Path: utzoo!attcan!uunet!cs.utexas.edu!tut.cis.ohio-state.edu!ucbvax!ANDREW.CMU.EDU!wjh+
From: wjh+@ANDREW.CMU.EDU (Fred Hansen)
Newsgroups: comp.soft-sys.andrew
Subject: (none)
Message-ID: <kZNlLw600VsPIJ5W1t@andrew.cmu.edu>
Date: 19 Nov 89 21:07:08 GMT
Sender: daemon@ucbvax.BERKELEY.EDU
Organization: The Internet
Lines: 28


{Susan,  I think this clears Bug beta-111.}

Thanks everyone who sent listings of malloc entry points.  I now have an
idea why certain things work and other do not.  I've made two changes
that should fix the problems.

a)  Moved the definitions of realloc_srchlen and morecore from
andrew/overhead/malloc/plumber.ci to andrew/overhead/malloc/malloc.ci. 
This will mean that whenever the ITC malloc is used, these symbols will
be defined.  In the old way, they would not get defined if
MALLOC_DEBUG_ENV was not set.

b) Revised the order of linking libraries in andrew/atk/apps.  Now the
X11 library is last;  after all, it should not depend on any CMU stuff. 
If ANDREW_MALLOC_ENV is set, the Andrew malloc will be loaded from the
preceding library and will supercede any malloc that may be defined in
Xlib.


Another set of changes dealt with making it possible to allocate memory
on double word boundaries for SPARCs et al.  This change was released
first in a bboard post and later in a patch.  I have had no negative
reports, so can only assume that it is satisfactory.

If anyone has any further problems, please let me know.

FredHansen