How to make D resolve C++ symbols by mangling symbols with the Itanium ABI on Windows

[thumbgun]

On Monday, 26 February 2024 at 20:07:55 UTC, user1234 wrote:
> On Monday, 26 February 2024 at 19:50:31 UTC, user1234 wrote:
>> On Monday, 26 February 2024 at 17:23:32 UTC, thumbgun wrote:
>>> [...]
>>
>> I see. That would be extremely painful. Could CTFE help here ?
>>
>> ```d
>> string getRightMangling(alias T)()
>> {
>>     return "todo" // ;)
>> }
>>
>> pragma(mangle, getRightMangling!(add))
>> extern(C++) int add(int lhs, int rhs);
>> ```
>>
>> That's not a serious suggestion but I'm still curious to see 
>> how that would scale.
>>
>> Also that looks weird that the pragma refers to what it is 
>> supposed to affect, not sure that would work, although in 
>> theory D manages forward refs.
>
> Turns out that the forward ref is not a problem. Remains the 
> question of writing a mangler working fast at compile time ;)

Although I'm not quite sure what available features are permitted 
for CTFE, I'll definitely have a look at that. Johan raised an 
important point though about calling conventions as they are 
different between the Itanium and the MSVC ABIs. So even if DMD 
can link against Itanium mangled symbols, it won't follow the 
same calling convention.

Another alternative apart from compiling libraries from source 
with MSVC would be to create a C API to wrap the C++ library but 
that's also time consuming.