This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Bug in mudflap?
I wrote:
> It looks like some instrumentation may be missing, related to the
> return-by-value of object instances (in this case, the std::vector's
> normal_iterator in its end() member function).
Specifically, this seems to occur because the local variable destined
for use as the return-by-value destination, is never marked
TREE_ADDRESSABLE by the front end. This, despite the fact that the
variable's address *is* taken and sent to the value-creating function
in its <retval> slot. Can a C++ person shed light on this?
- FChE