!standard  4.6(32)                                     04-04-23  AC95-00098/01
!class confirmation 04-04-23
!status received no action 04-04-23
!status received 04-04-22
!subject Unspecified behavior for numeric type conversions

!summary

!appendix

!topic unspecified behavior for numeric type conversions
!reference 4.6(32)
!from Dan Eilers 04-04-22
!discussion

RM 4.6(32) leaves unspecified whether truncation or rounding
(or some other mechanism) is used in certain numeric type conversions.
It would be helpful to use the word "unspecified", so that this
paragraph appears in the index under unspecified, as is done
for the similar paragraph 4.5.5(21), for example.

It might also be helpful to give implementation advice,
in the interest of portability.

****************************************************************

From: Pascal Leroy
Sent: Friday, April 23, 2004  4:07 AM

It is certainly not unspecified.  It is perfectly well specified when
you are in strict mode.  And there is nothing special wrt conversion
here: the semantics of addition have exactly the same property.

So I am objecting to the notion of adding "unspecified" to 4.6(32) or
anywhere in the core of the RM.  Maybe G.2(1) should have the word
"unspecified" in it when it describes the relaxed mode, but it seems
insufficiently broken to me.

****************************************************************