Before we didn't wait until the inspector window was hidden, which caused the
following tests to fail on our OS X buildbot:
==================================== ERRORS ====================================
_________ ERROR at teardown of test_fakekey_sending_key_to_the_website _________
[...]
tests/integration/testprocess.py:186: InvalidLine
----------------------------- Captured stdout call -----------------------------
[...]
INVALID: Traceback (most recent call last):
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/app.py", line 109, in <lambda>
IGNORED: target_arg=target_arg))
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/app.py", line 263, in process_pos_args
IGNORED: commandrunner.run_safely_init(cmd[1:])
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/commands/runners.py", line 266, in run_safely_init
IGNORED: self.run(text, count)
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/commands/runners.py", line 249, in run
IGNORED: result.cmd.run(self._win_id, args)
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/commands/command.py", line 505, in run
IGNORED: self.handler(*posargs, **kwargs)
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/browser/commands.py", line 1781, in fake_key
IGNORED: tab='current')
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/utils/objreg.py", line 215, in get
IGNORED: reg = _get_registry(scope, window, tab)
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/utils/objreg.py", line 202, in _get_registry
IGNORED: return _get_tab_registry(window, tab)
IGNORED: File "/Users/buildbot/buildbot/slave/osx/build/qutebrowser/utils/objreg.py", line 152, in _get_tab_registry
IGNORED: raise RegistryUnavailableError('tab')
INVALID: qutebrowser.utils.objreg.RegistryUnavailableError: tab
[...]
While this is a real issue, this test shouldn't affect the next one.
Since we have to wait for windows getting focused but Xvfb doesn't do that, we
skip this tests when running via Xvfb.
This worked before b0d4ebe because --debug was given and the
"args.loglevel.upper()" part was never executed.
However, loglevel is expected as a string, not as an int.
It's now based by the browser opening a second page after :stop is run rather
than timing-based, which hopefully should make it run more stable on Travis.
Instead, make it hashable and use collections.Counter instead of sorted
lists. This actually produces a nicer output for differences between
the counters.
webview uses QUrl to output the url and thus might convert some escape
sequences to their unicode characters. If we don't do the same
conversions here, we will "miss" the log message and fail the test.
When we don't include the start time in the measurement, the subprocess will
already start sleeping before we start measuring, which can lead to times
measured which are around 260-300ms.
Closes#1077.
There were two issues:
- Paths for coverage.py suddenly were absolute instead of relative, so we strip
the common base part if that happens.
/home/florian/proj/qutebrowser/git/qutebrowser/browser/cookies.py has 100%
coverage but is not in perfect_files!
/home/florian/proj/qutebrowser/git/qutebrowser/browser/http.py has 100%
coverage but is not in perfect_files!
[...]
- There was an io.UnsupportedOperationError in test_check_coverage.py because
of pytest-faulthandler, so we just don't load that in the subprocess.
This also makes the output of MHTMLWriter deterministic, by
1) Setting the boundary at object creation, allowing uuid.uuid4 to be
monkey patched
2) Outputting the files in sorted order (sorted by location), as python
dicts are unordered by default.
With this change, no test should affect another one anymore. Changes in detail:
- Before each test, invalid lines are cleaned so the invalid output from the
previous test doesn't affect the next one.
- Before each test, qutebrowser is restarted if it was quit.
- After each test, the data is cleared properly in every case.
- If there was an invalid output, the test waits for 1s for more output, and
then terminates qutebrowser so it's restarted with a clean state.
We need to search for lines in the history because we could miss something
otherwise, but for subsequent wait_for calls, we really don't want to wait for
the same thing again.
This should make test_backforward.py more stable as it *actually* waits when
going back now. Before, it did produce failures such as this one on OS X:
____________________________ test_going_backforward ____________________________
[..]
@bdd.then(bdd.parsers.parse("The requests should be:\n{pages}"))
def list_of_loaded_pages(httpbin, pages):
requests = [httpbin.Request('GET', '/' + path.strip())
for path in pages.split('\n')]
> assert httpbin.get_requests() == requests
E assert [Request(verb...rward/1.txt')] == [Request(verb=...rward/2.txt')]
E At index 3 diff: Request(verb='GET', path='/data/backforward/1.txt') != Request(verb='GET', path='/data/backforward/2.txt')
E Full diff:
E [Request(verb='GET', path='/data/backforward/1.txt'),
E Request(verb='GET', path='/data/backforward/2.txt'),
E Request(verb='GET', path='/data/backforward/1.txt'),
E - Request(verb='GET', path='/data/backforward/1.txt')]
E ? ^
E + Request(verb='GET', path='/data/backforward/2.txt')]
E ? ^
tests/integration/features/conftest.py:85: AssertionError
In test_cache_existing_metadata_file() we are now getting the correct
path to the metadata files, thus making the test useful. This was the
last missing test, the cache is now 100% covered (issue#999).
