[patch, fortran, committed] Document COMPLEX*k syntax quirk.

[Brooks Moses]

The attached patch documents the fact that "COMPLEX*k x" declares x to 
be of kind k/2, not of kind k.  The only rational way to explain this is 
to say that in a "TYPESPEC*k" declaration, the "k" is a byte size rather 
than a kind number -- which is, after all, the usual interpretation of 
that syntax anyway.

I also added a comment to the relevant function in decl.c, noting that 
it assumes a correspondence between byte sizes and kind numbers.

-------------------------------------------------------------------
2007-02-22  Brooks Moses  <brooks.moses@codesourcery.com>

	* gfortran.texi (Old-style kind specifications): Document
	special handling of old-style kind specifiers for COMPLEX.
	* decl.c (gfc_match_old_kind_spec): Documented kind/bytesize
	assumptions in comment.

-------------------------------------------------------------------

The documentation changes were tested with "make pdf" and "make info" on 
i686-pc-linux-gnu.  Committed to trunk and 4.2, under a blanket approval 
from Steve Kargl on the texi changes, and as obvious for the decl.c comment.

- Brooks