On 7/22/2017 2:04 AM, Martin Nowak wrote:
Also translating the C++ backend to D zero benefit to D users (at worst it introduces codegen bugs). I'm inclined to say we should rather spent our time on the various more important issues. It was a good move for the frontend as that will allow us to make use of D features to improve the code/architecture and to eventually turn it into a proper library. Converting the stable and well tested codebase of the backend, which will hardly receive any feature development, is unlikely to ever pay off.

It'll be converted anyway. :-)

I'm about half way through converting the DMC++ front end. It isn't as difficult as I expected. The -betterC is helping a lot, as well as the fact that DMC++ is still pretty much C code. Some years back Daniel Murphy and I removed a lot of the preprocessor trickery from it, which helps, too.

Note that DMC++ is now Boost licensed as well.

Putting the entire set in D (C compiler, C++ compiler, C preprocessor, htod converter, optimizer, code generator) makes the whole thing much more tractable, and who knows what we will be able to do with it!

Reply via email to