Commits

Show all
Author Commit Message Labels Comments Date
Antonio Cuni
support string arguments and results. Makes test_stringresult passing
Branches
jitypes2
Antonio Cuni
actually pass the *value* of the pointer buffer (i.e., the address of the param which has been wrapped with byref). test_voidresult passes
Branches
jitypes2
Antonio Cuni
merge from trunk: svn merge svn+ssh://codespeak.net/svn/pypy/trunk -r79770:HEAD .
Branches
jitypes2
Antonio Cuni
cast pointers to unsigned long, not signed. This make it possible to use app-level longs to represent them in _ffi
Branches
jitypes2
Antonio Cuni
add support for pointers as arguments/return value. Since _ffi is only a thin layer, pointers are represented just as integers at app-level
Branches
jitypes2
Antonio Cuni
make test_wchar_result passing. Probably the details of type conversion will be refactored later
Branches
jitypes2
Antonio Cuni
do the char-->int conversion based on the declared type, not on the actual value of the argument
Branches
jitypes2
Antonio Cuni
merge from trunk: svn merge http://codespeak.net/svn/pypy/trunk -r79389:HEAD in particular, this merges the _ctypes changes that amaury did on the fast-forward branch and then merged on trunk
Branches
jitypes2
Antonio Cuni
kill this hack. The goal now is to have something that works, we will make it jit-friendly later
Branches
jitypes2
Antonio Cuni
add support for converting arguments. The first test in test_functions (test_wchar_parm) passes
Branches
jitypes2
Antonio Cuni
add support for float arguments/return value also at applevel
Branches
jitypes2
Antonio Cuni
add support for single float parameters and result to rlib.libffi. A bit hackish because we have to work around the limitation of the JIT which does not support single floats
Branches
jitypes2
Antonio Cuni
raise an applevel AttributeError when trying to get a pointer to a non existent function
Branches
jitypes2
Antonio Cuni
write a test to check that a call through ctypes is fast, and also generally improve a bit the machinery in test_pypy_c
Branches
jitypes2
Antonio Cuni
merge from trunk: svn merge svn+ssh://codespeak.net/svn/pypy/trunk -r78974:HEAD
Branches
jitypes2
Antonio Cuni
veeeery IN-PROGRESS: use _ffi instead of _rawffi to call C functions; it's already much faster, but there is still an issue with calling funcptr(*args), as it forces a frame
Branches
jitypes2
Antonio Cuni
enable _ffi by default on this branch
Branches
jitypes2
Antonio Cuni
a branch where to try to have fast ctypes based on the _ffi module
Branches
jitypes2
Antonio Cuni
fix the test
Antonio Cuni
don't use NotImplementError for flow control, it's not fully rpython
Antonio Cuni
make "except {NotImplementedError,AssertionError}" explicitly not rpython (but still allow it for geninterp)
Antonio Cuni
don't rely on catching NotImplementedError, it's not rpython
Antonio Cuni
fix title, fix english, add two highlights
Maciej Fijalkowski
start drafting a release announcement
Maciej Fijalkowski
A day of wonders - make this test test anything
Maciej Fijalkowski
Make the test actually test *something*
Maciej Fijalkowski
Revert 78939, it still doesn't work. Add correct casts
Maciej Fijalkowski
add() was adding to self.pending, which was already something else
Maciej Fijalkowski
Make setupstate a newstyle class and make Repr a more-obvious newstyle class
Antonio Cuni
don't crash if you try to compare an r_singlefloat with == or !=. Some places in the translation rely on being able to compare random stuff without crashing. Make sure we still get a crash if we use == or != in rpython
  1. Prev
  2. Next