summaryrefslogtreecommitdiff
path: root/c
Commit message (Collapse)AuthorAgeFilesLines
* Allow varsized arrays to appear at a non-last position in case of aArmin Rigo2013-11-082-1/+40
| | | | verify()ed structure
* Implementation: support a few extra ways to give initialization arguments to ↵Armin Rigo2013-11-082-79/+240
| | | | | | ffi.new() for var-sized structs. See the test.
* Add PyPy support for big-endian bitfieldsArmin Rigo2013-10-231-2/+0
|
* A redo of pull request 19: dynamically determine if the C compilerArmin Rigo2013-09-252-6/+3
| | | | | supports "__thread" or not on this platform. Done slightly more simply in setup.py.
* Fix bitfields on big-endian machinesArmin Rigo2013-09-012-8/+61
|
* Fix for win64Armin Rigo2013-08-011-1/+6
|
* Py3k fixArmin Rigo2013-08-011-2/+2
|
* I have seriously no clue how Win64 ctypes workArmin Rigo2013-08-011-2/+2
|
* Backed out changeset a0bbd364d7a2Armin Rigo2013-08-011-3/+0
|
* FixArmin Rigo2013-08-011-0/+3
|
* Attempted fixArmin Rigo2013-08-011-0/+5
|
* Bah?Armin Rigo2013-08-011-1/+1
|
* I fail to see how "cif->bytes ? cif->bytes : 40" means "at least 40".Armin Rigo2013-08-012-2/+7
| | | | It seems to mean only "not 0" to me.
* Add three tests about callbacks receiving inlined structure argumentsArmin Rigo2013-07-181-0/+55
|
* Add a testArmin Rigo2013-07-181-0/+17
|
* Attempt a fix for Win64Armin Rigo2013-07-181-0/+9
|
* Attempt to get rid of USE_C_LIBFFI_MSVC by really fixing libffi_msvc.Armin Rigo2013-07-183-20/+18
|
* Check in win64.obj, thanks Matti :-)Armin Rigo2013-07-181-0/+0
|
* Found and fixed the bug.Armin Rigo2013-07-181-0/+2
|
* I think that this should clean up the (relative) mess of the GC onArmin Rigo2013-07-181-38/+91
| | | | some-but-not-all cdata objects. The test still doesn't pass though.
* A test, and start to work on a fix, which doesn't work so farArmin Rigo2013-07-162-1/+33
|
* Finish the fix for issue #91.Armin Rigo2013-06-192-12/+12
|
* merge headsArmin Rigo2013-06-192-69/+67
|\
| * Attack issue 91; first step.Armin Rigo2013-06-192-69/+67
| |
* | extend and fix test_c.py for ARMDavid Schneider2013-06-161-5/+16
| |
* | check for ARM using a flag like it is done for MSVCDavid Schneider2013-06-161-3/+6
|/
* Fix for the alignment calculation of anonymous and zero sized bitfields on ARM.David Schneider2013-06-161-0/+2
|
* Fix MSVC bitfields in all tested cases.Armin Rigo2013-06-012-10/+42
|
* MSVC-style bitfields, first attemptArmin Rigo2013-06-012-23/+77
|
* issue #82: implement bitfields in a way that is, as far as I can tell,Armin Rigo2013-05-302-88/+153
| | | | fully compatible with gcc.
* issue 87: first stabArmin Rigo2013-05-222-9/+106
|
* Bump the version number of 0.7.Armin Rigo2013-05-222-2/+2
|
* Test and fix: two pointers of different types can compare equal, butArmin Rigo2013-05-162-6/+12
| | | | they still had different hashes
* Add a few lines in a testArmin Rigo2013-04-291-0/+3
|
* don't assume errno carried over from the previous testPhilip Jenvey2013-04-281-0/+1
|
* Fix for issue 77: give cdatas __module__, __name__ and __doc__ attributes,Armin Rigo2013-04-152-0/+16
| | | | for convenience with functools.wraps().
* moreArmin Rigo2013-04-041-0/+4
|
* Add CFFI_MEM_DEBUG and CFFI_MEM_LEAK defines, to build versions ofArmin Rigo2013-04-041-0/+15
| | | | _cffi_backend that help in debugging early frees.
* Merge d0f6755466f7 into default:Armin Rigo2013-03-312-22/+21
|\ | | | | | | | | | | | | | | | | Cancel again these changes: we can't pass 0 or None to mean NULL. Just use NULL explicitly. (After discussion on IRC, notably with Amaury -- thanks) Removed some unrelated doc changes that describe something that was planned but never implemented.
| * PyPy compatibilityArmin Rigo2013-03-301-6/+7
| |
| * Fix(?) an issue that only showed up on PPC64 so far, aboutArmin Rigo2013-03-301-6/+1
| | | | | | | | char-returning C functions.
| * Fix for PPC: check that "long double" has more precision than "double"Armin Rigo2013-03-302-15/+18
| | | | | | | | if the type is larger, without assuming that it can hold bigger values.
* | Cancel again these changes: we can't pass 0 or None to mean NULL. JustArmin Rigo2013-03-072-46/+10
|/ | | | | use NULL explicitly. (After discussion on IRC, notably with Amaury -- thanks)
* Test for the 0-to-ptr convertion.Armin Rigo2013-03-021-0/+2
|
* Change the hack: instead of passing None for NULL pointers, we pass 0.Armin Rigo2013-02-282-10/+47
| | | | | This is more compatible with C. Also accept 0 at other places that expect a pointer.
* Python 3 fixes.Armin Rigo2013-02-281-0/+1
|
* Skip running this file on top of PyPy.Armin Rigo2013-02-221-0/+3
|
* Moving the determination of the base integer type of an enumArmin Rigo2013-02-162-139/+56
| | | | | | out of C code, into the 'cffi' package. This is work in progress; it should eventually permit the new test to fully pass. For now wrote a warning in the doc.
* Kill a debugging print (and make it py3k compatible again)Armin Rigo2013-02-151-1/+0
|
* Another test that version number matches.Armin Rigo2013-02-131-1/+5
|
View Lorry Controller Status