libstdc++: Avoid overflow in ranges::advance(i, n, bound)

When (bound - i) or n is the most negative value of its type, the
negative of the value will overflow. Instead of abs(n) >= abs(bound - i)
use n >= (bound - i) when positive and n <= (bound - i) when negative.
The function has a precondition that they must have the same sign, so
this works correctly. The precondition check can be moved into the else
branch, and simplified.

The standard requires calling ranges::advance(i, bound) even if i==bound
is already true, which is technically observable, but that's pointless.
We can just return n in that case. Similarly, for i!=bound but n==0 we
are supposed to call ranges::advance(i, n), but that's pointless. An LWG
issue to allow omitting the pointless calls is expected to be filed.

libstdc++-v3/ChangeLog:

	* include/bits/ranges_base.h (ranges::advance): Avoid signed
	overflow. Do nothing if already equal to desired result.
	* testsuite/24_iterators/range_operations/advance_overflow.cc:
	New test.
This commit is contained in:
Jonathan Wakely 2022-01-26 16:08:51 +00:00
parent 66b8617118
commit f21f22d1ba
2 changed files with 46 additions and 6 deletions

View file

@ -756,20 +756,23 @@ namespace ranges
{
const auto __diff = __bound - __it;
// n and bound must not lead in opposite directions:
__glibcxx_assert(__n == 0 || __diff == 0 || (__n < 0 == __diff < 0));
const auto __absdiff = __diff < 0 ? -__diff : __diff;
const auto __absn = __n < 0 ? -__n : __n;;
if (__absn >= __absdiff)
if (__diff == 0)
return __n;
else if (__diff > 0 ? __n >= __diff : __n <= __diff)
{
(*this)(__it, __bound);
return __n - __diff;
}
else
else if (__n != 0) [[likely]]
{
// n and bound must not lead in opposite directions:
__glibcxx_assert(__n < 0 == __diff < 0);
(*this)(__it, __n);
return 0;
}
else
return 0;
}
else if (__it == __bound || __n == 0)
return __n;

View file

@ -0,0 +1,37 @@
// { dg-options "-std=gnu++20" }
// { dg-do compile { target c++20 } }
// Public domain testcase from Casey Carter, send to LWG list on 2021-07-24.
//
// Here's a compile-only test case for which n is INT_MIN, which will overflow
// if simply negated to get |n|: https://godbolt.org/z/M7Wz1nW58.
#include <cassert>
#include <iterator>
#include <limits>
struct I {
using difference_type = int;
using value_type = int;
int x;
constexpr int operator*() const { return x; }
constexpr I& operator++() { ++x; return *this; }
constexpr I operator++(int) { ++x; return {x - 1}; }
constexpr bool operator==(const I&) const = default;
constexpr int operator-(const I& that) const { return x - that.x; }
constexpr I& operator--() { --x; return *this; }
constexpr I operator--(int) { --x; return {x - 1}; }
};
static_assert(std::bidirectional_iterator<I>);
static_assert(std::sized_sentinel_for<I, I>);
constexpr bool test() {
using L = std::numeric_limits<int>;
I i{-2};
return std::ranges::advance(i, L::min(), I{-4}) == L::min() + 2;
}
static_assert(test());