NEXT (In development)
---------------------
+BUG FIXES
+~~~~~~~~~
+
+* Perl files should now honour perl system config.
+ (Benedikt Morbach, #1233198)
+
+* Python 3.1 and 3.2 have an inconsistent memoryview implementation which
+ required a workaround for NUL byte detection. (Robert Collins, #1216246)
+
+* The test suite was failing 6 tests due to testtools changing it's output
+ formatting of exceptions. (Robert Collins)
+
+* V2 parser errors now set appropriate mime types for the encapsulated packet
+ data and the error message. (Robert Collins)
+
+* When tests fail to import ``python subunit.run -l ...`` will now write a
+ subunit file attachment listing the failed imports and exit 2, rather than
+ listing the stub objects from the importer and exiting 0.
+ (Robert Collins, #1245672)
+
+IMPROVEMENTS
+~~~~~~~~~~~~
+
+* Most filters will now accept a file path argument instead of only reading
+ from stdin. (Robert Collins, #409206)
+
+0.0.15
+------
+
+BUG FIXES
+~~~~~~~~~
+
+* Clients of subunit did not expect memoryview objects in StreamResult events.
+ (Robert Collins)
+
+* Memoryview and struct were mutually incompatible in 2.7.3 and 3.2.
+ (Robert Collins, #1216163)
+
+0.0.14
+------
+
+BUG FIXES
+~~~~~~~~~
+
+* Memoryview detection was broken and thus it's use was never really tested.
+ (Robert Collins, 1216101)
+
+* TestProtocol2's tag tests were set sort order dependent.
+ (Robert Collins, #1025392)
+
+* TestTestProtocols' test_tags_both was set sort order dependent.
+ (Robert Collins, #1025392)
+
+* TestTestProtocols' test_*_details were dictionary sort order dependent.
+ (Robert Collins, #1025392)
+
+* TestSubUnitTags's test_add_tag was also se sort order dependent.
+ (Robert Collins, #1025392)
+
+0.0.13
+------
+
+IMPROVEMENTS
+~~~~~~~~~~~~
+
+* subunit should now build with automake 1.11 again. (Robert Collins)
+
+* `subunit-stats` no longer outputs encapsulated stdout as subunit.
+ (Robert Collins, #1171987)
+
+* The logic for `subunit.run` is now importable via python -
+ `subunit.run.main`. (Robert Collins, #606770)
+
+BUG FIXES
+~~~~~~~~~
+
+* Removed GPL files that were (C) non Subunit Developers - they are
+ incompatible for binary distribution, which affects redistributors.
+ (Robert Collins, #1185591)
+
+0.0.12
+------
+
+BUG FIXES
+~~~~~~~~~
+
+* Subunit v2 packets with both file content and route code were not being
+ parsed correctly - they would incorrectly emit a parser error, due to trying
+ to parse the route code length from the first byes of the file content.
+ (Robert Collins, 1172815)
+
+0.0.11
+------
+
+v2 protocol draft included in this release. The v2 protocol trades off human
+readability for a massive improvement in robustness, the ability to represent
+concurrent tests in a single stream, cheaper parsing, and that provides
+significantly better in-line debugging support and structured forwarding
+of non-test data (such as stdout or stdin data).
+
+This change includes two new filters (subunit-1to2 and subunit-2to1). Use
+these filters to convert old streams to v2 and convert v2 streams to v1.
+
+All the other filters now only parse and emit v2 streams. V2 is still in
+draft format, so if you want to delay and wait for v2 to be finalised, you
+should use subunit-2to1 before any serialisation steps take place.
+With the ability to encapsulate multiple non-test streams, another significant
+cange is that filters which emit subunit now encapsulate any non-subunit they
+encounter, labelling it 'stdout'. This permits multiplexing such streams and
+detangling the stdout streams from each input.
+
+The subunit libraries (Python etc) have not changed their behaviour: they
+still emit v1 from their existing API calls. New API's are being added
+and applications should migrate once their language has those API's available.
+
+IMPROVEMENTS
+~~~~~~~~~~~~
+
+* ``subunit.run`` now replaces sys.stdout to ensure that stdout is unbuffered
+ - without this pdb output is not reliably visible when stdout is a pipe
+ as it usually is. (Robert Collins)
+
+* v2 protocol draft included in this release. (Python implementation only so
+ far). (Robert Collins)
+
+* Two new Python classes -- ``StreamResultToBytes`` and
+ ``ByteStreamToStreamResult`` handle v2 generation and parsing.
+ (Robert Collins)
+
+0.0.10
+------
+
+BUG FIXES
+~~~~~~~~~
+
+* make_stream_binary is now public for reuse. (Robert Collins)
+
+* NAME was not defined in the protocol BNF. (Robert Collins)
+
+* UnsupportedOperation is available in the Python2.6 io library, so ask
+ forgiveness rather than permission for obtaining it. (Robert Collins)
+
+* Streams with no fileno() attribute are now supported, but they are not
+ checked for being in binary mode: be sure to take care of that if using
+ the library yourself. (Robert Collins)
+
+0.0.9
+-----
+
+BUG FIXES
+~~~~~~~~~
+
+* All the source files are now included in the distribution tarball.
+ (Arfrever Frehtes Taifersar Arahesis, Robert Collins, #996275)
+
+* ``python/subunit/tests/test_run.py`` and ``python/subunit/filters.py`` were
+ not included in the 0.0.8 tarball. (Robert Collins)
+
+* Test ids which include non-ascii unicode characters are now supported.
+ (Robert Collins, #1029866)
+
+* The ``failfast`` option to ``subunit.run`` will now work. The dependency on
+ testtools has been raised to 0.9.23 to permit this.
+ (Robert Collins, #1090582)
+
+0.0.8
+-----
+
+IMPROVEMENTS
+~~~~~~~~~~~~
+
+* Perl module now correctly outputs "failure" instead of "fail". (Stewart Smith)
+
+* Shell functions now output timestamps. (Stewart Smith, Robert Collins)
+
+* 'subunit2csv' script that converts subunit output to CSV format.
+ (Jonathan Lange)
+
+* ``TagCollapsingDecorator`` now correctly distinguishes between local and
+ global tags. (Jonathan Lange)
+
+* ``TestResultFilter`` always forwards ``time:`` events.
+ (Benji York, Brad Crittenden)
+
+BUG FIXES
+~~~~~~~~~
+
+* Add 'subunit --no-xfail', which will omit expected failures from the subunit
+ stream. (John Arbash Meinel, #623642)
+
+* Add 'subunit -F/--only-genuine-failures' which sets all of '--no-skips',
+ '--no-xfail', '--no-passthrough, '--no-success', and gives you just the
+ failure stream. (John Arbash Meinel)
+
+* Python2.6 support was broken by the fixup feature.
+ (Arfrever Frehtes Taifersar Arahesis, #987490)
+
+* Python3 support regressed in trunk.
+ (Arfrever Frehtes Taifersar Arahesis, #987514)
+
+* Python3 support was insufficiently robust in detecting unicode streams.
+ (Robert Collins, Arfrever Frehtes Taifersar Arahesis)
+
+* Tag support has been implemented for TestProtocolClient.
+ (Robert Collins, #518016)
+
+* Tags can now be filtered. (Jonathan Lange, #664171)
+
+* Test suite works with latest testtools (but not older ones - formatting
+ changes only). (Robert Collins)
+
+0.0.7
+-----
+
The Subunit Python test runner ``python -m subunit.run`` can now report the
test ids and also filter via a test id list file thanks to improvements in
``testtools.run``. See the testtools manual, or testrepository - a major
user of such functionality.
+Additionally the protocol now has a keyword uxsuccess for Unexpected Success
+reporting. Older parsers will report tests with this status code as 'lost
+connection'.
+
IMPROVEMENTS
~~~~~~~~~~~~
+* Add ``TimeCollapsingDecorator`` which collapses multiple sequential time()
+ calls into just the first and last. (Jonathan Lange)
+
+* Add ``TagCollapsingDecorator`` which collapses many tags() calls into one
+ where possible. (Jonathan Lange, Robert Collins)
+
+* Force flush of writes to stdout in c/tests/test_child.
+ (Jelmer Vernooij, #687611)
+
+* Provisional Python 3.x support.
+ (Robert Collins, Tres Seaver, Martin[gz], #666819)
+
+* ``subunit.chunked.Decoder`` Python class takes a new ``strict`` option,
+ which defaults to ``True``. When ``False``, the ``Decoder`` will accept
+ incorrect input that is still unambiguous. i.e. subunit will not barf if
+ a \r is missing from the input. (Martin Pool)
+
* ``subunit-filter`` preserves the relative ordering of ``time:`` statements,
so you can now use filtered streams to gather data about how long it takes
to run a test. (Jonathan Lange, #716554)
+* ``subunit-ls`` now handles a stream with time: instructions that start
+ partway through the stream (which may lead to strange times) more gracefully.
+ (Robert Collins, #785954)
+
+* ``subunit-ls`` should handle the new test outcomes in Python2.7 better.
+ (Robert Collins, #785953)
+
+* ``TestResultFilter`` now collapses sequential calls to time().
+ (Jonathan Lange, #567150)
+
+* ``TestResultDecorator.tags()`` now actually works, and is no longer a buggy
+ copy/paste of ``TestResultDecorator.time()``. (Jonathan Lange, #681828)
+
+* ``TestResultFilter`` now supports a ``fixup_expected_failures``
+ argument. (Jelmer Vernooij, #755241)
+
* The ``subunit.run`` Python module supports ``-l`` and ``--load-list`` as
per ``testtools.run``. This required a dependency bump due to a small
API change in ``testtools``. (Robert Collins)
-* Force flush of writes to stdout in c/tests/test_child. (Jelmer Vernooij, #687611)
-
* The help for subunit-filter was confusing about the behaviour of ``-f`` /
``--no-failure``. (Robert Collins, #703392)
-* ``subunit.chunked.Decoder`` Python class takes a new ``strict`` option,
- which defaults to ``True``. When ``False``, the ``Decoder`` will accept
- incorrect input that is still unambiguous. i.e. subunit will not barf if
- a \r is missing from the input. (Martin Pool)
+* The Python2.7 / testtools addUnexpectedSuccess API is now supported. This
+ required adding a new status code to the protocol. (Robert Collins, #654474)
+
+CHANGES
+~~~~~~~
+
+* testtools 0.9.11 or newer is new needed (due to the Python 3 support).
+ (Robert Collins)
0.0.6
-----
git.samba.org / third_party / subunit / blobdiff