summaryrefslogtreecommitdiff
Commit message (Collapse)AuthorAgeFilesLines
* Detect and mark the final variable array in a varsized struct with ↵calculate_variable_array_lengthAndrew Leech2016-09-142-8/+11
| | | | | | BS_VARSIZESTRUCT_ARRAY Use this when returning this field/arrtibute of said struct
* Add some initial test assertions to check length details on variable length ↵Andrew Leech2016-09-131-3/+8
| | | | structs
* When using a variable length struct, calculate and enforce the length of the ↵Andrew Leech2016-09-131-1/+26
| | | | varsized_array when accessing this field.
* When getting repr on CDataObject_own_structptr, use length field for owned bytesnew_struct_allocated_sizeAndrew Leech2016-09-121-1/+3
|
* When getting ffi.buffer() on CDataObject_own_structptr, use length field for ↵Andrew Leech2016-09-121-1/+5
| | | | buffer size
* Add length field to CDataObject_own_structptr to hold the size of the memory ↵Andrew Leech2016-09-121-0/+2
| | | | | | block allocated for the struct. This is particularly useful for C99 variable-sized structures where the sizeof(<typedef>) does not directly inform the allocated size.
* hg merge release-1.8Armin Rigo2016-09-061-3/+3
|\
| * md5/shaArmin Rigo2016-09-061-3/+3
| |
| * Release branch, starting at 1.8.2 for CPythonArmin Rigo2016-09-060-0/+0
|/
* A direct testArmin Rigo2016-09-061-0/+19
|
* skip testArmin Rigo2016-09-061-0/+3
|
* Update and document v1.8.2Armin Rigo2016-09-068-8/+17
|
* Issue #283: initializer for nested anonymous structs inside unionsArmin Rigo2016-09-052-24/+51
|
* Skip test if setuptools cannot be importedArmin Rigo2016-09-041-1/+4
|
* Python 3 fixArmin Rigo2016-09-031-0/+6
|
* fixArmin Rigo2016-09-031-2/+4
|
* Issue #282: probable test and fixArmin Rigo2016-09-032-16/+23
|
* Un-document the removal of the ctypes backendArmin Rigo2016-09-031-4/+0
|
* Backed out changeset 0087e2aec9efArmin Rigo2016-09-039-198/+1431
| | | | Un-kill the ctypes backend. Issue #282 for a justification.
* Bump version number to 1.8.1Armin Rigo2016-09-037-8/+8
|
* - add ffi.compile(debug=flag)Armin Rigo2016-09-035-14/+33
| | | | | - this small addition, and a few others not in pypy 5.4, will make the next cffi release v1.8.1
* Document 0ff89f851d2cArmin Rigo2016-09-031-1/+3
|
* If we say Py_LIMITED_API and we're compiling with a debug version ofArmin Rigo2016-09-032-2/+18
| | | | CPython, "#include <Python.h>" crashes. Work around it the hard way.
* Windows funArmin Rigo2016-09-031-3/+3
|
* Kill the ctypes backend.Armin Rigo2016-09-0310-1431/+202
|
* Fix testArmin Rigo2016-09-031-1/+1
|
* update whatsnewArmin Rigo2016-08-242-3/+8
|
* merge headsArmin Rigo2016-08-227-40/+56
|\
| * Add two tests for 34b29a139894Armin Rigo2016-08-221-0/+18
| |
| * Update the docs according to issue #281Armin Rigo2016-08-223-35/+30
| |
| * Returning unions works fine in API mode, remove the checks.Armin Rigo2016-08-223-5/+6
| |
| * Avoid calling more CPython functions when there is an exception setArmin Rigo2016-08-221-0/+2
| |
* | If we don't know the version number of setuptools, we try to setArmin Rigo2016-08-221-1/+4
| | | | | | | | 'py_limited_api' anyway. At worst, we get a warning.
* | Remove the extra checks, which should be unnecessary and possibly causeArmin Rigo2016-08-222-4/+1
| | | | | | | | more confusion.
* | use py_limited_api flag when available.Daniel Holth2016-08-212-1/+43
|/ | | | | In setuptools>=26, Extension(..., py_limited_api=True) sets the .abi3.so filename, readable by Python 3.2+.
* Document some common mistakes of bad lifetime of ffi.new()Armin Rigo2016-08-171-0/+27
|
* Give an error when subtracting two pointers and the division's result isArmin Rigo2016-08-152-2/+23
| | | | | not exact (in gcc, we get nonsense, so it means it is undefined behavior for C, which is best handled by raising in cffi)
* Eventually, replace this UserWarning with an errorArmin Rigo2016-08-071-6/+4
|
* Avoid a combination where we end up with two incompatible "typedef xxxArmin Rigo2016-08-062-1/+15
| | | | _Bool;"
* Update docsArmin Rigo2016-08-063-6/+13
|
* Seems that as a #define, this hack breaks if a future header doesArmin Rigo2016-08-061-1/+1
| | | | | "typedef bool _Bool;". With a typedef, though, we get two identical typedefs, which at least g++ is perfectly happy with
* Update the version number to 1.8Armin Rigo2016-08-067-12/+12
|
* The null_byte_after_str branch of PyPy makes ffi.from_buffer(str)Armin Rigo2016-08-024-2/+22
| | | | possible.
* Issue #275: add Python 3.5 hereArmin Rigo2016-07-311-0/+1
|
* Document the fact that the C code contains '#define Py_LIMITED_API'.Armin Rigo2016-07-311-0/+11
|
* Test and fix: refuse to 'recompile' a cdef that declares a structArmin Rigo2016-07-312-0/+41
| | | | with an opaque struct as a field
* Use Py_LIMITED_API in the generated C extension modules, because doingArmin Rigo2016-07-193-22/+11
| | | | | so seems easy. Still need to check if and how we have to produce the actual .so files.
* Issue #271: in the test, call "cc" instead of "gcc", in case gcc isn'tArmin Rigo2016-07-181-1/+1
| | | | installed but another compiler is.
* Fix for a3708c91be7bArmin Rigo2016-07-161-5/+8
|
* Same hereArmin Rigo2016-07-111-1/+1
|
View Lorry Controller Status