Frigg assert / illegal instruction while running configure

[Dennisbonke]

While running configure in GNU sed, for the checking for working nanosleep check, an illegal instruction is triggered. IP points to libc.so, and addr2line gives /var/lib/managarm-buildenv/build/system-root/usr/share/frigg/include/frg/mutex.hpp:62 (discriminator 3) (source code here).

[Dennisbonke]

One of the tests that consistently reproduces this is the following code snippet

#include <float.h>
#include <math.h>
#include <string.h>
#if HAVE_DECL_ALARM
# include <signal.h>
# include <unistd.h>
#endif
/* HP cc on HP-UX 10.20 has a bug with the constant expression -0.0.
   ICC 10.0 has a bug when optimizing the expression -zero.
   The expression -DBL_MIN * DBL_MIN does not work when cross-compiling
   to PowerPC on Mac OS X 10.5.  */
#if defined __hpux || defined __sgi || defined __ICC
static double
compute_minus_zero (void)
{
  return -DBL_MIN * DBL_MIN;
}
# define minus_zero compute_minus_zero ()
#else
double minus_zero = -0.0;
#endif
int main()
{
  int result = 0;
  int i;
  volatile double x;
  double zero = 0.0;
#if HAVE_DECL_ALARM
  /* NeXTstep 3.3 frexp() runs into an endless loop when called on an infinite
     number.  Let the test fail in this case.  */
  signal (SIGALRM, SIG_DFL);
  alarm (5);
#endif
  /* Test on denormalized numbers.  */
  for (i = 1, x = 1.0; i >= DBL_MIN_EXP; i--, x *= 0.5)
    ;
  if (x > 0.0)
    {
      int exp;
      double y = frexp (x, &exp);
      /* On machines with IEEE754 arithmetic: x = 1.11254e-308, exp = -1022.
         On NetBSD: y = 0.75. Correct: y = 0.5.  */
      if (y != 0.5)
        result |= 1;
    }
  /* Test on infinite numbers.  */
  x = 1.0 / zero;
  {
    int exp;
    double y = frexp (x, &exp);
    if (y != x)
      result |= 2;
  }
  /* Test on negative zero.  */
  x = minus_zero;
  {
    int exp;
    double y = frexp (x, &exp);
    if (memcmp (&y, &x, sizeof x))
      result |= 4;
  }
  return result;
}