Question on tai to utc conversion

[timweckx]

In eraTaiutc, there is an iteration block that calls back to eraUtctai to get the leap seconds deltas. In the iteration, the deltas for both arguments are added to the 2nd argument. Should the first delta be applied to the first argument?

/* Iterate (though in most cases just once is enough). */
   for ( i = 0; i < 3; i++ ) {

   /* Guessed UTC to TAI. */
      j = eraUtctai(u1, u2, &g1, &g2);
      if ( j < 0 ) return j;

   /* Adjust guessed UTC. */
      u2 += a1 - g1;    --> should this be u1 += a1 - g1
      u2 += a2 - g2;
   }

In practice, I think the first delta is always zero so it doesn't affect the actual values.

[mhvk]

Thanks for the question! As you noticed, in general g1 and g2 are both small. Since u1 is normally a much larger number, than u1 adding to u1 thus risks rounding errors.