From @nwc10

sv.c
9837​: sv_pos_u2b(argsv, &p, 0); /* sticks at end */

This is deep in the bowls of printf​:

  case 's'​:
  if (vectorize)
  goto unknown;
  if (args) {
  eptr = va_arg(*args, char*);
  if (eptr)
  elen = strlen(eptr);
  else {
  eptr = (char *)nullstr;
  elen = sizeof nullstr - 1;
  }
  }
  else {
  eptr = SvPV_const(argsv, elen);
  if (DO_UTF8(argsv)) {
  STRLEN old_precis = precis;
  if (has_precis && precis < elen) {
  STRLEN ulen = sv_len_utf8(argsv);
  I32 p = precis > ulen ? ulen : precis;
  sv_pos_u2b(argsv, &p, 0); /* sticks at end */
  precis = p;
  }
  if (width) { /* fudge width (can't fudge elen) */
  if (has_precis && precis < elen)
  width += precis - old_precis;
  else
  width += elen - sv_len_utf8(argsv);
  }
  is_utf8 = TRUE;
  }
  }

I *think* all the surroundings translate as %s, but with a length constraint,
eg %42s, where the constraint is less than the (octed) length of a UTF-8
string. In which case, work out where the 42nd character goes up to.

So, %2147483648s and similar aren't going to work (well) on UTF-8 strings
over 2GB.

Nicholas Clark

From @cpansprout

Exact copy of 72776.

@cpansprout - Status changed from 'new' to 'resolved'