https://gcc.gnu.org/bugzilla/show_bug.cgi?id=102064
Bug ID: 102064
Summary: Wrong assignable check in uninitialized_fill and
uninitialized_fill_n
Product: gcc
Version: 12.0
Status: UNCONFIRMED
Keywords: rejects-valid
Severity: normal
Priority: P3
Component: libstdc++
Assignee: unassigned at gcc dot gnu.org
Reporter: redi at gcc dot gnu.org
Target Milestone: ---
In r204615 I added assignable checks to various algos to fix PR
libstdc++/58982.
The checks in uninitialized_fill and uninitialized_fill_n use
is_copy_assignable, which is wrong because the value being filled doesn't have
to be the value type of the output range.
This valid code fails to compile with libstdc++:
#include <memory>
#include <algorithm>
struct X
{
X() = default;
X(int) { }
X(const X&) = default;
X& operator=(const X&) = default;
X& operator=(int) = delete;
};
static_assert( std::is_trivial<X>::value, "" );
int main()
{
unsigned char buf[sizeof(X)];
std::uninitialized_fill_n((X*)buf, 1, 99);
}
/usr/include/c++/11/bits/stl_algobase.h:924:18: error: use of deleted function
‘X& X::operator=(int)’
924 | *__first = __tmp;
| ~~~~~~~~~^~~~~~~
The problem is that we decide it's OK to use std::fill_n because X is trivial
and is_copy_assignable<X> is true, but then std::fill_n assigns int to X.
We need to check is_assignable<X&, const int&> instead.
