BetterC Is Not Ready

[Dukc]

On Thursday, 2 June 2022 at 19:40:49 UTC, Jack Stouffer wrote:
> It's currently practically impossible to translate existing D 
> code to use BetterC in a large code base (haven't tried 
> existing C code, I imagine it's easier). I have an existing app 
> which is ~100,000 lines of D code and I've tried and failed to 
> convert it to BetterC with my own custom runtime.

I have also used `-betterC` to some extent. Unfortunately I have 
to agree that you're mostly right.

IMO `-betterC` works somewhat well when you don't use Phobos 
much. But if you're using it to any significant extent it's far 
from working out of the box. I learned that one should copypaste 
and adapt stuff from Phobos and DRuntime when they don't work 
right away. `-i` to the standard library or `-allinst` might also 
help, have not tried them myself.

In principle all the ranges stuff that does not use the GC should 
work seamlessly with `-betterC` In practice the problem is that 
some of the `-betterC` features are implemented by simply not 
linking the Phobos binary in. Result is that the Phobos source 
files are processed as header files, but without anything to link 
to if anything precompilable is found in the files. It might be 
most practical to simply use the compiler switch to link the 
Phobos binary in with `-betterC`. On the other hand, I don't know 
how much this defeats the purpose of using `-betterC` in the 
first place.