Re: [PATCH] Re: .macro behavior

[Hans-Peter Nilsson <hp@bitrange.com>]

On Tue, 8 Feb 2005, Jan Beulich wrote:
> >>> Ian Lance Taylor <ian@airs.com> 14.01.05 17:50:05 >>>
> >"Jan Beulich" <JBeulich@novell.com> writes:
> >
> >> Is it intentional that .macro
> >>
> >> - ignores the (configurable) set of symbol name characters and
> instead
> >> only allows [[:alpha:]_$][[:alnum:]_$]*

I think this restriction is actually good.  Calls for the least
amount of surprises when applying the same macro to different
GAS ports.

> 	* gas/mmix/relax2.s: Use .altmacro and & to separate macro
> parameter
> 	references.

Ew.  Please just get rid of the ":"s instead (my mistake, as
it's a symbol character, not the usual label-ending thingy).

And a nitpick:

> /home/jbeulich/src/binutils/mainline/2005-02-01/gas/testsuite/gas/mmix/relax2.s	2003-10-18
> 18:00:21.000000000 +0200
> +++ 2005-02-01/gas/testsuite/gas/mmix/relax2.s	2005-02-08
> 09:38:12.000000000 +0100
> @@ -1,35 +1,37 @@
>  # PUSHJ stub border-cases: two with either or both stubs unreachable,
>  # local symbols, ditto non-local labels, similar with three PUSHJs.
>
> +.altmacro
> +

Pseudos shouldn't be first on a line. Accepting it in gas is ok,
but we shouldn't get sloppy in the test-suite.  For consistency,
it should also be aligned with the other pseudos in that file.
(I guess you don't need the ".altmacro" pseudo if you just
remove the ":"s so this is mostly for the record.)

brgds, H-P