Itcl - the [incr Tcl] extension

In ItclClassBaseCmd(), the flags field of an ItclMemberFunc
is tested for whether the flag ITCL_IMPLEMENT_NONE is set.

http://core.tcl.tk/itcl/info/07f70b5b42?ln=779

Trouble is that this flag value is only ever set in the flags
field of an ItclMemberCode struct, *not* an ItclMemberFunc struct.

This turns the conditional test into one that is always true, which
appears to be against the coded intent.

Suspect that what is meant here is to test this instead:

if (!(imPtr->codePtr->flags & ITCL_IMPLEMENT_NONE)) {

so that an additional dereference gets us a struct of the right type.

Note that would be better written with use of the macro:

http://core.tcl.tk/itcl/info/a6c3f38589?ln=779

which improves the clarity.

So why bother with a ticket?  Because making this "obvious" fix
produces failures and segfaults in the test suite.

The segfault comes from something apparently totally broken.

the "functions" hash table stores values of (ItclMemberFunc *).

However the routine ItclBiObjectUnknownCmd can pull
values from that hash table, and then start acting on them
as if they were (ItclDelegatedFunction *) values.

The two structs do not share a common structure.  Hilarity ensues.

It seems that without my flags experimenting the test suite just
doesn't cover cases broken by this error.

Appears the attempt to keep unimplemented methods
secret from the TclOO machinery was just the wrong
idea to begin with, so the buggy failure to do so
was accidentally making things work.

And masking the broken bit of the unknown method
handler, that is now fixed.