FORMAT spec

[Sriram Thaiyar <sri@asu.edu>]

The expression (format "%.2 " -1.0) segfaults GNU Emacs 21.3.1 (under
sparc-sun-solaris2.8) and GNU Emacs 21.2.1 (under GNU/Linux).

This seems to occur because the when the argument to format is a float,
there is no check to see if the format char is a valid character.
Doing the following seems to work (in Fformat in editfns.c):

	else if (FLOATP (args[n]) && *format != 's')
	  {
            /** CHECK:  Adding the following seems to work. */ 
            if (*format != 'd' && *format != 'o' && *format != 'x'
                && *format != 'i' && *format != 'X' && *format != 'c'
                && *format != 'e' && *format != 'f' && *format != 'g')
	     error ("Invalid format operation %%%c", *format);

            
	    if (! (*format == 'e' || *format == 'f' || *format == 'g'))
	      args[n] = Ftruncate (args[n], Qnil);

	    /* Note that we're using sprintf to print floats,
	       so we have to take into account what that function
	       prints.  */
	    thissize = MAX_10_EXP + 100 + precision;
	  }

I simply cut and paste the check from the INTEGERP line, and added the
characters 'e', 'f', and 'g'.  I haven't checked to see if the above patch
conflicts with anything else.  Hope this helps.

-sri