Re: ctypes-users Digest, Vol 49, Issue 4

"Mark Tolonen" <[email protected]> wrote in message 
news:[email protected]...
>
> "Mads Kiilerich" <[email protected]> wrote in message
> news:[email protected]...
>> Adam Morris wrote, On 06/12/2010 06:44 AM:
>>>
>>>     AFAIK ctypes should handle GIL correctly automatically, also when
>>>     used from threads.
>>>
>>>
>>> Sure am glad to hear that!
>>>
>>>     To me it sounds more like a problem with bad alignment or bad
>>>     parameter
>>>
>>>     profile of the callback function.
>>>
>>>
>>> I don't know about bad alignment, but you may be right around the bad
>>> parameter profile. I think I've narrowed down the problem to the
>>> context argument that is being passed to the SetCallback function of
>>> the dll. If I pass it null or 0 or pointer() or even byref(), the
>>> result is a bus error of 0x00 when it actually calls the callback, but
>>> if I pass it 10 it gives a bus error with address of 0x0a. I'm
>>> confused because the example files of the SDK show the C code passing
>>> a "0" for the context argument, so then why would it spell trouble
>>> using that from Python? Even weirder, if I don't pass any context
>>> parameter at all to the SetCallback function, I get the exact same
>>> error (or is this because I don't understand how parameter lists work
>>> in C?)
>>>
>>
>> It seems like the lib uses and dereferences pContext. You have to give
>> it something it likes.
>>
>> You could try to make a little C application that tests this API and
>> calls the callback. Try to do it without using the SDK .h files - create
>> your own and simplify them as much as possible. That will help you get
>> familiar with the API and take ctypes out of the equation. If you can't
>> create such an application then you probably also can't do it with
>> ctypes - unless you are lucky.
>>
>>> Here's all the relevant typedefs as defined in the SDK:
>>>
>>>
>>> #Storage declarations:
>>>
>>> typedef signed int      ax_int32_t
>>>
>>> typedef unsigned short  ax_char16_t
>>>
>>> typedef unsigned int    ax_handle_t
>>>
>>> # can't find AX_CALLCONV nor AX_API in the SDK
>>>
>>>
>>> #Callback declaration:
>>>
>>> typedef void(AX_CALLCONV * ax_vote_callback_t)(ax_handle_t hubHandle,
>>> ax_int32_t deviceSerialNumber, ax_int32_t vote, void *pContext)
>>>
>>>
>>> Here's the SetCallback method I have to use, notice the context 
>>> argument:
>>>
>>>
>>> AX_API ax_result_t AX_CALLCONV  Ax_SetVoteCallback (ax_vote_callback_t
>>> pfnCallback, void *pContext
>>>
>>>
>>> Here's what I do in my own code, python 3.1 on Mac OS X, after "from
>>> cytpes import *":
>>>
>>>
>>>     def set_vote_callback(self, python_function):
>>>
>>>         self.keep_wrap_function_reference = CFUNCTYPE(c_void_p,
>>> c_uint, c_int, c_int, c_void_p)
>>>
>>
>> Shouldn't the first parameter (ax_handle_t) be an integer?
>>
>>>         context = c_int(0)  # no idea what to do here, honestly
>>>
>>
>> Try to use None instead.
>>
>>>         result =
>>> self.cdll.Ax_SetVoteCallback(self.keep_wrap_function_reference(python_function),
>>> context)
>
> The problem is in the line above.
> self.keep_wrap_function_reference(python_function) creates a CFUNCTYPE
> wrapper for the python function, but right after this call completes there
> are no references to this wrapper and the object is deleted.  Assign this
> wrapped function to a variable that is still in scope when the callback
> actually occurs and it will work.  Something like:
>
> self.function_instance = 
> self.keep_wrap_function_reference(python_function)
> result = self.cdll.Ax_SetVoteCallback(self.function_instance,context)
>
> HTH,
> Mark

Oh, and here's the relevant docs (at the very end of the section):

http://docs.python.org/library/ctypes.html?highlight=ctypes%20tutorial#callback-functions

"Important note for callback functions:

Make sure you keep references to CFUNCTYPE objects as long as they are used 
from C code. ctypes doesn't, and if you don't, they may be garbage 
collected, crashing your program when a callback is made."

-Mark 



------------------------------------------------------------------------------
ThinkGeek and WIRED's GeekDad team up for the Ultimate 
GeekDad Father's Day Giveaway. ONE MASSIVE PRIZE to the 
lucky parental unit.  See the prize list and enter to win: 
http://p.sf.net/sfu/thinkgeek-promo