summaryrefslogtreecommitdiff
path: root/testing/test_function.py
Commit message (Collapse)AuthorAgeFilesLines
* Skip the sinf test on windowsArmin Rigo2014-03-071-0/+2
|
* fix for win32Matti Picus2014-03-021-11/+15
|
* Fix ffi.dlopen(None): we can't replace it with ffi.dlopen("c") on POSIXArmin Rigo2014-02-281-15/+7
| | | | | | | systems. It has a different meaning: getting a handle that works for *any* already-loaded library, not just "libc.so". This breaks obscurely a test that is already obscure, so ignoring that (I checked that in C it would break the same way anyway).
* Don't care too much about the ctypes backendArmin Rigo2013-08-011-0/+2
|
* Add the (now-passing) test from issue #92Armin Rigo2013-07-181-0/+21
|
* Fix for issue 77: give cdatas __module__, __name__ and __doc__ attributes,Armin Rigo2013-04-151-0/+16
| | | | for convenience with functools.wraps().
* Try to work around issues when running in a buildbot: only writes to stderr.Armin Rigo2013-04-091-58/+45
|
* Merged in pjenvey/cffi (pull request #13)Armin Rigo2013-04-061-0/+25
|\
| * Test and fix in the ctypes backend for calling "void *" functions with aArmin Rigo2013-04-061-0/+25
| | | | | | | | Python string argument.
* | fix handling w/ py3 hasattr, which only swallows AttributeErrorsPhilip Jenvey2013-04-051-0/+8
|/
* Linux isn't the only platform using glibc. Debian gnukfreebsd and hurd ports ↵Stefano Rivera2013-03-301-1/+1
| | | | do too
* Test and fix for pull request #11: anonymous enums.Armin Rigo2013-03-291-1/+4
|
* Issue #67Armin Rigo2013-03-271-0/+9
| | | | Copy the enum's values on the library object returned by dlopen().
* (lazka) improve the test, and fix a typo (bad arigo)Armin Rigo2013-02-221-3/+9
|
* (lazka, arigo) Change dlopen() to accept either a full path or a libraryArmin Rigo2013-02-221-0/+11
| | | | name.
* Bah, seems Yet Another ctypes limitation of PyPy. Don't fully care :-/Armin Rigo2013-02-131-0/+3
|
* Add the .DLL after KERNEL32.Armin Rigo2012-11-251-1/+1
|
* FILE limited supportArmin Rigo2012-10-191-0/+18
|
* Change the default dlopen() flags from RTLD_LAZY to RTLD_NOW.Armin Rigo2012-09-241-0/+9
| | | | Give access to all flags, for more precise control.
* hack hack hackArmin Rigo2012-08-121-8/+8
|
* hg merge default, and try to resolve all conflicts.Armin Rigo2012-08-121-13/+26
|\ | | | | | | Not tested so far.
| * Try to fix the tests on OS/X, mainly by skipping the ABI tests thatArmin Rigo2012-08-121-5/+7
| | | | | | | | | | depend on the 'stdout' and 'stderr' symbols being present (under this name) in the libc.
| * Add two tests, one passing and one (probably definitely) skipped.Armin Rigo2012-08-121-0/+11
| |
| * As discussed on the mailing list: str() -> ffi.string()Armin Rigo2012-08-031-2/+2
| |
* | An attempt to port cffi to python3.Amaury Forgeot d'Arc2012-07-291-33/+38
|/ | | | Most tests are passing, yeah!
* This test works (tested elsewhere).Armin Rigo2012-07-181-3/+0
|
* Change ffi.new() to take a pointer-to-X instead of directly X,Armin Rigo2012-07-161-1/+1
| | | | in order to match directly the returned type.
* Test and fix: from callbacks with 'void' as the result type, you shouldArmin Rigo2012-07-051-1/+20
| | | | really return None and not anything else.
* Fix(?) the ctypes backend, as far as the tests are concerned.Armin Rigo2012-06-261-1/+1
|
* Convert the front-end. Missing backend_ctypes.Armin Rigo2012-06-251-10/+10
|
* Test and fix.Armin Rigo2012-06-161-0/+10
|
* A passing test.Armin Rigo2012-06-161-0/+13
|
* Fix tests for windows.Armin Rigo2012-06-161-2/+14
|
* Rename ffi.rawload() to ffi.dlopen().Armin Rigo2012-06-141-13/+13
|
* Fix tests.Armin Rigo2012-06-121-2/+13
|
* Rename the package from "ffi" to "cffi".Armin Rigo2012-06-051-2/+2
| | | | | Use local imports inside the package; anyway it has no chance to work on Python < 2.6 because pycparser doesn't.
* Complete the tests.Armin Rigo2012-06-031-2/+9
| | | | Works on _ffi_backend, but not on backend_ctypes...
* Shuffle tests.Armin Rigo2012-06-011-7/+7
|
* Variadic calls.Armin Rigo2012-06-011-1/+2
|
* Extent the API to allow None to variadic funcs.Armin Rigo2012-06-011-1/+3
| | | | | | It is somehow a simplification because some operations on pointers will return None to mean "the C NULL". Passing them to variadic functions as None makes sense.
* Not implementing this feature.Armin Rigo2012-06-011-2/+4
|
* Fix test.Armin Rigo2012-06-011-1/+1
|
* Import other tests here too.Armin Rigo2012-06-011-164/+167
|
* Finish to fix the existing tests.Armin Rigo2012-05-271-11/+11
|
* In-progressArmin Rigo2012-05-271-1/+1
|
* Rename malloc() to new().Armin Rigo2012-05-271-7/+16
| | | | | Keep naming the type we want a new() copy of, rather than the resulting pointer type.
* Small edits.Armin Rigo2012-05-261-0/+6
|
* More or less adapt the tests.Armin Rigo2012-05-261-12/+12
|
* Merge branch 'master' of github.com:arigo/ffifijal2012-05-231-0/+17
|\
| * Test and fix.Armin Rigo2012-05-231-0/+3
| |