This is likely a bug in dmd, probably related to http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.D.bugs/1810

dmd 0.101 on Linux.


#import mintl.util;
#
#/* Segmentation fault when Base is an interface. Okay when Base is a #class. */
#interface Base
#//class Base
#{
#	int id();
#//	int id() { return 0; }
#}
#
#class Foo : Base
#{
#	int id() { return 123; }
#}
#
#class Bar : Base
#{
#	int id() { return 987; }
#}
#
#int main()
#{
#	Foo foo = new Foo;	// Declaring foo and bar as Base
#	Bar bar = new Bar;	// does not help.
#
#	SortedSet!(Base) set;
#	
#	set[foo] = foo;
#	set[bar] = bar;
#
#	printf("Length set = %d.\n", set.length());
#	
#	foreach( Base b; set ) printf("id: = %d.\n", b.id() );
#	return 0;
#}

Bummer. It looks like TypeInfo compare functions are busted for interfaces.
If you try to run
 TypeInfo ti = typeid(foo);
 ti.compare(&foo,&bar);
you'll error out. On win32 I get a Win32 Exception. Probably since the
declaration of compare is with void* instead of a specific type it gets all
confused about calling opCmp. Nasty. I'm cross-posting to bugs.

"Bastiaan Veelo" <Bastiaan.N.Veelo@ntnu.no> wrote in message news:ci495h$1alr$1@digitaldaemon.com...
> This is likely a bug in dmd, probably related to http://www.digitalmars.com/drn-bin/wwwnews?digitalmars.D.bugs/1810
>
> dmd 0.101 on Linux.
>
>
> #import mintl.util;
> #
> #/* Segmentation fault when Base is an interface. Okay when Base is a
> #class. */
> #interface Base
> #//class Base
> #{
> # int id();
> #// int id() { return 0; }
> #}
> #
> #class Foo : Base
> #{
> # int id() { return 123; }
> #}
> #
> #class Bar : Base
> #{
> # int id() { return 987; }
> #}
> #
> #int main()
> #{
> # Foo foo = new Foo; // Declaring foo and bar as Base
> # Bar bar = new Bar; // does not help.
> #
> # SortedSet!(Base) set;
> #
> # set[foo] = foo;
> # set[bar] = bar;
> #
> # printf("Length set = %d.\n", set.length());
> #
> # foreach( Base b; set ) printf("id: = %d.\n", b.id() );
> # return 0;
> #}