https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85873
Jason Merrill <jason at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords|wrong-code |missed-optimization
Depends on| |67445
--- Comment #2 from Jason Merrill <jason at gcc dot gnu.org> ---
This is actually only a missed optimization, your code has undefined behavior.
11.6.4 [dcl.init.list]:
An object of type std::initializer_list<E> is constructed from an initializer
list as if the implementation generated and materialized (7.4) a prvalue of
type “array of N const E”, where N is the number of elements in the initializer
list....
The array has the same lifetime as any other temporary object (15.2), except
that initializing an initializer_list object from the array extends the
lifetime of the array exactly like binding a reference to a temporary....
[ Note: The implementation is free to allocate the array in read-only memory if
an explicit array with the same initializer could be so allocated. — end note ]
So the array's lifetime only lasts until the end of the return-statement; when
the function returns, the array is gone, and so the range for iterates a
dangling pointer. As the note says, a compiler is permitted to put it in
.rodata as an optimization, just as it can for
const int ar[] = { 1,2,3 };
but it isn't required to do so.
Bug 67445 asks for a warning about this situation, which I have now implemented
and will check in before long.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=67445
[Bug 67445] New warning: returning std::initializer_list bound to temporary
