Add support for templates about castxml

[bradking]

Strange, gccxml never supported uninstantiated templates at all. What does pygccxml expect that gccxml provides and CastXML doesn't?

[iMichka]

Ok, sorry I was confused and did not post everything.
In fact the real code which being tested by pygccxml is the following:

template <typename T> struct A {};
template <int N> struct A<const char[N]>
{ static int size(const char[N]) { return N - 1; } };

The resulting xml file with castxml:
https://gist.github.com/iMichka/e1faf2d67076260539c0

The resulting xml file with gccxml:
https://gist.github.com/iMichka/af55f8e8a2c56d8ed1fa

With gccxml I got:

<Struct id="_140" name="A&lt;const char [N]&gt;" context="_1" mangled="1AIAT__KcE" location="f1:4" file="f1" line="4" artificial="1" size="0" align="8" members="_160 " bases=""/>

Pyggcxml is expecting the to find the struct named in the global namespace: 'A<const char [N]>'

[bradking]

This was a case where gccxml would incorrectly output a partial template specialization as if it were a full specialization. This was a bug that we never fixed in gccxml. CastXML does not have the bug.

So, pygccxml should be taught to accept the lack of this output as correct or have the test case simply removed.

[iMichka]

Thanks. I found the discussion on the mailing list.
http://public.kitware.com/pipermail/gccxml/2010-January.txt

I guess the tests were just added to make sure if the bug was fixed one day in gccxml, it would detect it and trigger an error; telling the maintainer to look what has changed. It was confusing as there were no comments and I was not sure if this was the wanted output.

I kept the test for the moment and modified it for CastXML, so that it makes sures this output can not be found:
CastXML/pygccxml@a2b1230

Thanks for the help