Fixed bug 14195: Added C++ mangling for passing function signatures

[Ingrater]

Added support for mangling the C++ equivalent type of:

template <typename T>
struct Delegate1 {};

template <typename R1>
struct Delegate1 < R1() > {};

template <typename T1, typename T2>
struct Delegate2 {};

template < typename R1, typename T1, typename T2, typename R2, typename T3, typename T4 >
struct Delegate2<R1(T1, T2), R2(T3, T4)> {};

void test14195a(Delegate1<void()> func){}

void test14195b(Delegate2<int(float, double), int(float, double)> func){}

The corresponding D code can be found in cppmangle.d. I hit this issue when trying to bind a C++ Delegate implementation to D.

[yebblies]

How hard is it to do linux mangling as well? I'd prefer the tests were in cppa/cppb so that the mangling is automatically verified.

[WalterBright]

using writestring("$$A6") would be more compact.

[WalterBright]

I'd like to see a test for this in cppa/cppb as well for the same reason @yebblies stated. And such would provide motivation, and a test case, for doing the Linux mangling as well.

[Ingrater]

Juding from the source code this is already implemented for linux. I will move the tests over to cppa/cppb as soon as I find time.

[Ingrater]

I tried enabeling the test on non-windows systems and got the following error message from the build machines: undefined reference to `DelegateTest1(Delegate1<void ()>)'
While this tells me that it is almost correctly implemented (because the linker can demangle it), the fact that it does not link tells me its only 'almost' correct. I don't have a linux development environment setup at the moment and don't have time nor interrest to fix this on non-windows systems.

[Ingrater]

Sorry, forget what I said. It is already correctly implemented on linux, I just made a mistake transitioning the test to cppa/cppb.

[yebblies]

In future pull requests please remember to add a bugzilla link to the description.

[yebblies]

Auto-merge toggled on