• Walter Bright (6/6) Sep 10 2014 C++ name mangling distinguishes between a class and a struct. This disti...
• FreeSlave (11/18) Sep 10 2014 Just for clarification. I should notice it relates only to msvc
• Daniel Murphy (6/13) Sep 10 2014 pragma(cpp_mangle_as_struct)
• Lionello Lunesu (6/10) Sep 11 2014 and/or?
• Daniel Murphy (3/8) Sep 11 2014 Yeah. No need to add new syntax, and it means adding a way to make D li...
• Andrei Alexandrescu (2/17) Sep 11 2014 I think attributes would be the go-to approach. -- Andrei
• Daniel Murphy (3/4) Sep 11 2014 Why?
• Andrei Alexandrescu (2/6) Sep 11 2014 Because they postdate pragma :o). I'm fine either way. -- Andrei
• Dicebot (6/14) Sep 11 2014 Those are different tools. Attributes are part of language
• Idan Arye (6/23) Sep 11 2014 Also, there already is a pragma for controlling the mangling, so

Walter Bright <newshound2 digitalmars.com> writes:

C++ name mangling distinguishes between a class and a struct. This distinction 
has no semantic meaning, but there it is in the name mangling. In order to 
interop, we have to have a way to tell D to mangle a struct as a 'class' when 
interfacing with C++.

I have some ideas, but they're all kinda ugly.

Any ideas?

"FreeSlave" <freeslave93 gmail.com> writes:

On Wednesday, 10 September 2014 at 20:43:44 UTC, Walter Bright 
wrote:
 C++ name mangling distinguishes between a class and a struct. 
 This distinction has no semantic meaning, but there it is in 
 the name mangling. In order to interop, we have to have a way 
 to tell D to mangle a struct as a 'class' when interfacing with 
 C++.

 I have some ideas, but they're all kinda ugly.

 Any ideas?

Just for clarification. I should notice it relates only to msvc 
mangling model, which does not respect equivalence of struct and 
class and use different symbols to mangle them (U and V). There 
is also warning C4099 because of it.
g++ seems to use same mangling regardless of how you declare 
class (with class or struct keyword).

Even if you find good solution by extending language (for 
example, by adding extern(C++-struct)) it would make sense only 
for Windows platforms.

"Daniel Murphy" <yebbliesnospam gmail.com> writes:

"Walter Bright"  wrote in message news:luqd60$2u0r$1 digitalmars.com...

 C++ name mangling distinguishes between a class and a struct. This
 distinction has no semantic meaning, but there it is in the name mangling. 
 In order to
 interop, we have to have a way to tell D to mangle a struct as a 'class' 
 when interfacing with C++.

 I have some ideas, but they're all kinda ugly.

 Any ideas?

pragma(cpp_mangle_as_struct)
class Foo
{
}

This is a low-level feature, and should be ugly. 

Lionello Lunesu <lionello lunesu.remove.com> writes:

On 11/09/14 14:32, Daniel Murphy wrote:
 pragma(cpp_mangle_as_struct)
 class Foo
 {
 }

and/or?

pragma(cpp_mangle_as_class)
struct Bar
{
}

"Daniel Murphy" <yebbliesnospam gmail.com> writes:

"Lionello Lunesu"  wrote in message news:luriec$1081$1 digitalmars.com...

 and/or?

 pragma(cpp_mangle_as_class)
 struct Bar
 {
 }

Yeah.  No need to add new syntax, and it means adding a way to make D link 
with C++ rather than adding C++ features to D. 

Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:

On 9/10/14, 11:32 PM, Daniel Murphy wrote:
 "Walter Bright"  wrote in message news:luqd60$2u0r$1 digitalmars.com...

 C++ name mangling distinguishes between a class and a struct. This
 distinction has no semantic meaning, but there it is in the name
 mangling. In order to
 interop, we have to have a way to tell D to mangle a struct as a
 'class' when interfacing with C++.

 I have some ideas, but they're all kinda ugly.

 Any ideas?

 pragma(cpp_mangle_as_struct)
 class Foo
 {
 }

 This is a low-level feature, and should be ugly.

I think attributes would be the go-to approach. -- Andrei

"Daniel Murphy" <yebbliesnospam gmail.com> writes:

"Andrei Alexandrescu"  wrote in message 
news:lurk9p$123f$1 digitalmars.com...

 I think attributes would be the go-to approach. -- Andrei

Why? 

Andrei Alexandrescu <SeeWebsiteForEmail erdani.org> writes:

On 9/11/14, 12:57 AM, Daniel Murphy wrote:
 "Andrei Alexandrescu"  wrote in message
 news:lurk9p$123f$1 digitalmars.com...

 I think attributes would be the go-to approach. -- Andrei

 Why?

Because they postdate pragma :o). I'm fine either way. -- Andrei

"Dicebot" <public dicebot.lv> writes:

On Thursday, 11 September 2014 at 07:58:14 UTC, Andrei
Alexandrescu wrote:
 On 9/11/14, 12:57 AM, Daniel Murphy wrote:
 "Andrei Alexandrescu"  wrote in message
 news:lurk9p$123f$1 digitalmars.com...

 I think attributes would be the go-to approach. -- Andrei

 Why?

 Because they postdate pragma :o). I'm fine either way. -- Andrei

Those are different tools. Attributes are part of language
semantics, pragmas are to work with compiler internals. It is not
like we have `getPragmas` trait.

I agree that this is better fit for a pragma.

"Idan Arye" <GenericNPC gmail.com> writes:

On Thursday, 11 September 2014 at 09:09:41 UTC, Dicebot wrote:
 On Thursday, 11 September 2014 at 07:58:14 UTC, Andrei
 Alexandrescu wrote:
 On 9/11/14, 12:57 AM, Daniel Murphy wrote:
 "Andrei Alexandrescu"  wrote in message
 news:lurk9p$123f$1 digitalmars.com...

 I think attributes would be the go-to approach. -- Andrei

 Why?

 Because they postdate pragma :o). I'm fine either way. -- 
 Andrei

 Those are different tools. Attributes are part of language
 semantics, pragmas are to work with compiler internals. It is 
 not
 like we have `getPragmas` trait.

 I agree that this is better fit for a pragma.

Also, there already is a pragma for controlling the mangling, so
it'll be easier to extend it. Something like:

      extern (C++) pragma(mangle, struct) interface E {
          int bar(int i, int j, int k);
      }