Support for "volatile" keyword

[jdemeyer]

No description provided.

[scoder]

From a code design perspective, it feels like there should be two separate subclasses for the CConstOrVolatileType, even if most of the implementation is in a shared base class. Also, I'm sure I'll fail to remember that "cv" stands for "const or volatile" when I look at the code later. And I don't see a real need for a flag that merges the two - why not just test both flags in conditions that apply to both?

[scoder]

Thinking about this some more, I'm unsure if const and volatile are really the same kind of thing from Cython's perspective. They are both C modifiers, ok, and const is certainly part of the type, but is volatile also part of the type? It feels more like a part of a variable (declaration) than a part of a data type.

[jdemeyer]

From a C/C++ point of view, "const" and "volatile" are often considered together. The name "cv" isn't my invention, it's regularly used when talking about const/volatile modifiers (just one example: http://en.cppreference.com/w/cpp/language/cv). But of course I can change cv to something else for Cython.

[jdemeyer]

And I don't see a real need for a flag that merges the two - why not just test both flags in conditions that apply to both?

The problem is that in the current code, is_const is really used for two things: on the one hand, it is used for semantic const-ness (to prevent assignment for example). On the other hand, it is used as replacement for an isinstance(..., CConstType) check or a hasttr(... , "const_base_type") check. For the former, I am using is_const/is_volatile and for the latter I am using is_const_or_volatile.

[jdemeyer]

Is there any chance to revisit this?

[scoder]

Ok, I'm fine with this change. Only found a little improvement to simplify the parsing code.

[scoder]

Proposal:

is_const = is_volatile = False
while s.sy == 'IDENT' and s.systring in ('const', 'volatile'):
    ...
if s.sy != 'IDENT':
    error(...)
if is_const or is_volatile:
    ...

[jdemeyer]

Fun fact: cdef const const int a is currently not an error, but I guess it should be :-)

[jdemeyer]

Why is Travis CI no longer testing this PR?

[scoder]

It is, its last test run was half an hour ago.