Works great! LGTM! =)

I'm not giving karma because it solves the isnan.c reproducer, but doesn't solve the similar problem I'm facing.

I will try to submit a reproducer to the bugzilla ticket, but I see no reason to block this release since it's already an improvement.

There you go, reproducer:

$ cat isnan.c 
#include <assert.h>
#include <math.h>

int main(void)
{
  double foo = 1.0;

  assert(!isnan(foo));
  return 0;
}
$ clang -Wdouble-promotion -Wall -o isnan isnan.c
isnan.c:8:17: warning: implicit conversion increases floating-point precision: 'double' to 'long double' [-Wdouble-promotion]
  assert(!isnan(foo));
          ~~~~~~^~~~
/usr/include/math.h:928:46: note: expanded from macro 'isnan'
#  define isnan(x) __MATH_TG ((x), __isnan, (x))
                   ~~~~~~~~~~~~~~~~~~~~~~~~~~^~~
/usr/include/math.h:846:16: note: expanded from macro '__MATH_TG'
   : FUNC ## l ARGS)
     ~~~~~~~~~ ^~~~
/usr/include/assert.h:106:11: note: expanded from macro 'assert'
      if (expr)                                                         \
          ^~~~
1 warning generated.

I will add it to the bugzilla ticket.

works for me

no regressions noted