ISO/ IEC JTC1/SC22/WG14 N848

					Document: N848
						  J11/98-047
-------------------------------------------------------

J11 Issues re Disposition of U.S. Public Comments
  Rex Jaeschke, J11 Chair
  Sep 6, 1998

1) Fred Tydeman voted NO on the following issue:

In PC #7, comment 29, change 'unlimited' to 'large'.

-------------------------------------------------------

2) SDRC/Jones voted NO on the following issue:

The response to 7.48 is clearly incorrect (and I would argue that the
response to DR #81 to which it refers is also incorrect, but that's not
nearly so clear).  The draft explicitly states that left shifting a
nonnegative signed value such that the theoretical result does not fit in
the return type results in undefined behavior; it is *NOT*
implementation-defined as stated in the response.  I suggest the following
response instead:

As you note, C90 didn't mention the signed case at all, making it at least
unspecified behavior, if not implicitly undefined behavior.  The intent in
the current draft is to make the noncontroversial cases well- defined. The
reason for making the above undefined is that many existing
implementations use arithmetic shift instructions to implement signed
shifts, and it is not unusual for arithmetic shift instructions to cause
an exception if the result overflows.

-------------------------------------------------------

3) U.S. Army Research Lab/Gwyn votes YES with the following comment:

I am concerned at the absence of explanation for the response to PC #10,
Comment 14.  I suggest something like the following be added, without
objection:

 We believe the new extensions are compatible with the
 previous facilities and constitute a definite
 improvement, even if not to the extent you would like.
 This specification does not preclude other extensions,
 which could perhaps be specified by other standards.
 Neither alternative, removing the improvements nor
 delaying this revision of the standard until a more
 perfect set of extensions are devised, seemed
 preferable to proceeding with the current specification.

-------------------------------------------------------