282 – Bizarre circular import nested name invisibility issue

I really don't know what else to call this, and it's a bit tedious to get to happen. So here's the general idea:
------ [modules/test1.d] --------
module modules.test1;
import modules.test2;
struct A
{
	enum B
	{
		One,
		Two
	}
}
void bar()
{
	foo();
}
--------- [modules/test2.d] ---------
module modules.test2;
import modules.test1;
A.B foobar; // error, no property B for type A?!
void foo()
{
	int x;
	if(x == A.B.One) // but this is OK
		x = 1;
}
---------------------------------
Some interesting facts:
- This only seems to happen if the modules are in a package. If they're two top-level modules, it works fine.
- Notice I can't use A.B as a type, but I can in fact access A.B's members.
- Using static imports/aliased imports does not help.
I don't know how to get it to not work the way it does in MiniD - that is, I can't even use A.B in the same module that A.B is defined!
This is, for me, something of a showstopper.. I'll probably be using 0.162 for a bit, which is sad, considering how cool the 0.163 features are! I don't know if I should change the priority of this bug, but this is something that can't be worked around easily, so I'd put it pretty high up there.

Related to #204. Maybe even same problem.

(In reply to comment #1)
> Related to #204. Maybe even same problem.
> 
Err. #102 . Sorry.

Created attachment 455 [details] 
ensure semantic run when searching struct for members
When searching members of a class that is not yet semantically analyzed, the analysis is run on it. This is not done for structs or unions, which is probably an omission.

Changeset 318

http://d.puremagic.com/issues/show_bug.cgi?id=282 

(In reply to comment #5)
> http://d.puremagic.com/issues/show_bug.cgi?id=282 
Damn! http://www.dsource.org/projects/dmd/changeset/318 

Fixed dmd 1.054 and 2.038