Remove FunctionRecord objects in favor of existing PyMethodDef globals

[Reid Kleckner, 2009-12-16 23:11:05]

PTAL.  This assumes that all PyMethodDefs are global and live forever, which is
certainly the case in CPython, and should be for all extension modules.  I can
imagine someone allocating a PyMethodDef and taking a weakref to the PyCFunction
(does it even support that?) and then deallocating the methoddef when it's
freed, but that's insane.

[Collin Winter, 2009-12-17 00:23:43]

http://codereview.appspot.com/179086/diff/8/10
File Util/RuntimeFeedback.cc (left):

http://codereview.appspot.com/179086/diff/8/10#oldcode59
Util/RuntimeFeedback.cc:59: else if (src.InFuncMode() &&
src.data_[i].getPointer() != NULL) {
You can't Py_INCREF a PyMethodDef. FuncMode needs to stick around.

[Reid Kleckner, 2009-12-17 01:20:48]

http://codereview.appspot.com/179086/diff/8/10
File Util/RuntimeFeedback.cc (left):

http://codereview.appspot.com/179086/diff/8/10#oldcode59
Util/RuntimeFeedback.cc:59: else if (src.InFuncMode() &&
src.data_[i].getPointer() != NULL) {
On 2009/12/17 00:23:43, Collin Winter wrote:
> You can't Py_INCREF a PyMethodDef. FuncMode needs to stick around.

I didn't remove func mode, so don't worry, it doesn't try to incref the
PyMethodDef.  It just copies the data as normal.

[collinwinter, 2009-12-17 20:38:38]

LGTM