[Issue 887] New: TypeInfo does not correctly override opCmp, toHash

[d-bugmail at puremagic.com]

http://d.puremagic.com/issues/show_bug.cgi?id=887

           Summary: TypeInfo does not correctly override opCmp, toHash
           Product: D
           Version: 1.001
          Platform: PC
        OS/Version: All
            Status: NEW
          Severity: normal
          Priority: P2
         Component: Phobos
        AssignedTo: bugzilla at digitalmars.com
        ReportedBy: fvbommel at wxs.nl


-----
urxae at urxae:~/tmp$ cat ti_comp.d 
import std.stdio;

class Foo {}
class Bar {}

void main() {
    TypeInfo ti_foo = typeid(Foo);
    TypeInfo ti_bar = typeid(Bar);

    writefln("typeid(Foo).toHash: ", ti_foo.toHash());
    writefln("typeid(Bar).toHash: ", ti_bar.toHash());

    writefln("opEquals: ", ti_foo == ti_bar ? "equal" : "not equal");
    writefln("opCmp: ", ti_foo.opCmp(ti_bar) == 0 ? "equal" : "not equal");
}

urxae at urxae:~/tmp$ dmd -run ti_comp.d 
typeid(Foo).toHash: 522954235
typeid(Bar).toHash: 522954235
opEquals: not equal
opCmp: equal
-----

TypeInfo.opCmp compares only the class, not the instance. This may be correct
for basic types such as char, int and long, but it's plain wrong for classes,
structs, pointers, and other user-defined or derived types.
opEquals *is* overridden correctly AFAICT. So basically any TypeInfo_* class
with an overridden opEquals needs to also override opCmp.

toHash is less serious, since it isn't *required* to be distinct but it has
essentially the same issue.

The combined effect of these defects is that TypeInfo is useless as an
associative array key type. A workaround is to use char[] keys, and use
TypeInfo.toString(). (toString seems to be overridden just fine as well)

A quick fix for both of these problems would be to replace
'this.classinfo.name' in both TypeInfo.opCmp and TypeInfo.toHash with
'this.toString()', though that's probably not the most efficient method.


--