[OT] Re: C's Biggest Mistake on Hacker News

Abdulhaq alynch4047 at gmail.com
Fri Aug 3 20:07:33 UTC 2018


On Tuesday, 31 July 2018 at 22:55:08 UTC, Laeeth Isharc wrote:

> Dpp doesn't work with STL yet.  I asked Atila how long to 
> #include vector and he thought maybe two months of full-time 
> work.  That's not out of the question in time, but we have too 
> much else to do right now.  I'm not sure if recent mangling 
> improvements help and how much that changes things.  But DPP 
> keeps improving as does extern (C++) and probably one way and 
> another it will work for quite a lot.  Calypso makes cpp 
> classes work as both value and reference types.  I don't know 
> the limit of what's possible without such changes - seems like 
> C++ mangling is improving by leaps and bounds but I don't know 
> when it will be dependable for templates.
>

Yes OK, thanks.

> It's not that relevant what Andrei or Walter might think 
> because it's a community-led project and we will make progress 
> if somebody decides to spend their time working on it, or a 
> company lends a resource for the same purpose.  I'm sure they 
> are all in favour of greater cpp interoperability, but I don't 
> think the binding constraint is will from the top, but rather 
> people willing and able to do the work.
>

I think the DIP system has greatly improved the situation, but 
for anyone thinking of embarking on a lot of work for something 
like e.g. the GC, you do need to feel that there will be a good 
chance of it being adopted - otherwise all that work could go to 
waste.

> And if one wants to see it go faster then one can logically 
> find a way to help with the work or contribute financially.  I 
> don't think anything else will make a difference.
>

Agreed entirely.

> Same thing with Calypso.  It's not ready yet to be integrated 
> in a production compiler so it's an academic question as to the 
> leadership's view about it.

Where I'm coming from is that writing and maintaining something 
as large and complex as Calypso requires a whole heap both of 
motivation and also of encouragement from the sidelines - and 
especially from Walter and/or Andrei. If someone starts to feel 
that the backing is not there then it's very very hard to 
maintain motivation, particularly on infrastructure related code 
that if not integrated by Walter will always be hard for people 
to use and therefore not be widely adopted.

To be fair to Walter though, this is a really intractable problem 
for him. He could adopt something like Calypso, and then find the 
original maintainer loses interest. That would leave Walter 
either needing to maintain someone else's complex code, or try to 
extricate himself from code having already integrated it. Also, 
there is no guarantee, in this particular case, that as C++ 
evolves it will still be possible to use Calypso's strategy. Of 
course there are other very good reasons for why adopting it is 
problematic. Still, it leaves the developer struggling, I expect, 
to maintain motivation.

Considering the above, then knowing the general direction that 
Walter/Andrei want to take D, would be a great help in deciding 
what larger projects are worth undertaking. It seems to me, 
anyway (big caveat).






More information about the Digitalmars-d mailing list