On 15/07/16 13:34 -0400, Ed Smith-Rowland wrote:

OK for trunk, thanks.

I didn't see a feature test in any of the SD-6 papers or P0025.

p0096r3 proposes __cpp_lib_clamp = 201603.
I added the feature macro and committed the attached as 238383.

Thanks,

Ed


2016-07-15  Edward Smith-Rowland  <3dw...@verizon.net>

        Implement C++17 P0025 clamp.
        * include/bits/algorithmfwd.h: Declare clamp overloads.
        * include/bits/stl_algo.h: Implement clamp.  Feature __cpp_lib_clamp.
        * testsuite/25_algorithms/clamp/1.cc: New test.
        * testsuite/25_algorithms/clamp/2.cc: New test.
        * testsuite/25_algorithms/clamp/constexpr.cc: New test.
        * testsuite/25_algorithms/clamp/requirements/explicit_instantiation/
        1.cc: New test.
        * testsuite/25_algorithms/clamp/requirements/explicit_instantiation/
        pod.cc: New test.


+  const int xc = std::clamp(1, 2, 4, std::greater<int>());
+  const int yc = std::clamp(3, 2, 4, std::greater<int>());
+  const int zc = std::clamp(5, 2, 4, std::greater<int>());

These all violate the precondition that !comp(hi, lo)
i.e. !greater<int>(1, 0)

The arguments need to be re-arranged, and then these are wrong:

+  VERIFY( xc == 4 );
+  VERIFY( yc == 2 );
+  VERIFY( zc == 2 );


+static_assert(std::clamp(2, 0, 1) == 1, "");
+static_assert(std::clamp(2, 0, 1, std::greater<int>()) == 0, "");

Same here. If the arguments are clamp(2, 1, 0, greater<int>()) then it
returns 1.

Reply via email to