There seem to be multiple issues with generic functions as at
several places the generic symbols attributes (expr->symtree->n.sym)
and not the one of the specific (expr->value.functions.esym) is used.
The attached patch fixes some of those problems, but more remain as
the PR shows; in particular, the program in comment 0 of the PR
still does not work.
With regards to wrong-code, the issue is mostly whether the address
of the pointer or of the pointer target is used.
Bootstrapped and regtested on x86-64-linux.
OK for the trunk? What do you think about 4.4/4.3 backports?
Tobias