On 04/08/2014 04:20 AM, Nicolas Richard wrote:
> Eli Zaretskii <eliz@gnu.org> writes:
> 
>>> From: Nicolas Richard <theonewiththeevillook@yahoo.fr>
>>> Date: Fri, 04 Apr 2014 17:26:36 +0200
>>>
>>> So I tried this:
>>> $ gdb --args emacs --batch  -f batch-byte-compile ../lisp/emacs-lisp/tcover-ses.el
>>>
>>> and here's what happens:
>>>
>>> Breakpoint 1, terminate_due_to_signal (sig=6, backtrace_limit=40) at emacs.c:355
>>> 355	  signal (sig, SIG_DFL);
>>>
>>> it aborts because of this in data.c:
>>>   /* Convert to eassert or remove after GC bug is found.  In the
>>>      meantime, check unconditionally, at a slight perf hit.  */
>>>   if (valid_lisp_object_p (definition) < 1)
>>>     emacs_abort ();
>>>
>>> Here's the backtrace:

Interesting. It works for me --- 64- and 32-bit builds. What happens if
you add that Ffset check to emacs-24? What symbol are we trying to set?
(Use xsymbol.) What does valid_lisp_object_p (definition) actually
return when you call it? Can you share your binary and core dump?