Subject: too good to be true, more than one bug in Xsun[Mono]
To: NetBSD/sparc Discussion List <port-sparc@NetBSD.ORG>
From: Greg A. Woods <>
List: current-users
Date: 04/23/2002 16:47:35
Well I should have known it would be too good to be true.  I've had four
more crashes of my Xserver since "avoiding" the easily repeatable
problem with unicode encoded fonts.

Two of the core dumps, triggered by SIGBUS, look good on first glance
with just a backtrace, but further examination shows the problem was
dereferencing an array with '-1', which came from a local
(i.e. on-the-stack) integer that was supposedly incremented from 0 in a
for loop, and there are no other obvious manipulations of that index
variable anywhere else within the loop (though the code right there,
WaitFor.c:480, is an ugly mess of #ifdefs).

I'm going to try to build a really -current kernel and see if any of the
changes that have been made since my last merge will make any
difference.....  (I'm up to 1.193 of sparc/pmap.c, and in other files
the only interesting changes I see are "use pmap_k{enter_pa,remove}()
where appropriate.", and the supposed no-op for the new pmap_update()
updated-page parameter.)

								Greg A. Woods

+1 416 218-0098;  <>;  <>;  <>
Planix, Inc. <>; VE3TCP; Secrets of the Weird <>