[Bug-gnulib] Rounding bug in human_readable from human.c

[Lute Kamstra]

Dear people,

There seems to be a rounding bug in human_readable from human.c.

The following code:

  char buf[LONGEST_HUMAN_READABLE + 1];
  int n;
  for (n = 10000; n <= 11000; n++)
    printf ("%d is written as %s\n", 
            n, 
            human_readable (n, buf, 
                            human_round_to_nearest |
                            human_autoscale | 
                            human_SI,
                            1, 1));

produces this output:

  10000 is written as 10k

  [...]

  10550 is written as 10k
  10551 is written as 11k

  [...]

  11000 is written as 11k

I would expect the transition from 10k to 11k at 10500 as opposed to
10550.


Kinds regards,

  Lute.