Commit message (Collapse) | Author | Age | Files | Lines | |
---|---|---|---|---|---|
* | Tentative fix for issue #542 | Armin Rigo | 2022-07-18 | 1 | -1/+2 |
| | |||||
* | #453 | Armin Rigo | 2020-05-20 | 1 | -2/+5 |
| | | | | | | Special-case typedefs of arrays with '...' length, so that they can be used through recompiler.py as if they had a specified length in most cases. | ||||
* | Issue 364 | Armin Rigo | 2018-06-16 | 1 | -4/+7 |
| | | | | Add support for MSVC's "#pragma pack(n)" in ABI mode | ||||
* | Issue #357: fix the out-of-line ABI mode when we see structs | Armin Rigo | 2018-01-31 | 1 | -9/+8 |
| | | | | containing anonymous unions or vice-versa | ||||
* | Issue #319 | Armin Rigo | 2017-06-15 | 1 | -10/+14 |
| | | | | | | | | | Fix a race condition: we can't stick an attribute to ffi._backend, because another thread might be in the middle of ``for name in ffi._backend.__dict__``. To minimize damage if I reintroduce an attribute-adding function in the future, also tweak the ``for``. | ||||
* | Tests and fixes | Armin Rigo | 2017-05-31 | 1 | -0/+2 |
| | |||||
* | fix | Armin Rigo | 2017-05-29 | 1 | -1/+2 |
| | |||||
* | More tweaks, more tests, try harder to avoid including <complex.h> | Armin Rigo | 2017-05-29 | 1 | -2/+4 |
| | | | | because it is not necessarily there | ||||
* | Merge default in. | Tom Krauss | 2017-03-12 | 1 | -0/+2 |
|\ | |||||
| * | initial support for 'float _Complex' and 'double _Complex' | Tom Krauss | 2017-01-15 | 1 | -0/+2 |
| | | | | | | | | Work In Progress - sources have printfs that will be removed. | ||||
* | | Avoid cyclic imports by moving exceptions to a separate module | Ronan Lamy | 2017-01-20 | 1 | -17/+10 |
|/ | |||||
* | Revert b84710ae130a again. Ensure we get a warning for every opaque | Armin Rigo | 2016-10-20 | 1 | -4/+12 |
| | | | | | enum, but then fall back to 'unsigned int'. See documentation for motivation. | ||||
* | Eventually, replace this UserWarning with an error | Armin Rigo | 2016-08-07 | 1 | -6/+4 |
| | |||||
* | Give a warning instead of crashing; in this way we can call the next | Armin Rigo | 2015-11-06 | 1 | -4/+6 |
| | | | | | release 1.3.1 as planned and give some advance warning to users relying on this. | ||||
* | Complain more readily about opaque enums, by refusing to guess which | Armin Rigo | 2015-11-05 | 1 | -2/+5 |
| | | | | integer type (unsigned/signed, int/long) it is meant to be. | ||||
* | Propagate "volatile" in addition to "const" and "restrict" | Armin Rigo | 2015-11-02 | 1 | -0/+3 |
| | |||||
* | Support directly __stdcall or WINAPI (or __cdecl, ignored) inside | Armin Rigo | 2015-10-06 | 1 | -16/+7 |
| | | | | cparser. | ||||
* | in-progress, but found a problem | Armin Rigo | 2015-10-06 | 1 | -3/+4 |
| | |||||
* | forgot pdb | Armin Rigo | 2015-10-05 | 1 | -1/+0 |
| | |||||
* | non-windows fixes | Armin Rigo | 2015-10-05 | 1 | -4/+6 |
| | |||||
* | Starting, with exactly two function types: no-abi (i.e. cdecl on windows), ↵ | Armin Rigo | 2015-10-05 | 1 | -5/+22 |
| | | | | or stdcall. | ||||
* | A note | Armin Rigo | 2015-09-30 | 1 | -0/+3 |
| | |||||
* | Seems that "__restrict" is recognized in all gcc modes I could try, | Armin Rigo | 2015-09-30 | 1 | -1/+1 |
| | | | | and it should also be recognized on MSVC | ||||
* | Tests and fixes for "const" fields | Armin Rigo | 2015-09-30 | 1 | -1/+2 |
| | |||||
* | in-progress: add qualifiers through model.py, cparser.py, and recompiler.py | Armin Rigo | 2015-09-30 | 1 | -18/+33 |
| | |||||
* | Test the various cases, fix issues with rounding, etc. | Armin Rigo | 2015-08-28 | 1 | -7/+1 |
| | | | | | | A subtle point is that your version uses is_float_type(), which is true also for "long double", and so it will loose precision on platforms like x86 Linux where "long double" is larger than "double". | ||||
* | Add support for `typedef float... foo_t`. | Max Belanger | 2015-08-27 | 1 | -0/+17 |
| | |||||
* | Refactor the way global variables are accessed. Now, every access to | Armin Rigo | 2015-07-03 | 1 | -20/+1 |
| | | | | | "lib.var" calls a generated C function to get "&var". This fixes issue #212 and also make it work in more cases, like __thread variables. | ||||
* | Issues 69, 73: add the syntax "typedef int... foo_t;". | Armin Rigo | 2015-05-30 | 1 | -5/+25 |
| | |||||
* | Issue #193: if we use a struct between the first cdef() where it is | Armin Rigo | 2015-05-20 | 1 | -4/+5 |
| | | | | | declared and another cdef() where its fields are defined, then the definition was ignored. | ||||
* | hg merge default | Armin Rigo | 2015-04-27 | 1 | -2/+35 |
|\ | |||||
| * | Check sizeof(global variables) when possible | Armin Rigo | 2015-04-27 | 1 | -0/+21 |
| | | |||||
| * | oups | Armin Rigo | 2015-04-25 | 1 | -0/+2 |
| | | |||||
| * | fixes for anonymous nested structs | Armin Rigo | 2015-04-25 | 1 | -0/+6 |
| | | |||||
| * | hg merge default | Armin Rigo | 2015-04-24 | 1 | -10/+0 |
| |\ | |||||
| * \ | hg merge default | Armin Rigo | 2015-04-24 | 1 | -2/+6 |
| |\ \ | |||||
| | * | | in-progress | Armin Rigo | 2015-04-24 | 1 | -1/+1 |
| | | | | |||||
| | * | | fixes | Armin Rigo | 2015-04-18 | 1 | -0/+1 |
| | | | | |||||
| | * | | the '[...]' syntax | Armin Rigo | 2015-04-18 | 1 | -1/+1 |
| | | | | |||||
| | * | | Starting working on the "recompiler" | Armin Rigo | 2015-04-15 | 1 | -0/+3 |
| | | | | |||||
* | | | | Test and fix: the new types like "int_least8_t" were actually not | Armin Rigo | 2015-04-27 | 1 | -0/+18 |
| |_|/ |/| | | | | | | | | available because model.py didn't list them | ||||
* | | | Reimplement vengine_gen returning incomplete structs. | Armin Rigo | 2015-04-24 | 1 | -10/+0 |
|/ / | |||||
* | | Skip an existing test after figuring out that it is indeed likely to | Armin Rigo | 2015-04-24 | 1 | -0/+10 |
|/ | | | | | crash in test_vgen, depending on details of the platform. Changed it to clearly flag this as invalid already in the cdef, and why. | ||||
* | improve error message a bit | Armin Rigo | 2015-01-02 | 1 | -1/+1 |
| | |||||
* | Fight a lot the CPython buffer/memoryview interface until we get a | Armin Rigo | 2014-12-28 | 1 | -0/+2 |
| | | | | | hopefully not-too-buggy ffi.from_buffer(). Obviously it's only half-tested because 2.7 doesn't really have built-in memoryview objects. | ||||
* | Handle the case where someone has made `import weakref` return a weird proxy.alex_gaynor/handle-the-case-where-someone-has-made-i-1391634819444 | Alex Gaynor | 2014-02-05 | 1 | -2/+3 |
| | | | | Fixes: https://bugs.pypy.org/issue1688 | ||||
* | Issue 131: support ffi.cdef("...", packed=True) | Armin Rigo | 2014-01-14 | 1 | -1/+6 |
| | |||||
* | Kill is_{signed,unsigned}_type() | Armin Rigo | 2013-12-15 | 1 | -17/+13 |
| | |||||
* | Due to a CPython bug, we cannot use setdefault() here | Armin Rigo | 2013-11-10 | 1 | -3/+10 |
| | |||||
* | Bah, setdefault() is not atomic on WeakValueDictionary. | Armin Rigo | 2013-11-10 | 1 | -2/+9 |
| |