https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65780
--- Comment #32 from Jeffrey A. Law <law at redhat dot com> --- HJ, you're missing my point. I need to understand the meaning of the variable defined_locally to move forward with the patch. Is it a must property (ie, if true, the symbol is always defined locally), or is it a may property (ie, if set, the symbol may be defined locally, but might also be defined externally). Much of the correctness of the patch hinges on that question.
