No, I don't have an easy test case, though I suspect it would be very 
easy to trigger by adding one more DEFVAR_LISP() to the main source 
code of 21.4.12; since the dyamic loading of my .ell file triggered it 
on its first DEFVAR_LISP(). Things to look for: staticpros has a length 
of 1346 elements before the DEFVAR_LISP(), and this additional 
staticpro causes it to lengthen. You probably would have hit this in 
21.4.13 anyway. (actually now that I think about it some more, there 
must be a lot of defvar_lisp() equivalents in all the .el files i load 
in my init.el that would contribute to getting this array that high. 
hmph.)
Andy
On Thursday, January 30, 2003, at 06:25  PM, Stephen J. Turnbull wrote:
 RECOMMEND 21.4
 [reply-to set to xemacs-beta]
 I still don't have a test case (Andrew, do you?) but as Martin would
 say this is "obviously correct".  ben?
>>>>> "Andrew" == Andrew Begel <abegel(a)cs.berkeley.edu>
writes:
     Andrew> In XEmacs 21.4.12:
     Andrew> There is a type mismatch between bytes and # elements in
     Andrew> the function dynarr.c:Dynarr_realloc().  [....]  Here's a
     Andrew> patch that fixes it:
 
-----------
Andrew Begel
Ph.D. Candidate
Computer Science Division
University of California, Berkeley