We decided this would be a better thing to do that bug 983719.

Lots of small fixups; they all seem to be fallout from the s/nsILocalFile/nsIFile/ conversion. Detecting conversions to superclasses would be easy, I think, but we'll leave that as a followup.

Comment on attachment 8392305 [details] [diff] [review] change CallQueryInterface to assert in cases of trivial conversions Review of attachment 8392305 [details] [diff] [review]: ----------------------------------------------------------------- ::: xpcom/glue/nsISupportsUtils.h @@ +132,5 @@ > + // We permit nsISupports-to-nsISupports here so that one can still obtain > + // the canonical nsISupports pointer with CallQueryInterface. > + static_assert(!mozilla::IsSame<T, DestinationType>::value || > + mozilla::IsSame<DestinationType, nsISupports>::value, > + "don't use CallQueryInterface for compile-time-determinable casts"); I assume you have bsmedberg's blessing on this, but this looks sane.

Yes, this did have Benjamin's blessing. :) https://hg.mozilla.org/integration/mozilla-inbound/rev/e17405b3cc01