https://gcc.gnu.org/bugzilla/show_bug.cgi?id=99851
Bug ID: 99851
Summary: Warn about operator new that takes std::nothrow_t but
is potentially-throwing
Product: gcc
Version: 11.0
Status: UNCONFIRMED
Keywords: diagnostic
Severity: enhancement
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: redi at gcc dot gnu.org
Blocks: 87403
Target Milestone: ---
This program crashes with a segfault:
#include <new>
void* null() { return nullptr; }
struct X
{
void* operator new[](std::size_t, const std::nothrow_t&) {
return null();
}
unsigned data = 0;
};
int main()
{
new(std::nothrow) X[2];
}
The problem is that the new overload is not noexcept, so the compiler assumes
it can't return null. The user probably intended it to be a non-throwing form
of operator new (as implied by the nothrow_t parameter), so we should warn that
it isn't noexcept.
N.B. if the function return nullptr directly then we warn:
new.C: In static member function ‘static void* X::operator new [](std::size_t,
const std::nothrow_t&)’:
new.C:6:12: warning: ‘operator new’ must not return NULL unless it is declared
‘throw()’ (or ‘-fcheck-new’ is in effect)
6 | return nullptr;
| ^~~~~~~
That should be updated to say 'noexcept' not 'throw()' and we might want the
two warnigns to use similar phrasing. That warning should also say "a null
pointer" not NULL.
Referenced Bugs:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87403
[Bug 87403] [Meta-bug] Issues that suggest a new warning
