| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
open files. (Necessary on Windows for os.remove() to succeed)
|
| |
|
| |
|
|
|
|
| |
safely reverted now.
|
|
|
|
|
|
| |
completely
equal (the disable() method is reported on a differently-named class).
|
|
|
|
| |
This contains fixes to allow building C extensions on Darwin, and a few extra C APIs as well.
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
By default, 'import foo' will search extension modules named
foo.pypy-14.so
This is an implementation of PEP 3149, with two differences:
- the default 'foo.so' is not considered
- the feature is available on Windows ('foo.pypy-14.pyd')
The tag name can be specified at transation time, with the option --soabi:
--soabi=pypy-14
is the default.
It it not recommmended to set:
--soabi=
this will disable the feature, allow the pypy interpreter to import 'foo.so',
and probably crash when this file has been compiled for another interpreter...
|
| |
|
| |
|
|
|
|
| |
with slots without __weakref__
|
| |
|
|
|
|
|
|
| |
The comment (re-appearing because of the revert) explains why.
I'm not sure I understand why it was un-skipped but nothing
at all was done to fix the massive number of failures.
|
| |
|
|
|
|
| |
now fully pass.
|
| |
|
| |
|
|
|
|
| |
segfault. Made with at the Europython sprint with arigo's help.
|
| |
|
| |
|
|
|
|
| |
'import site' in the case we are using the builtin path
|
|
|
|
|
|
|
|
|
| |
automatically copied there from cpyext/include during translation. The
generated pypy_decl.h and pypy_macros.h are also put there, instead of the
now-gone pypy/_interfaces.
The goal is to have the svn checkout as similar as possible as release
tarballs and virtualenvs, which have an include/ dir at the top
|
|
|
|
|
|
|
|
|
| |
is too hard to keep in sync all the various tool that needs to know where the
directories are, such as distutils, virtualenv, etc.
Now the only supported hierarchy is the one adopted in the svn checkout (i.e.,
lib-python, lib_pypy and site-packages needs to be on some parent directory
that the one where pypy-c resides)
|
|
|
|
| |
again now
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The most important changes are:
1) trunk/pypy/lib has been moved to trunk/lib_pypy: the code there is really
pure python and should not depend on pypy. However, some of the tests existed
as AppLevel tests: since they need to be run inside the pypy/ package, they
have been moved to pypy/module/test_lib_pypy. Also, a custom conftest makes
py.test module/test_lib_pypy also run the tests in ../lib_pypy
2) sys.pypy_prefix has been removed. Instead, sys.prefix and sys.exec_prefix
(which are always the same) are added. The prefix is searched at runtime,
starting from the directory where the pypy executable resides and walking up
until it finds the expected directoris (lib-python and lib_pypy)
The layout of the svn checkout has been designed in a way that it will be
possibile to run virtualenv -p translator/goal/pypy-c, without needing to
install pypy system-wide
|
| |
|
| |
|
|
|
|
|
|
|
| |
the interpreter.
Don't rely on sys.version (pypy does not indicates the compiler version),
directly fetch the manifest embedded in the program.
|
|
|
|
| |
library
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Tests are now collected more "normally" in their respective
directories 2.5.2/test and modified-2.5.2/test instead of "faking"
an artifical tree at lib-python level.
Note that you will get a collection error if you run a specific test via
"py.ttest lib-python/2.5.2/test/test_XYZ.py" if there is a test_XYZ in
the modified-2.5.2 directory.
Running "py.test lib-python" or "py.test --pypy=pypy-c lib-python" will
run both modified and unmodified tests as is to be expected and filenames
during testing progress clearly indicate where a test comes from.
|
|
|
|
|
|
| |
I think (by running pypy-c py.test --collect) that it is because in this
case it causes py.test to try to collect far more stuff, including both
the 2.5.2/test and the modified-2.5.2/test version of the same test.
|
| |
|
| |
|
|
|
|
|
| |
globally importable 'test' can get in the way in
some environments.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* fixes keyboardinterrupt issues - hopefully strange internal TBs
when doing ctrl-c are a thing of the past now
* be more robust against bogus source code and fix/improve
reporting of failing applevel tests
* fix some xfail/skipping issues and introduce a new
imperative way to xfail a test by
py.test.xfail() # or py.test.xfail("reasonstring")
I strongly recommend to use this instead of py.test.skip("xxx") because
it more clearly communicates that not just some platform/dependency
problem but an implementation issue is present. Note that it is also
possible to do:
@py.test.mark.xfail
def test_function()
and with Python2.6 it can also be applied to a test class like this:
@py.test.mark.xfail(reason="xyz needs improvement/refactoring")
class TestClass:
def test_...
And btw, you can issue "py.test --runxfail" to still run those tests and see
tracebacks. And finally,
@py.test.mark.xfail(run=False)
def test_function()
will not try to run the function at all.
|
|
|
|
|
|
|
|
|
|
|
|
| |
* introduces spaceconfig attribute to specify which
space to use for a certain test class.
* removes deprecated usage of py.test features
* removes internal imports of py lib
* various little test functionality related cleanups
note: the merging was going odd with svn throwing
assertion errors and other issues - so i merged
in two steps - hope this worked fine.
|
|
|
|
|
|
| |
installed there.
Obviously this function should return the version used to compile the current
executable, but pypy does not gives this information yet (in sys.version).
|
| |
|
| |
|
| |
|
|
|
|
| |
in various versions of CPython.
|
| |
|
| |
|
| |
|
|
|
|
| |
easy_install instead of exploding.
|
|
|
|
|
| |
remove all exec there and simply write down methods (also docstrings don't
have to be in _socket)
|
| |
|
| |
|