4:01 p.m.
PATCH 21.5
This patch (attached to avoid gmail corruption) solves the problem
reported here:
http://list-archive.xemacs.org/pipermail/xemacs-beta/2011-August/021780.html
Briefly, when using lisp-disunion.h, it is possible to lose bits when
shifting, so we need to make sure the value to be shifted has enough
bits. I'm not sure the make_char_1 part is really necessary, but it
doesn't hurt anything and may prevent a problem. The lisp-union.h
definitions do not have this problem, since there is no explicit
shifting.
--
Jerry James
http://www.jamezone.org/
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://lists.xemacs.org/mailman/listinfo/xemacs-patches
5:54 p.m.
Ar an t-ochtú lá is fiche de mí na Nollaig, scríobh Jerry James:
PATCH 21.5
This patch (attached to avoid gmail corruption) solves the problem
reported here:
http://list-archive.xemacs.org/pipermail/xemacs-beta/2011-August/021780.html
Briefly, when using lisp-disunion.h, it is possible to lose bits when
shifting, so we need to make sure the value to be shifted has enough
bits. I'm not sure the make_char_1 part is really necessary, but it
doesn't hurt anything and may prevent a problem. The lisp-union.h
definitions do not have this problem, since there is no explicit
shifting.
This seems perfectly reasonable, and given that this behaviour is the same
as the union build has always given, we have some assurance it doesn’t
introduce any new bugs.
--
‘Iodine deficiency was endemic in parts of the UK until, through what has been
described as “an unplanned and accidental public health triumph”, iodine was
added to cattle feed to improve milk production in the 1930s.’
(EN Pearce, Lancet, June 2011)
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://lists.xemacs.org/mailman/listinfo/xemacs-patches
5:44 a.m.
Jerry James writes:
Briefly, when using lisp-disunion.h, it is possible to lose bits
when
shifting, so we need to make sure the value to be shifted has enough
bits.
Does this ever actually happen? (Just curious.) I thought in C the
RHS of the shift operator will be an int, so the LHS will get promoted
to int if it was char or short. When do we make_fixnum on anything big
enough to overflow on a one bit shift? That sounds like a bug in
itself to me.
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://lists.xemacs.org/mailman/listinfo/xemacs-patches
7:49 p.m.
On Wed, Dec 28, 2011 at 10:44 PM, Stephen J. Turnbull
<stephen(a)xemacs.org> wrote:
Does this ever actually happen? (Just curious.) I thought in C the
RHS of the shift operator will be an int, so the LHS will get promoted
to int if it was char or short. When do we make_fixnum on anything big
enough to overflow on a one bit shift? That sounds like a bug in
itself to me.
Yes, it actually happened. See the link I posted with the patch:
http://list-archive.xemacs.org/pipermail/xemacs-beta/2011-August/021780.html
That broke the XFT build on 64-bit platforms. I guess my
characterization of the bug was a bit off (get it?). The real problem
is a one bit being shifted into the sign bit position, thereby turning
a positive number into a negative number.
--
Jerry James, who is now going to eat lunch for a bit
http://www.jamezone.org/
_______________________________________________
XEmacs-Patches mailing list
XEmacs-Patches(a)xemacs.org
http://lists.xemacs.org/mailman/listinfo/xemacs-patches
4754
days inactive
4755
days old
3 comments
3 participants
participants (3)
-
Aidan Kehoe -
Jerry James -
Stephen J. Turnbull