Bug 8286 – Type resolution issue during extra inlining semantic

This bug was originally reported by David Simcha as a LDC issue (https://github.com/ldc-developers/ldc/issues/62), but it occurs with DMD as well: --- a.d import b; --- --- b.d import hash; void fun() { alias StackSet!(string) SS; } --- --- hash.d struct HashRange(K, S, ) { size_t _length; this(S* set) { this._length = set.length; } } struct StackSet(K) { HashRange!(K, typeof(this)) elems() { assert(0); } @property size_t length() { assert(0); } } --- With DMD 2.059, dmd -release -inline -noboundscheck a.d yields: --- hash.d(5): Error: cannot implicitly convert expression ((*set).length()) of type size_t to ulong hash.d(11): Error: template instance hash.HashRange!(string,StackSet!(string)) error instantiating b.d(4): instantiated from here: StackSet!(string) b.d(4): Error: template instance hash.StackSet!(string) error instantiating --- With DMD 2.060 Git (ad7637f), it instead segfaults in AggregateDeclaration::semantic3.

With 2.060 release, no longer ICEs. And 2.062 and later, the OP code compiles successfully.