Paul Gilmartin wrote: >Um. No. Not unless the top halves and bottom halves are identical.
I said it wrong. What I meant is, if you expect a halfword and I pass a fullword, you’ll see zero for values up to 65K. Then you’ll start seeing values mod(65K). So you’ll *notice* right away. Similarly, if you expect a fullword and I pass a halfword, you’ll also notice right away, ‘cause you’ll see values+65K. Either way, basic testing will reveal the problem. With little-endian, as Charles noted, it’ll work fine up until 65K+1. So testing that doesn’t cover this case won’t notice the bug. Better? ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
