[Swig-user] SWIG -python -modern, 0-argument function

[Arnaud D. <ar...@he...>]

Hello SWIG users,

We just switched to using SWIG 2.0.12 (we were using 2.0.0 before), and 
noticed that the "check" code SWIG generates for python, in -modern 
mode, is different than it used to be.

For example, if I have the following in :test.hpp":
---
> int foo(void);
> int bar(int b);
---

the following wrappers will be generated (snipped for brevity):
---
> SWIGINTERN PyObject *_wrap_foo(PyObject *SWIGUNUSEDPARM(self), PyObject *args) {
>   PyObject *resultobj = 0;
>   int result;
>   {
>     SWIG_PYTHON_THREAD_BEGIN_ALLOW;
> [...]
>
> SWIGINTERN PyObject *_wrap_bar(PyObject *SWIGUNUSEDPARM(self), PyObject *args) {
>[...]
>   if(!PyArg_UnpackTuple(args,(char *)"bar",1,1,&obj0)) SWIG_fail;
>   ecode1 = SWIG_AsVal_int(obj0, &val1);
>   if (!SWIG_IsOK(ecode1)) {
>     SWIG_exception_fail(SWIG_ArgError(ecode1), "in method '" "bar" "', argument " "1"" of type '" "int""'");
>   }
>   arg1 = static_cast< int >(val1);
>   {
>     SWIG_PYTHON_THREAD_BEGIN_ALLOW;
>   [...]
> }
---
(the command line was: /opt/swig-2.0.12-install/bin/swig -modern -python 
-threads -c++ -shadow -D__GNUC__ -I/tmp -o test.cpp test.i)

Notice how, '_wrap_foo' doesn't validate that the number of arguments is 
0 (i.e., no call to 'PyArg_UnpackTuple'), as it used to.

I was wondering why, and tracked the change to this: 
http://sourceforge.net/p/swig/code/12332/
More specifically, this piece of the diff:
---
> -      } else {
> -	Printf(parse_args, "if(!PyArg_UnpackTuple(args,(char *)\"%s\",%d,%d", iname, num_required, num_arguments);
> +      } else if (tuple_arguments > 0) {
> +	Printf(parse_args, "if(!PyArg_UnpackTuple(args,(char *)\"%s\",%d,%d", iname, tuple_required, tuple_arguments);
---

This is in the "szager-python-builtin" branch, and unfortunately the 
commit message doesn't quite explain the decision of adding the 
"tuple_arguments > 0" condition.

Then, that branch was brought into the trunk, and the rest is history.

To my SWIG-novice eyes, it looks like a bug, but ... would anyone happen 
to know why that condition was added?

Thanks!

Best regards,
      Arnaud


-- 
Arnaud Diederen
Hex-Rays SA: state of the art binary software analysis tools