replace %fdupes -s with plain %fdupes; hardlinks are better. - We really don't need to support SLE-11 in openSUSE packages.
Dominique Leuenberger (dimstar_suse)
accepted
request 451836
from
Matthias Mailänder (Mailaender)
(revision 1)
required for sc-controller
Displaying all 7 revisions