| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| |
|
|
| |
Pylint before 1.4.4 reads our version from there, let's not break it.
|
| | |
|
| | |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
#294756)
Virtualenv symlinks a bunch of stdlib modules which confuse
modutils.is_standard_module. The better way would obviously be to fix
modutils to properly detect virtualenvs, but life is too short.
As for why virtualenv actually needs to symlink a good chunk of the
stdlib, anyone's guess is as good as mine. Here's the upstream commit
that added hashlib.py to the list of needed symlinks:
https://github.com/pypa/virtualenv/commit/10ba3f3cc0d7a254dd72e2bb9557006da2f85e00#diff-7e83770aa980bd4327db90f4eafeffdfR126
And for the record, 'venv' (the virtualenv-lookalike shipped since
python 3.3) does not do _any_ symlink to stdlib modules.
|
| |
|
|
|
| |
Unsupported since the very first few commits of the 0.63.0 release
(38c568b29cf3 to be precise).
|
| | |
|
| | |
|
| | |
|
| |
|
|
| |
It's spring^Wsummer cleaning time.
|
| |
|
|
|
| |
setup_module() and teardown_module() support have been deprecated 4
years ago in commit 06755c704e60.
|
| | |
|
| |
|
|
| |
Closes #105845.
|
| |
|
|
|
|
|
|
| |
type(x | y) where type(x) is a subclass of set is now set, not type(x).
See http://bugs.python.org/issue1721812
Related to #105845.
|
| |
|
|
|
|
| |
The latter does not exist in python 3.
Related to #105845.
|
| |
|
|
|
|
|
|
|
| |
Python 3 no longer has the concept of unbound methods, so we can't get
from the method to the class defining it. This is only an issue in the
test, not in actual usage, because does_match_tags should only ever deal
with instantiated TestCases.
Related to #105845
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Quoting `pydoc __import__`: "When importing a module from a package,
note that __import__('A.B', ...) returns package A".
`importlib.import_module` will always return module A.B in this respect.
The underlying problem is that, when running test using something else
than pytest this `__import__` call will generate wrong results. For
instance, considering from CubicWeb tests, running:
python -m unittest cubicweb.test.unittest_cwconfig.CubicWebConfigurationTC.test_appobjects_path
__import__('cubicweb.test.unittest_cwconfig') will return the 'cubicweb'
module, not 'cubicweb.test.unittest_cwconfig'.
Since the module is imported (`cls` is a subclass coming from it), we
can just look it up in sys.modules.
This is part of an attempt to use something else than pytest to run
tests in CubicWeb.
|
| | |
|
| |
|
|
|
|
| |
If everything goes well, it doesn't matter, nobody will look at the log.
If something goes wrong, -q will just result in missing information and
a frustrated developer. Frustration can kill, let's save some kittens.
|
| |
|
|
|
| |
- stop installing __pkginfo__ (move it back to root dir)
- adapt all packaging to use setuptools
|
| | |
|
| |
|
|
|
| |
BUILTIN_MODULES could actually be just a set, but I'm not sure what the
API implications are, so let's keep it as a dict.
|
| |
|
|
|
|
| |
Nested namespaces (zope is a big fan) probably don't work.
Closes #122443.
|
| | |
|
| | |
|
| |
|
|
|
| |
Stop using lgc's own __pkginfo__ (which we may stop shipping/installing
at some point).
|
| |
|
|
|
|
| |
The logilab.common package now lives in a logilab/common directory to make it pip compliant.
Related to #294479.
|
| |
|
|
| |
logilab-common-debian-version-0.63.2-1, logilab-common-centos-version-0.63.2-1 for changeset dddc0f200882
|
| | |
|
| |
|
|
| |
(closes #281909)
|
| |
|
|
|
|
|
|
| |
The refactoring of the coverage suspend/resume tools in pytest done in
31725b8fa3f5 breaks the compatibility of theses tools with coverage.py,
due to a bug in this later project, see:
https://bitbucket.org/ned/coveragepy/issue/123
|
| |
|
|
| |
logilab-common-debian-version-0.63.1-1, logilab-common-centos-version-0.63.1-1 for changeset 0f392891a907
|
| |
|
|
| |
required to run tests, hence to be able to build the deb, on jessie
|
| |
|
|
| |
It was removed in 0.63.
|
| | |
|
| |
|
|
| |
Closes #278550
|
| |
|
|
| |
No reason to hide the pid file / dir behind restrictive permissions.
|
| |
|
|
|
|
|
|
| |
Dates with a tzinfo get messed up by utcdatetime.
The unittest in this patch depends on pytz.
Closes #280794
|
| |
|
|
|
|
|
| |
The former implementation was not restoring properly the trace function on python 2.7 at least.
This cleaner implementation uses context-manager, and deprecates the pause_tracing/resume_tracing couple.
Closes #280757
|
| |
|
|
|
|
|
|
|
|
|
| |
_ts_wrapped_run is a modified version of the upstream
unittest.suite.TestSuite.run method,
but the recent (python 2.7) evolutions of this later had never been 'integrated'
in lgc.
This mismatch broke tests using a tearDownModule function.
This patch is... humm, sorry...
|
| | |
|
| | |
|
| |
|
|
| |
This file uses the Python 3-style print function, so >> won't work.
|
| |
|
|
| |
logilab-common-debian-version-0.63.0-1, logilab-common-centos-version-0.63.0-1 for changeset 468b7a368a0d
|
| | |
|
| |
|
|
|
| |
Add six build-dep so we can run tests, and build the lib where
doc/makefile expects it before building API docs.
|
| | |
|
| |
|
|
|
|
| |
Instead of having two copies of part of the packaging, that don't stay
in sync, just keep one, and only build the python3 version of the
package if we have new enough python3.
|
| | |
|
| | |
|
