I cannot build 21.0-b40 on FreeBSD 2.2.6 RELEASE.
abbreviated log is like this.
% uname -m -s -r
FreeBSD 2.2.6-RELEASE i386
% make
[......]
EMACSBOOTSTRAPLOADPATH="/usr3/src/xemacs-21.0-b40/src/../lisp/:/usr3/src/xemacs-21.0-b40"
./temacs -batch -l /usr3/src/xemacs-21.0-b40/src/../lisp/update-elc.el
Loading /usr3/src/xemacs-21.0-b40/src/../lisp/update-elc.el...
[......]
Loading cl.el...
Loading cl-extra.el...Floating point exception - core dumped
*** Error code 136
Stop.
*** Error code 1
Stop.
This seems to be the same problem as of
<199805120211.LAA11075(a)jjctsn10.pnc.go.jp> '[Failure] b40-pre1 on
Alpha/Linux'.
In cl-float-limits,
1998-05-10 Hrvoje Niksic <hniksic(a)srce.hr>
* cl-extra.el (cl-float-limits): Removed; move code to top-level.
The code in ex. cl-float-limits causes 'Floating point exception' in
some systems. This exception is normally caught by arith_error (in data.c).
But when temacs executes update-elc.el in batch-mode, SIGFPE is not caught.
See the following function in src/data.c.
void
init_data_very_early (void)
{
/* Don't do this if just dumping out.
We don't want to call `signal' in this case
so that we don't have trouble with dumping
signal-delivering routines in an inconsistent state. */
#ifndef CANNOT_DUMP
if (!initialized)
return;
#endif /* CANNOT_DUMP */
signal (SIGFPE, arith_error);
#ifdef uts
signal (SIGEMT, arith_error);
#endif /* uts */
}
--
SAKIYAMA Nobuo nobuo(a)db3.so-net.or.jp