D should disallow forward references

[Don]

Leandro Lucarella wrote:
> Don, el 26 de agosto a las 17:19 me escribiste:
>> Leandro Lucarella wrote:
>>> Don, el 26 de agosto a las 09:37 me escribiste:
>>>> I have a copy of Walter's internal DMD test suite, so I could actually
>>> What was the reason not to release the test suite? Copyright?
>> Yes.
>>
>>> Because Walter has claimed frequently that the submitted patches are not
>>> well tested. It would be a huge help for people contributing patches to
>>> have the "official" test suite to be sure their patches would be accepted.
>> Yes, though you don't need the test suite, just the results. You just
>> need to know: "that passed all tests" or "here are test cases which it
>> fails".
> 
> That would improve the situation, but not being able to see what's in the
> test case that failed could be a little.... frustrating. 

Of course you'd see the test case which failed. Most of them are in 
Bugzilla, so it'd usually just be a bugzilla link. It's only the ones 
from the early days of D which are problematic, I think. And they are 
unlikely to fail.
BTW, I don't have Walter's whole test suite, eg I certainly don't have 
any of his C++ test suite which is where the serious copyright question 
marks lie.

And it would
> require some sort of public "testing" facility where people can test their
> patches, with could be a little.... hard to implement =)

Can be manual. I don't imagine there's going to be THAT many patches.

Actually, I've been doing this a bit anyway, there are several 
contributed patches I've tested and then passed onto Walter. From memory 
there will be 3 of those in the next release. (BTW, patches by Shin are 
always excellent, I wish he'd do more of them).

> Ok, you have a point here, seeing the DMD BE merely as a personal tool
> for doing FE development might have sense. But it's really a little weird.
> That's not how free/open source development usually works. I'm not saying
> it *can't* work, but it will be certainly harder to find people willing to
> work that way...

Yeah, probably true, although LDC is available too. I've submitted over 
80 patches in 20 weeks, which is about the rate that new unfixed bugs 
appear. So we need probably two or three more people working at the same 
rate to clear the backlog in a decent amount of time. Which is a very 
big ask.

Actually I still think we need library developers more than compiler 
bgufixers.