This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [C++ Patch/RFC] Final part of c++/26099, __is_convertible_to
Mark Mitchell wrote:
I see no inherent reason why we don't call build_user_type_conversion_1
when expr == NULL_TREE; I expect that when we expr == NULL_TREE, we must
know that we don't have class types involved. But, yes, this would have
to change, and, yes, you may need a dummy object at some point. But,
creating it later (in build_user_type_conversion_1, for example) would
be better because at least you would not create dummy objects for easy
conversions. And, if you do need a dummy object, you should be using
build_dummy_object.
Ok, thanks for the additional details. Of course I agree that delaying
the creation of a dummy object as much as possible is an excellent idea.
Note that really the problem isn't only with user type conversions, it's
also with access checking, ambiguities... in short, as far as I can see,
convert_like_real, called by perform_implicit_conversions, does *a lot*
of work.
Paolo.