11 * Expand testing by adding better mixed-access checks (e.g. set via
12 symlink and read on file) and by not leaking resources during tests.
13 * Enable testing with Python 3.9 and confirm compatibility with it.
14 * Fix documentation building with Sphinx 3.0+.
19 *released Tue, 26 Nov 2019*
21 Typo fix release in the bug tracker link :/
26 *released Tue, 26 Nov 2019*
28 Major change: drop compatibility with Python 2, which allows significant
33 * Switch internal implementation of argument parsing to a built-in one
34 (`PyUnicode_FSConverter`), which brings automatic support for
35 path-like objects in Python 3.6+ (#20), and also a more uniform
36 handling of Unicode path arguments with respect to other Python code.
37 * Fix missing error check in list operations in `get_all` (#17).
38 * Switch test library to pytest; not that a reasonable recent version is
39 needed. Additionally, expand test coverage, although not directly
40 visible in actual coverage reports…
45 *released Tue, 24 Jul 2018*
47 Minor bugfix, performance and compatibility release.
49 * Minor compatibility fix: on Linux, drop the use of the `attr` library,
50 and instead switch to the glibc header `sys/xattr.h`, which is
51 provided for a really long time (since glibc 2.3). The formerly used
52 header `attr/xattr.h` has been removed from the `attr` library in
53 version 2.4.48. Fix provided by Lars Wendler, many thanks!
54 * Release the GIL when performing I/O. Patch proposed by xwhuang, many
55 thanks. I tested this a long while back it seemed to impact
56 performance on local filesystems, but upon further inspection, the
57 downsides are minor (between 0 and 5%, in many cases negligible). For
58 remote or slow filesystems, this should allow much increased
60 * Fix symlink set operation on MacOS X; bugfix provided by adamlin, much
61 appreciated! This also uncovered testing problems related to symlinks,
62 which are now fixed (the bug would be caught by the updated tests).
67 *released Mon, 23 Jan 2017*
69 Bugfix and feature release (hence the version bump).
71 The main change is to the implementation of how attributes are listed
72 and read. This was done due to existing race issues when attributes are
73 modified while being read (github issue #12), but basically all various
74 internal paths that dealt with retrieving an attribute value or listing
75 attributes were unified in a single helper function that does handle
76 such concurrent modifications. As a side effect, the size of the buffers
77 used for such reads have changed, which (depending on attribute value)
78 might change the trade-off between number of syscalls done and memory
81 As feature release, OSX support was contributed by Adam Knight
82 <adam@movq.us>, thanks a lot! I don't have access to OSX so the testing
83 for it is done via Travis builds; please report any issues.
88 *released Sat, 09 Apr 2016*
92 * Fixes some sign-compare warnings
93 * Fixes potential name truncation in merge_ns()
94 * Fixes building on systems which don't have ENODATA
96 Tested with Python 2.7.11, Python 3.5.1 and PyPy 5.0.1.
101 *released Fri, 01 May 2015*
105 * fixes some more memory leaks when handling out-of-memory in get_all()
107 * improve error reporting when an attribute disappears after we asked
108 for its length but before we managed to read it
109 * fix int/size_t issues found by RedHat/Fedora,
110 https://bugzilla.redhat.com/show_bug.cgi?id=1127310; the fix is
111 different than their fix, but it should accomplish the same thing
112 * convert all code to only do explicit casts after checking boundaries,
113 making the code `-Wconversion`-clean (although that warning is not
119 *released Thu, 30 Apr 2015*
121 Fix memory leaks on some of the error-handling paths of the `get()`
127 *released Fri, 23 May 2014*
129 Small optimisations release:
131 * ari edelkind contributed a speed-up optimisation for handling of files
132 without xattrs (which is, in general, the expected case)
133 * Jonas Borgström contributed a behaviour change to the handling of file
134 names: under Python 3 and up, unicode paths are encoded/decoded using
135 the 'surogatee' handler, instead of the 'strict' handler; while this
136 can hide encoding errors, it mirrors what Python libraries do
137 (e.g. see os.fsencode/fsdecode)
138 * Sean Patrick Santos contributed improvements to the test suite so that
139 it can be used even on files systems which have built-in attributes
140 (e.g. when using SELinux, or NFSv4); to enable this, define the
141 attributes in the TEST_IGNORE_XATTRS environment variable
146 *released Thu, 03 Jan 2013*
148 Bug-fix release. Thanks to Michał Górny, it looked like the library had
149 problem running under pypy, but actually there was a bug in the
150 PyArg_ParseTuple use of et# (signed vs. unsigned, and lack of compiler
151 warnings). This was fixed, and now the test suite passed with many
152 CPython versions and PyPy (version 1.9).
157 *released Wed, 16 May 2012*
159 Bug-fix release. Thanks to Dave Malcolm and his cpychecker tool, a
160 number of significant bugs (refcount leaks and potential NULL-pointer
161 dereferences) have been fixed.
163 Furthermore, compatibility with Python 3 has been improved; this however
164 required changing the meaning of the ``namespace`` argument to the
165 functions: if passed, None is no longer a valid value; pass an empty
166 string if (due to the structure of your program) you have to pass this
167 argument but want to specify no namespace.
169 Also, the project home page has changed from SourceForge to GitHub, and
170 the documentation has been converted from epydoc-based to sphinx.
176 *released Sun, 27 Dec 2009*
178 Implemented support for Python 3. This required a significant change to
179 the C module, hence the new version number.
184 *released Mon, 30 Jun 2008*
189 The old functions ({get,set,list,remove}xattr) are deprecated and replaced with
190 a new API that is namespace-aware and hopefully will allow other OSes (e.g.
191 FreeBSD) to be supported more naturally.
193 Both the old and the new API are supported in the 0.4 versions, however users
194 are encouraged to migrate to the new API.
199 A new bulk get function called get_all() has been added that should be somewhat
200 faster in case of querying files which have many attributes.
205 Since LGPLv3 is not compatible with GPLv2 (which unfortunately I didn't realize
206 before), the license was changed to LGPLv2.1 or later.
211 Unittest coverage was improved.
216 *released Sun, 09 Mar 2008*
218 * changed licence from GPL to LGPL (3 or later)
219 * changed listxattr return type from tuple to a list
220 * developer-related: added unittests
225 *released Sun, 01 Jul 2007*
227 * fixed listing symlink xattrs
232 *released Sat, 11 Feb 2006*
234 * fixed a bug when reading symlink EAs (you weren't able to
236 * fixed a possible memory leak when the actual read of the EA
237 failed but the call to get the length of the EA didn't