summaryrefslogtreecommitdiff
path: root/testing/cffi1
Commit message (Collapse)AuthorAgeFilesLines
* skip on windowsArmin Rigo2017-03-131-0/+2
|
* silence warningsArmin Rigo2017-03-131-5/+5
|
* issue #255: comparing primitive cdatasArmin Rigo2017-02-191-10/+10
|
* Make 'ffi.buffer' be the type of minibuffers returned by cffi, andArmin Rigo2017-02-041-0/+1
| | | | 'ffi.buffer()' now calls the constructor.
* complain clearly if set_source() is given a /-separated nameArmin Rigo2017-01-261-0/+5
|
* Add '#pragma GCC visibility push(default)' in case the userArmin Rigo2017-01-171-0/+13
| | | | specifies -fvisibility=hidden
* py3 test fixArmin Rigo2017-01-021-1/+1
|
* Python 2.6 compatArmin Rigo2016-12-301-1/+5
|
* Explicitly flush the import machinery cache, in case the filesystem doesn't ↵Stefano Rivera2016-12-301-0/+3
| | | | have enough mtime resolution to notice change
* Detect packed structs. Improve error messages and test them.Armin Rigo2016-12-222-13/+84
|
* A failing testArmin Rigo2016-12-221-0/+22
|
* Fix/test/expand the error messages we get by using structs/unions asArmin Rigo2016-12-221-0/+76
| | | | arguments/return values of functions
* fix testArmin Rigo2016-12-071-1/+2
|
* Fix: the condition "offsetof == sizeof" for being a var-sized arrayArmin Rigo2016-10-292-3/+7
| | | | | is bogus. See test for a case where it is not the case because of alignment.
* One more case, this time in CompiledFFI.sizeof().Armin Rigo2016-10-281-4/+9
|
* Fix testArmin Rigo2016-10-251-5/+8
|
* Fix what is hopefully the last remaining issues with ↵Armin Rigo2016-10-251-1/+10
| | | | calculate_variable_array_length
* typedef int foo_t[...];Armin Rigo2016-09-202-0/+30
|
* Windows funArmin Rigo2016-09-031-3/+3
|
* Add two tests for 34b29a139894Armin Rigo2016-08-221-0/+18
|
* Avoid a combination where we end up with two incompatible "typedef xxxArmin Rigo2016-08-061-0/+12
| | | | _Bool;"
* Test and fix: refuse to 'recompile' a cdef that declares a structArmin Rigo2016-07-311-0/+32
| | | | with an opaque struct as a field
* Use Py_LIMITED_API in the generated C extension modules, because doingArmin Rigo2016-07-191-3/+6
| | | | | so seems easy. Still need to check if and how we have to produce the actual .so files.
* Test and fix: the "bool" type in C++Armin Rigo2016-06-051-0/+7
|
* Haaaack! Have 'lib.__class__' return &PyModule_Type. It makesArmin Rigo2016-05-251-2/+2
| | | | | | | | | help(lib) behave and display a nice module-like view of your compiled lib. Note that this is seriously unexpected, but I hope nobody should get hit by it by looking at 'lib.__class__' directly... Might be reverted if I hear about someone :-)
* Fix tests in 'testing' to account for e7ca388b0197 (I only fixed theArmin Rigo2016-05-081-3/+9
| | | | tests in 'c/test_c.py')
* Re-add this hack, removed in 51f1337c9b4c; but this time only in testsArmin Rigo2016-04-211-0/+8
|
* Windows fix: 'struct a' is a great name for being predefined and not reusableArmin Rigo2016-04-211-9/+9
|
* Support help(lib.foo)Armin Rigo2016-04-191-3/+15
|
* Issue #254: extern "Python+C"Armin Rigo2016-04-181-1/+32
|
* Remove again ffi.rawstring(), and implement instead ffi.unpack().Armin Rigo2016-04-161-9/+4
| | | | | | | | | | | | | Pre-documentation notes: (hi Amaury :-) * ffi.unpack(<cdata 'char'>, n) == ffi.buffer(<cdata 'char'>, n)[:] but I hope it is a little bit more natural * ffi.unpack(<cdata 'wchar_t'>, n): this is the original motivation, because it has no previous equivalent * ffi.unpack(<cdata 'int'>, n) == list(<cdata 'int'>[0:n]) but should be much faster on CPython
* Python3 compatArmin Rigo2016-04-151-3/+3
|
* ffi.rawstring(), with a minimal interfaceArmin Rigo2016-04-151-0/+12
|
* Change the API of ffi.list_types()Armin Rigo2016-03-301-10/+10
|
* ffi.list_types()Armin Rigo2016-03-272-13/+124
|
* Issue #245: __stdcall not generated correctly on extern "Python" functionsArmin Rigo2016-02-041-0/+30
|
* Allow testing against installed package (which has egg_info already)Stefano Rivera2016-01-171-1/+2
|
* fix when running this test in the same process as ../cffi0/test_verify.py ↵Armin Rigo2016-01-161-5/+5
| | | | (found on os/x, but I don't know why it doesn't also fail on linux)
* here tooArmin Rigo2016-01-161-2/+2
|
* win32Armin Rigo2016-01-161-1/+1
|
* Trying a different hack: stop patching SO and EXT_SUFFIX inArmin Rigo2016-01-161-22/+6
| | | | | | sysconfigdata, and instead patch directly a method in distutils.command.build_ext. Motivation: Windows, where the previous solution makes it add the wrong 'init...' in exports_symbols
* refactor details, start writing docsArmin Rigo2016-01-131-4/+61
|
* Add a passing test about "from mymodule import *"Armin Rigo2016-01-061-0/+7
|
* Add a __class__ attribute on lib objectsArmin Rigo2015-12-112-0/+3
|
* py3k compatArmin Rigo2015-12-091-1/+5
|
* hg merge static-callbackArmin Rigo2015-12-072-4/+224
|\
| * py3k compatArmin Rigo2015-12-072-9/+14
| |
| * hg merge defaultArmin Rigo2015-11-231-0/+39
| |\
| * | Change the @ffi.def_extern() decorator to not automatically replace theArmin Rigo2015-11-201-14/+16
| | | | | | | | | | | | | | | | | | function with the cdata. You need to get the cdata from the lib explicitly. This should make it clearer that there is only one cdata, even if you apply the decorator again.
| * | Write one error message directly to stderr instead of sys.stderr. ThisArmin Rigo2015-11-181-10/+2
| | | | | | | | | | | | | | | lets us avoid taking the GIL, which might crash in case the Python interpreter is not initialized at all.
View Lorry Controller Status