No description provided.
An interesting idea, but modern nqp is a much larger project than nqp-rx. It not only has a more-or-less completely separate object system, 6model, but also separate dynamic call and big integer libraries.
That said, I think we do need to seriously look at stealing back everything from nqp into core. I'd have no arguments if dyncall, tommath, and 6model all were integrated into Parrot. I'd prefer integration over simple bundling.
I agree with benabik. We already want to "borrow" 6model from nqp into Parrot core. Getting some of the other features like dyncall and tommath would be interesting ideas as well.
NQP-rx is old and showing its age. A more modern NQP would bring a lot of improvements and benefits. Not the least of which is more motivation to finally migrate 6model to Parrot core.
I didn't said how it should be done. Yes, we do need to borrow "6model". We do need "bigint" libraries. Etc, etc, etc.
Target goal stays same - have modern nqp bundled.
@bacek how do you suggest we actually bundle a modern nqp? Git submodule? How much does our build/configure system need to change?
I suppose it's a wording nit, but "bundle" made me think "just include their code", as we do for nqp-rx and winxed already. I've seen requests like that in the channel before and I wanted to make it clear that I don't like that approach but wholeheartedly endorse grabbing the things that nqp has done and making it core.
Ah, ok. Sorry for misleading with ticket wording. I definitely doesn't want "bolt on" approach with "nqp bundling". We have to have proper integration of 6model, provide enough hooks for different MMD, etc.
I've created a page in the wiki to hold wishlist and roadmap items relating to Rakudo and Perl6 integration ideas such as this. I'm closing this ticket and we can continue the discussion on the wiki