Re: Xwpe: Saving The World

[Dennis Payne <payned@rpi.edu>]

James Curbo had some problems with xwpe crashing back in December.
Well I just recently stumbled across some new info.

> > Could you try commenting out the line:
> >  sigaction(SIGSEGV, &act, NULL);
> > in we_unix.c, recompile, and crash it.  That should generate a core file.
> > 'gdb wpe core' will load the core dump.  Type 'where' which should
> > give you the stack and then send that to the xwpe mailing list.
> 
> I did so, but crashing it won't give me a core file. Ulimit's set ok and
> everything. (I usually have ulimit -c 0) It does do the same recursive
> thing though.. I commented out line 93 of we_unix.c, which is the one
> above. It still gives me:

It appears we_term.c sets the seg fault signal as well in e_begscr.
The offending line being:
 signal(SIGSEGV, e_exit);

(I decided to take a break from my computer game and spent last night
ripping apart xwpe.  I managed to split the curses/termcap and X11 functions
into separate shared libraries that way the curses version doesn't require
installing X.  It is extremely quick and dirty and requires dl* functions
which aren't available on all unixes so it still needs work.  No one has
ever complained to me about this but I always hated that RedHat shipped
a non-X11 xwpe package but it requires the X11 libraries to run.)

Dennis Payne
payned@rpi.edu