Regression on beta for trait bounds on supertrait's associated items

[Ryman]

The following playpen works on 1.0, Stable(1.1) and Nightly(1.3 bf3c979 2015-06-30) but fails to compile on current Beta (1.2).

http://is.gd/cQJ1Xk

It would be good if we could backport whatever fixed the behavior for 1.3.

cc @nagisa

[huonw]

triage: I-nominated

Seems to be a regression.

[nikomatsakis]

It's unclear what's changed here. @nrc do you think this could be the DST coercion stuff? It wasn't obvious to me that it was, but it's a possible candidate, in which case @eddyb was perhaps right ;P.

[nikomatsakis]

triage: P-high (regression)

[nikomatsakis]

Not sure who to assign this one too! I'll take it for now.

[nrc]

Seems unlikely to be DST-related to me

[brson]

@nikomatsakis Any new insight into this one? I believe this is about to hit stable in 1.3.

[eddyb]

@brson It works on beta and nightly on the playpen, but not stable, so presumably it will work again in 1.3.

[brson]

@eddyb Oh, so this is fixed? Can we close?

[eddyb]

cc @nikomatsakis Can you confirm? I'm not entirely sure what the story was here.

[nikomatsakis]

I'm not 100% sure what happened here. It sure appears that there was a transient bug which was fixed, but I'm not sure what that bug was. The code appears legit at first glance.

[nikomatsakis]

So basically I think yes we can close.