On Fri, Jul 9, 2010 at 6:42 AM, toki doki <tokido...@gmail.com> wrote: > I have been bitten by this twice this week, so I thought I would tell about > it: > > Contrary to C, C++ has a boolean type, using the same keyword as > Python: "bool". When declaring a C++ function with a (C++) bool > argument, like for example " int myFunc(int val, bool exact)", it > seems logical (although it is not, in insight) to declare it that way: > > cdef extern from "xxx.hpp": > int myFunc(int,bool) > > However, cython will understand that "bool" declare a python object of > (python) type bool. Thus, when doing "myFunc(5,False)", myFunc receive > as argument a pointer to the python object "False". And here is the > catch: gcc will implicitly convert this pointer to a C++ "True" > value. After checking, this implicit conversion seems to be part of > the C++ standard. > > Therefore, the problem is as follow: > Cython and gcc will emit no warning when a C++ bool is declared by > "bool" in cython. But when a python "False" is given as an argument > of a C++ function typed by "bool", the C++ function will receive a C++ > "true". > > Of course, the right way to do it is to declare the C++ bool as a > "bint". But since the other C/C++ types can be declared by their > normal C names, I think many people will assume that "bool" is the > right keyword here. And since Cython and g++ do not give any warning > about it, this might lead to some slightly puzzling bugs in users > code.
Hmm... yes, this is particularly bad, as it takes the pointer which should never be NULL (= False) > I am not sure you can do something about it (maybe handle it the same > way that you handle C int and Python int ?). But if you cannot, it > deserves at least a warning in the docs. bint and bool are semantically different types, and to introduce a bool type we would have to either disallow or emulate it in non-C++ code. (This is why I called it bint rather than bool in the first case.) For the moment, though, I'm wondering why we even bother letting people declare the type as bool--it's not a very useful (Python) type to statically declare, but can cause a lot of confusion. > (and for the record, although I have been posting quite a lot this > week about bugs and limitations of the c++ mode of cython, I am > overall delighted with it; it works much better and has much more > features than I was expecting when I started using it last week :-) Glad to hear. And we love bug reports, it helps us make the product better. - Robert _______________________________________________ Cython-dev mailing list Cython-dev@codespeak.net http://codespeak.net/mailman/listinfo/cython-dev
