New __traits

[BLM768 via Digitalmars-d]

On Thursday, 26 November 2015 at 20:56:39 UTC, Jonathan M Davis 
wrote:
> Yes, code can forward-reference an import. e.g. this code 
> compiles just fine:
>
>     void main()
>     {
>         writeln("Where's my import?");
>     }
>
>     import std.stdio;
>
> Now, when the import is inside of a function, then it can't be 
> forward-referenced, but that's in line with how variable 
> declarations work.
>
> - Jonathan M Davis

Oh, duh. I should have remembered that.

In any case, I my tests work when I forward-reference, too, so 
I'll probably just put an assert on it and call it good. I've got 
some unit tests on my code now, so it looks like it's almost time 
for my first PR.

I don't know if this is at all related to how top-level packages 
aren't resolving to Package objects by the time the __traits run, 
but while testing this code, I found what seems to be a bug 
related to __traits(allMembers). Specifically, this code produces 
an extremely strange output:

---
import std.stdio;

pragma(msg, __traits(allMembers, std));

void main() {}
---

It lists a bunch of symbols that most certainly _aren't_ direct 
ancestors of the "std" package: "object", "core", "std", 
"KeepTerminator", "GCC_IO", "HAS_GETDELIM", "FSChar", and a bunch 
of others.

That's a bug, right?