6
The Importance of Testing
7
=========================
9
Reliability is a critical success factor for any Version Control System.
10
We want Bazaar to be highly reliable across multiple platforms while
11
evolving over time to meet the needs of its community.
13
In a nutshell, this is what we expect and encourage:
15
* New functionality should have test cases. Preferably write the
16
test before writing the code.
18
In general, you can test at either the command-line level or the
19
internal API level. See `Writing tests`_ below for more detail.
21
* Try to practice Test-Driven Development: before fixing a bug, write a
22
test case so that it does not regress. Similarly for adding a new
23
feature: write a test case for a small version of the new feature before
24
starting on the code itself. Check the test fails on the old code, then
25
add the feature or fix and check it passes.
27
By doing these things, the Bazaar team gets increased confidence that
28
changes do what they claim to do, whether provided by the core team or
29
by community members. Equally importantly, we can be surer that changes
30
down the track do not break new features or bug fixes that you are
33
As of September 2009, Bazaar ships with a test suite containing over
34
23,000 tests and growing. We are proud of it and want to remain so. As
35
community members, we all benefit from it. Would you trust version control
36
on your project to a product *without* a test suite like Bazaar has?
39
Running the Test Suite
40
======================
42
As of Bazaar 2.1, you must have the testtools_ library installed to run
45
.. _testtools: https://launchpad.net/testtools/
47
To test all of Bazaar, just run::
51
With ``--verbose`` bzr will print the name of every test as it is run.
53
This should always pass, whether run from a source tree or an installed
54
copy of Bazaar. Please investigate and/or report any failures.
57
Running particular tests
58
------------------------
60
Currently, bzr selftest is used to invoke tests.
61
You can provide a pattern argument to run a subset. For example,
62
to run just the blackbox tests, run::
64
./bzr selftest -v blackbox
66
To skip a particular test (or set of tests), use the --exclude option
67
(shorthand -x) like so::
69
./bzr selftest -v -x blackbox
71
To ensure that all tests are being run and succeeding, you can use the
72
--strict option which will fail if there are any missing features or known
75
./bzr selftest --strict
77
To list tests without running them, use the --list-only option like so::
79
./bzr selftest --list-only
81
This option can be combined with other selftest options (like -x) and
82
filter patterns to understand their effect.
84
Once you understand how to create a list of tests, you can use the --load-list
85
option to run only a restricted set of tests that you kept in a file, one test
86
id by line. Keep in mind that this will never be sufficient to validate your
87
modifications, you still need to run the full test suite for that, but using it
88
can help in some cases (like running only the failed tests for some time)::
90
./bzr selftest -- load-list my_failing_tests
92
This option can also be combined with other selftest options, including
93
patterns. It has some drawbacks though, the list can become out of date pretty
94
quick when doing Test Driven Development.
96
To address this concern, there is another way to run a restricted set of tests:
97
the --starting-with option will run only the tests whose name starts with the
98
specified string. It will also avoid loading the other tests and as a
99
consequence starts running your tests quicker::
101
./bzr selftest --starting-with bzrlib.blackbox
103
This option can be combined with all the other selftest options including
104
--load-list. The later is rarely used but allows to run a subset of a list of
105
failing tests for example.
110
To test only the bzr core, ignoring any plugins you may have installed,
113
./bzr --no-plugins selftest
115
Disabling crash reporting
116
-------------------------
118
By default Bazaar uses apport_ to report program crashes. In developing
119
Bazaar it's normal and expected to have it crash from time to time, at
120
least because a test failed if for no other reason.
122
Therefore you should probably add ``debug_flags = no_apport`` to your
123
``bazaar.conf`` file (in ``~/.bazaar/`` on Unix), so that failures just
124
print a traceback rather than writing a crash file.
126
.. _apport: https://launchpad.net/apport/
129
Test suite debug flags
130
----------------------
132
Similar to the global ``-Dfoo`` debug options, bzr selftest accepts
133
``-E=foo`` debug flags. These flags are:
135
:allow_debug: do *not* clear the global debug flags when running a test.
136
This can provide useful logging to help debug test failures when used
137
with e.g. ``bzr -Dhpss selftest -E=allow_debug``
139
Note that this will probably cause some tests to fail, because they
140
don't expect to run with any debug flags on.
146
Bazaar can optionally produce output in the machine-readable subunit_
147
format, so that test output can be post-processed by various tools.
149
.. _subunit: https://launchpad.net/subunit/
156
Normally you should add or update a test for all bug fixes or new features
160
Where should I put a new test?
161
------------------------------
163
Bzrlib's tests are organised by the type of test. Most of the tests in
164
bzr's test suite belong to one of these categories:
167
- Blackbox (UI) tests
168
- Per-implementation tests
171
A quick description of these test types and where they belong in bzrlib's
172
source follows. Not all tests fall neatly into one of these categories;
173
in those cases use your judgement.
179
Unit tests make up the bulk of our test suite. These are tests that are
180
focused on exercising a single, specific unit of the code as directly
181
as possible. Each unit test is generally fairly short and runs very
184
They are found in ``bzrlib/tests/test_*.py``. So in general tests should
185
be placed in a file named test_FOO.py where FOO is the logical thing under
188
For example, tests for merge3 in bzrlib belong in bzrlib/tests/test_merge3.py.
189
See bzrlib/tests/test_sampler.py for a template test script.
195
Tests can be written for the UI or for individual areas of the library.
196
Choose whichever is appropriate: if adding a new command, or a new command
197
option, then you should be writing a UI test. If you are both adding UI
198
functionality and library functionality, you will want to write tests for
199
both the UI and the core behaviours. We call UI tests 'blackbox' tests
200
and they belong in ``bzrlib/tests/blackbox/*.py``.
202
When writing blackbox tests please honour the following conventions:
204
1. Place the tests for the command 'name' in
205
bzrlib/tests/blackbox/test_name.py. This makes it easy for developers
206
to locate the test script for a faulty command.
208
2. Use the 'self.run_bzr("name")' utility function to invoke the command
209
rather than running bzr in a subprocess or invoking the
210
cmd_object.run() method directly. This is a lot faster than
211
subprocesses and generates the same logging output as running it in a
212
subprocess (which invoking the method directly does not).
214
3. Only test the one command in a single test script. Use the bzrlib
215
library when setting up tests and when evaluating the side-effects of
216
the command. We do this so that the library api has continual pressure
217
on it to be as functional as the command line in a simple manner, and
218
to isolate knock-on effects throughout the blackbox test suite when a
219
command changes its name or signature. Ideally only the tests for a
220
given command are affected when a given command is changed.
222
4. If you have a test which does actually require running bzr in a
223
subprocess you can use ``run_bzr_subprocess``. By default the spawned
224
process will not load plugins unless ``--allow-plugins`` is supplied.
227
Per-implementation tests
228
~~~~~~~~~~~~~~~~~~~~~~~~
230
Per-implementation tests are tests that are defined once and then run
231
against multiple implementations of an interface. For example,
232
``per_transport.py`` defines tests that all Transport implementations
233
(local filesystem, HTTP, and so on) must pass. They are found in
234
``bzrlib/tests/per_*/*.py``, and ``bzrlib/tests/per_*.py``.
236
These are really a sub-category of unit tests, but an important one.
238
Along the same lines are tests for extension modules. We generally have
239
both a pure-python and a compiled implementation for each module. As such,
240
we want to run the same tests against both implementations. These can
241
generally be found in ``bzrlib/tests/*__*.py`` since extension modules are
242
usually prefixed with an underscore. Since there are only two
243
implementations, we have a helper function
244
``bzrlib.tests.permute_for_extension``, which can simplify the
245
``load_tests`` implementation.
251
We make selective use of doctests__. In general they should provide
252
*examples* within the API documentation which can incidentally be tested. We
253
don't try to test every important case using doctests |--| regular Python
254
tests are generally a better solution. That is, we just use doctests to
255
make our documentation testable, rather than as a way to make tests.
257
Most of these are in ``bzrlib/doc/api``. More additions are welcome.
259
__ http://docs.python.org/lib/module-doctest.html
265
``bzrlib/tests/script.py`` allows users to write tests in a syntax very close to a shell session,
266
using a restricted and limited set of commands that should be enough to mimic
267
most of the behaviours.
269
A script is a set of commands, each command is composed of:
271
* one mandatory command line,
272
* one optional set of input lines to feed the command,
273
* one optional set of output expected lines,
274
* one optional set of error expected lines.
276
Input, output and error lines can be specified in any order.
278
Except for the expected output, all lines start with a special
279
string (based on their origin when used under a Unix shell):
281
* '$ ' for the command,
283
* nothing for output,
286
Comments can be added anywhere, they start with '#' and end with
289
The execution stops as soon as an expected output or an expected error is not
292
When no output is specified, any ouput from the command is accepted
293
and execution continue.
295
If an error occurs and no expected error is specified, the execution stops.
297
An error is defined by a returned status different from zero, not by the
298
presence of text on the error stream.
300
The matching is done on a full string comparison basis unless '...' is used, in
301
which case expected output/errors can be less precise.
305
The following will succeeds only if 'bzr add' outputs 'adding file'::
310
If you want the command to succeed for any output, just use::
314
The following will stop with an error::
318
If you want it to succeed, use::
321
2> bzr: ERROR: unknown command "not-a-command"
323
You can use ellipsis (...) to replace any piece of text you don't want to be
326
$ bzr branch not-a-branch
327
2>bzr: ERROR: Not a branch...not-a-branch/".
329
This can be used to ignore entire lines too::
335
# And here we explain that surprising fourth line
342
You can check the content of a file with cat::
347
You can also check the existence of a file with cat, the following will fail if
348
the file doesn't exist::
352
The actual use of ScriptRunner within a TestCase looks something like
355
def test_unshelve_keep(self):
358
sr.run_script(self, '''
360
$ bzr shelve --all -m Foo
363
$ bzr unshelve --keep
374
`bzrlib.tests.test_import_tariff` has some tests that measure how many
375
Python modules are loaded to run some representative commands.
377
We want to avoid loading code unnecessarily, for reasons including:
379
* Python modules are interpreted when they're loaded, either to define
380
classes or modules or perhaps to initialize some structures.
382
* With a cold cache we may incur blocking real disk IO for each module.
384
* Some modules depend on many others.
386
* Some optional modules such as `testtools` are meant to be soft
387
dependencies and only needed for particular cases. If they're loaded in
388
other cases then bzr may break for people who don't have those modules.
390
`test_import_tarrif` allows us to check that removal of imports doesn't
393
This is done by running the command in a subprocess with
394
``--profile-imports``. Starting a whole Python interpreter is pretty
395
slow, so we don't want exhaustive testing here, but just enough to guard
396
against distinct fixed problems.
398
Assertions about precisely what is loaded tend to be brittle so we instead
399
make assertions that particular things aren't loaded.
401
Unless selftest is run with ``--no-plugins``, modules will be loaded in
402
the usual way and checks made on what they cause to be loaded. This is
403
probably worth checking into, because many bzr users have at least some
404
plugins installed (and they're included in binary installers).
406
In theory, plugins might have a good reason to load almost anything:
407
someone might write a plugin that opens a network connection or pops up a
408
gui window every time you run 'bzr status'. However, it's more likely
409
that the code to do these things is just being loaded accidentally. We
410
might eventually need to have a way to make exceptions for particular
413
Some things to check:
415
* non-GUI commands shouldn't load GUI libraries
417
* operations on bzr native formats sholudn't load foreign branch libraries
419
* network code shouldn't be loaded for purely local operations
421
* particularly expensive Python built-in modules shouldn't be loaded
422
unless there is a good reason
428
In our enhancements to unittest we allow for some addition results beyond
429
just success or failure.
431
If a test can't be run, it can say that it's skipped by raising a special
432
exception. This is typically used in parameterized tests |--| for example
433
if a transport doesn't support setting permissions, we'll skip the tests
434
that relating to that. ::
437
return self.branch_format.initialize(repo.bzrdir)
438
except errors.UninitializableFormat:
439
raise tests.TestSkipped('Uninitializable branch format')
441
Raising TestSkipped is a good idea when you want to make it clear that the
442
test was not run, rather than just returning which makes it look as if it
445
Several different cases are distinguished:
448
Generic skip; the only type that was present up to bzr 0.18.
451
The test doesn't apply to the parameters with which it was run.
452
This is typically used when the test is being applied to all
453
implementations of an interface, but some aspects of the interface
454
are optional and not present in particular concrete
455
implementations. (Some tests that should raise this currently
456
either silently return or raise TestSkipped.) Another option is
457
to use more precise parameterization to avoid generating the test
461
The test can't be run because a dependency (typically a Python
462
library) is not available in the test environment. These
463
are in general things that the person running the test could fix
464
by installing the library. It's OK if some of these occur when
465
an end user runs the tests or if we're specifically testing in a
466
limited environment, but a full test should never see them.
468
See `Test feature dependencies`_ below.
471
The test exists but is known to fail, for example this might be
472
appropriate to raise if you've committed a test for a bug but not
473
the fix for it, or if something works on Unix but not on Windows.
475
Raising this allows you to distinguish these failures from the
476
ones that are not expected to fail. If the test would fail
477
because of something we don't expect or intend to fix,
478
KnownFailure is not appropriate, and TestNotApplicable might be
481
KnownFailure should be used with care as we don't want a
482
proliferation of quietly broken tests.
486
We plan to support three modes for running the test suite to control the
487
interpretation of these results. Strict mode is for use in situations
488
like merges to the mainline and releases where we want to make sure that
489
everything that can be tested has been tested. Lax mode is for use by
490
developers who want to temporarily tolerate some known failures. The
491
default behaviour is obtained by ``bzr selftest`` with no options, and
492
also (if possible) by running under another unittest harness.
494
======================= ======= ======= ========
495
result strict default lax
496
======================= ======= ======= ========
497
TestSkipped pass pass pass
498
TestNotApplicable pass pass pass
499
UnavailableFeature fail pass pass
500
KnownFailure fail pass pass
501
======================= ======= ======= ========
504
Test feature dependencies
505
-------------------------
507
Writing tests that require a feature
508
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
510
Rather than manually checking the environment in each test, a test class
511
can declare its dependence on some test features. The feature objects are
512
checked only once for each run of the whole test suite.
514
(For historical reasons, as of May 2007 many cases that should depend on
515
features currently raise TestSkipped.)
519
class TestStrace(TestCaseWithTransport):
521
_test_needs_features = [StraceFeature]
523
This means all tests in this class need the feature. If the feature is
524
not available the test will be skipped using UnavailableFeature.
526
Individual tests can also require a feature using the ``requireFeature``
529
self.requireFeature(StraceFeature)
531
The old naming style for features is CamelCase, but because they're
532
actually instances not classses they're now given instance-style names
535
Features already defined in ``bzrlib.tests`` and ``bzrlib.tests.features``
543
- UnicodeFilenameFeature
545
- CaseInsensitiveFilesystemFeature.
548
Defining a new feature that tests can require
549
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
551
New features for use with ``_test_needs_features`` or ``requireFeature``
552
are defined by subclassing ``bzrlib.tests.Feature`` and overriding the
553
``_probe`` and ``feature_name`` methods. For example::
555
class _SymlinkFeature(Feature):
558
return osutils.has_symlinks()
560
def feature_name(self):
563
SymlinkFeature = _SymlinkFeature()
565
A helper for handling running tests based on whether a python
566
module is available. This can handle 3rd-party dependencies (is
567
``paramiko`` available?) as well as stdlib (``termios``) or
568
extension modules (``bzrlib._groupcompress_pyx``). You create a
569
new feature instance with::
571
# in bzrlib/tests/features.py
572
apport = tests.ModuleAvailableFeature('apport')
575
# then in bzrlib/tests/test_apport.py
576
class TestApportReporting(TestCaseInTempDir):
578
_test_needs_features = [features.apport]
581
Testing exceptions and errors
582
-----------------------------
584
It's important to test handling of errors and exceptions. Because this
585
code is often not hit in ad-hoc testing it can often have hidden bugs --
586
it's particularly common to get NameError because the exception code
587
references a variable that has since been renamed.
589
.. TODO: Something about how to provoke errors in the right way?
591
In general we want to test errors at two levels:
593
1. A test in ``test_errors.py`` checking that when the exception object is
594
constructed with known parameters it produces an expected string form.
595
This guards against mistakes in writing the format string, or in the
596
``str`` representations of its parameters. There should be one for
597
each exception class.
599
2. Tests that when an api is called in a particular situation, it raises
600
an error of the expected class. You should typically use
601
``assertRaises``, which in the Bazaar test suite returns the exception
602
object to allow you to examine its parameters.
604
In some cases blackbox tests will also want to check error reporting. But
605
it can be difficult to provoke every error through the commandline
606
interface, so those tests are only done as needed |--| eg in response to a
607
particular bug or if the error is reported in an unusual way(?) Blackbox
608
tests should mostly be testing how the command-line interface works, so
609
should only test errors if there is something particular to the cli in how
610
they're displayed or handled.
616
The Python ``warnings`` module is used to indicate a non-fatal code
617
problem. Code that's expected to raise a warning can be tested through
620
The test suite can be run with ``-Werror`` to check no unexpected errors
623
However, warnings should be used with discretion. It's not an appropriate
624
way to give messages to the user, because the warning is normally shown
625
only once per source line that causes the problem. You should also think
626
about whether the warning is serious enought that it should be visible to
627
users who may not be able to fix it.
630
Interface implementation testing and test scenarios
631
---------------------------------------------------
633
There are several cases in Bazaar of multiple implementations of a common
634
conceptual interface. ("Conceptual" because it's not necessary for all
635
the implementations to share a base class, though they often do.)
636
Examples include transports and the working tree, branch and repository
639
In these cases we want to make sure that every implementation correctly
640
fulfils the interface requirements. For example, every Transport should
641
support the ``has()`` and ``get()`` and ``clone()`` methods. We have a
642
sub-suite of tests in ``test_transport_implementations``. (Most
643
per-implementation tests are in submodules of ``bzrlib.tests``, but not
644
the transport tests at the moment.)
646
These tests are repeated for each registered Transport, by generating a
647
new TestCase instance for the cross product of test methods and transport
648
implementations. As each test runs, it has ``transport_class`` and
649
``transport_server`` set to the class it should test. Most tests don't
650
access these directly, but rather use ``self.get_transport`` which returns
651
a transport of the appropriate type.
653
The goal is to run per-implementation only the tests that relate to that
654
particular interface. Sometimes we discover a bug elsewhere that happens
655
with only one particular transport. Once it's isolated, we can consider
656
whether a test should be added for that particular implementation,
657
or for all implementations of the interface.
659
The multiplication of tests for different implementations is normally
660
accomplished by overriding the ``load_tests`` function used to load tests
661
from a module. This function typically loads all the tests, then applies
662
a TestProviderAdapter to them, which generates a longer suite containing
663
all the test variations.
665
See also `Per-implementation tests`_ (above).
671
Some utilities are provided for generating variations of tests. This can
672
be used for per-implementation tests, or other cases where the same test
673
code needs to run several times on different scenarios.
675
The general approach is to define a class that provides test methods,
676
which depend on attributes of the test object being pre-set with the
677
values to which the test should be applied. The test suite should then
678
also provide a list of scenarios in which to run the tests.
680
Typically ``multiply_tests_from_modules`` should be called from the test
681
module's ``load_tests`` function.
687
We have a rich collection of tools to support writing tests. Please use
688
them in preference to ad-hoc solutions as they provide portability and
689
performance benefits.
692
TestCase and its subclasses
693
~~~~~~~~~~~~~~~~~~~~~~~~~~~
695
The ``bzrlib.tests`` module defines many TestCase classes to help you
699
A base TestCase that extends the Python standard library's
700
TestCase in several ways. It adds more assertion methods (e.g.
701
``assertContainsRe``), ``addCleanup``, and other features (see its API
702
docs for details). It also has a ``setUp`` that makes sure that
703
global state like registered hooks and loggers won't interfere with
704
your test. All tests should use this base class (whether directly or
707
TestCaseWithMemoryTransport
708
Extends TestCase and adds methods like ``get_transport``,
709
``make_branch`` and ``make_branch_builder``. The files created are
710
stored in a MemoryTransport that is discarded at the end of the test.
711
This class is good for tests that need to make branches or use
712
transports, but that don't require storing things on disk. All tests
713
that create bzrdirs should use this base class (either directly or via
714
a subclass) as it ensures that the test won't accidentally operate on
715
real branches in your filesystem.
718
Extends TestCaseWithMemoryTransport. For tests that really do need
719
files to be stored on disk, e.g. because a subprocess uses a file, or
720
for testing functionality that accesses the filesystem directly rather
721
than via the Transport layer (such as dirstate).
723
TestCaseWithTransport
724
Extends TestCaseInTempDir. Provides ``get_url`` and
725
``get_readonly_url`` facilities. Subclasses can control the
726
transports used by setting ``vfs_transport_factory``,
727
``transport_server`` and/or ``transport_readonly_server``.
730
See the API docs for more details.
736
When writing a test for a feature, it is often necessary to set up a
737
branch with a certain history. The ``BranchBuilder`` interface allows the
738
creation of test branches in a quick and easy manner. Here's a sample
741
builder = self.make_branch_builder('relpath')
742
builder.build_commit()
743
builder.build_commit()
744
builder.build_commit()
745
branch = builder.get_branch()
747
``make_branch_builder`` is a method of ``TestCaseWithMemoryTransport``.
749
Note that many current tests create test branches by inheriting from
750
``TestCaseWithTransport`` and using the ``make_branch_and_tree`` helper to
751
give them a ``WorkingTree`` that they can commit to. However, using the
752
newer ``make_branch_builder`` helper is preferred, because it can build
753
the changes in memory, rather than on disk. Tests that are explictly
754
testing how we work with disk objects should, of course, use a real
757
Please see bzrlib.branchbuilder for more details.
759
If you're going to examine the commit timestamps e.g. in a test for log
760
output, you should set the timestamp on the tree, rather than using fuzzy
767
The ``TreeBuilder`` interface allows the construction of arbitrary trees
768
with a declarative interface. A sample session might look like::
770
tree = self.make_branch_and_tree('path')
771
builder = TreeBuilder()
772
builder.start_tree(tree)
773
builder.build(['foo', "bar/", "bar/file"])
774
tree.commit('commit the tree')
775
builder.finish_tree()
777
Usually a test will create a tree using ``make_branch_and_memory_tree`` (a
778
method of ``TestCaseWithMemoryTransport``) or ``make_branch_and_tree`` (a
779
method of ``TestCaseWithTransport``).
781
Please see bzrlib.treebuilder for more details.
784
.. |--| unicode:: U+2014
added
removed
doc/developers/testing.txt
