Re: [vile] xvile 9.7y core dumps on *.rc files

[Thomas Dickey]

On Thu, 31 Dec 2009, Gary Jennejohn wrote:

> On Thu, 31 Dec 2009 06:25:48 -0500 (EST)
> Thomas Dickey <address@hidden> wrote:
>
> > On Thu, 31 Dec 2009, Gary Jennejohn wrote:
> >
> > > I just tried to look at /usr/local/share/vile/vileinit.rc with xvile
> > > and got a core dump.  I've only seen this with *.rc files so far.
> > > C, shell scripts and mail are OK.
> > >
> > > If I comment out "set cs light" in my .vilerc then it works.
> > >
> > > However, vile works just fine with "set cs light".
> > >
> > > Is there a way to test whether vile or xvile is running so I can "set
> > > cs light" only for vile?  xvile actually works OK without that setting.
> > >
> > > Here a bt from gdb:
> > > (gdb) bt
> > > #0  0x000000080161f41c in kill () from /lib/libc.so.7
> > > #1  0x000000080161e21b in abort () from /lib/libc.so.7
> > > #2  0x000000000044942b in imdying ()
> > > #3  <signal handler called>
> > > #4  0x000000000041eb83 in get_color_gc ()
> >
> > I don't see the cause, but it could be an out-of-bounds index into the
> > color-arrays.  This would help in that case:
> >
> > diff -u -r1.354 x11.c
> > --- x11.c       2009/12/09 01:47:36     1.354
> > +++ x11.c       2009/12/31 11:22:21
> > @@ -2,7 +2,7 @@
> >    *     X11 support, Dave Lemke, 11/91
> >    *     X Toolkit support, Kevin Buettner, 2/94
> >    *
> > - * $Header: /usr/build/vile/vile/RCS/x11.c,v 1.354 2009/12/09 01:47:36 tom Exp$
> > + * $Header: /usr/build/vile/vile/RCS/x11.c,v 1.356 2009/12/31 11:22:21 tom Exp$
> >    *
> >    */
> >
> > @@ -2180,10 +2180,15 @@
> >   static GC
> >   get_color_gc(int n, Boolean normal)
> >   {
> > -    ColorGC *data = (normal
> > -                    ? &(cur_win->fore_color[n])
> > -                    : &(cur_win->back_color[n]));
> > +    ColorGC *data;
> >
> > +    assert(n >= 0 && n < NCOLORS);
> > +
> > +    if (n < 0 || n > NCOLORS)
> > +       n = 0;                  /* shouldn't happen */
> > +    data = (normal
> > +           ? &(cur_win->fore_color[n])
> > +           : &(cur_win->back_color[n]));
> >       if (cur_win->screen_depth == 1) {
> >          data->gc = (normal
> >                      ? cur_win->textgc
>
> Yup, this patch fixes it.

thanks.  I put in the assert so that I might stumble over the actual cause 
when I'm debugging.  (The assert's are turned on when tracing is 
compiled-in).

--
Thomas E. Dickey
http://invisible-island.net
ftp://invisible-island.net