On Thursday, 17 September 2020 at 09:53:57 UTC, Remi wrote:
My problem here is mostly understanding the __initZ symbol and where it comes from.
The compiler generates that when it spits out something that uses TypeInfo, which is a lot of things: ~= and ~ operators on built in arrays, the new operator, class dynamic casting, exception catching, and more. Annoyingly, even having a class member of a type will trigger a TypeInfo dependency too. Why? idk, probably to do with precise GC hook support.
The error you mentioned in your other message about the order of files is interesting.. I suspect that is because the compiler creates these symbols when it first sees a need... then tries to reuse it, and that must have made the conflict here since it made it in the wrong context... or something.
The error says something about constness being different, well, why the eff would an init symbol ever be mutable? And ldc is so strict about this stuff too where dmd just doesn't care and that can be annoying too.
idk this probably doesn't help you much since I don't know why it is generating it this way either. It is possible it is actually pulling the definition from the original druntime or something.
(on my webassembly.arsdnet.net site btw I made my own little on-demand build web service to pass the right args. p convenient but also specialized for me and prolly not too general purpose)
I'll keep investigating as I have time, but I do kinda prefer th idea of improving the system rather than hacking around it lol
I'll probably try what you describe in your "Zero-runtime classes" actually, and hopefully I can get away without the string manipulation or replace it with my own version of it.
oh the string manip there is just looking up the symbol, it isn't too bad.
The betterC thing also kinda sorta works with classes btw if you combine this with extern(C++) class definitions.
