[Bug libstdc++/58800] [4.8 regression, I think] std::nth_element segfaults on valid input

[paolo.carlini at oracle dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=58800

--- Comment #5 from Paolo Carlini <paolo.carlini at oracle dot com> ---
Don't tell me is just this:

Index: include/bits/stl_algo.h
===================================================================
--- include/bits/stl_algo.h    (revision 203835)
+++ include/bits/stl_algo.h    (working copy)
@@ -1917,7 +1917,7 @@
                 _RandomAccessIterator __last, _Compare __comp)
     {
       _RandomAccessIterator __mid = __first + (__last - __first) / 2;
-      std::__move_median_to_first(__first, __first + 1, __mid, (__last - 2),
+      std::__move_median_to_first(__first, __first + 1, __mid, (__last - 1),
                   __comp);
       return std::__unguarded_partition(__first + 1, __last, __first, __comp);
     }

??

Andy, while waiting for Chris' feedback, you may want to test the above on your
real applications. In 4.8.x some nth_algorithm details are different (the above
is versus mainline), but you can quickly find a couple of (__last - 2) which
I'm asking you to change to (__last - 1)