This is specialized enough that I just put it inline rather than splitting it out to a new function.

I forgot the runtime parameter in the assertion and then didn't bother compile testing such a trivial debug-only change.

Comment on attachment 745399 [details] [diff] [review] v1: facepalm Review of attachment 745399 [details] [diff] [review]: ----------------------------------------------------------------- ::: js/src/jsobj.cpp @@ +1776,5 @@ > { > Rooted<TypeObject*> typeObj(cx); > typeObj = cx->global()->getOrCreateObjectPrototype(cx)->getNewType(cx, &ObjectClass); > + > + AllocKind kind = GetGCObjectFixedSlotsKind(srcObj->numFixedSlots()); I think you want GuessObjectGCKind. Otherwise this will fail if the literal has more than 16 slots. @@ +1777,5 @@ > Rooted<TypeObject*> typeObj(cx); > typeObj = cx->global()->getOrCreateObjectPrototype(cx)->getNewType(cx, &ObjectClass); > + > + AllocKind kind = GetGCObjectFixedSlotsKind(srcObj->numFixedSlots()); > + if (CanBeFinalizedInBackground(kind, srcObj->getClass())) I think our literals are always of class js::ObjectClass. Can you assert that and then use the background kind unconditionally? @@ +1779,5 @@ > + > + AllocKind kind = GetGCObjectFixedSlotsKind(srcObj->numFixedSlots()); > + if (CanBeFinalizedInBackground(kind, srcObj->getClass())) > + kind = GetBackgroundAllocKind(kind); > + JS_ASSERT_IF(!IsInsideNursery(cx->runtime, srcObj), kind == srcObj->tenuredGetAllocKind()); It seems more natural to use srcObj->isTenured for the test here.

Green try run: https://tbpl.mozilla.org/?tree=Try&rev=8ea3d8d7ce1d