ANSIfication patches
Matthieu Herrb
matthieu.herrb at laas.fr
Fri Feb 18 13:33:35 PST 2005
Adam Jackson wrote:
> There are a few patches sitting in bugzilla regarding converting old ugly
> K&R-style code to ANSI:
>
> https://bugs.freedesktop.org/show_bug.cgi?id=2560
> https://bugs.freedesktop.org/show_bug.cgi?id=2566
>
> What are the issues to look out for here? Can we pretty much assume that
> non-API code is safe to convert and that the only dangerous areas would be
> (say) Xlib? And how does the implicit argument passing convention for K&R
> differ from ANSI? (Probably assumes 32-bit width everywhere, which really
> only matters for float/double...)
These should be treated carefully. Changing a K&R declaration of a less
than 32 bits integer parameter (or a float parameter) does change the
actual type that is passed, and the size of the parameter. Some subtile
breakage may happen. I generally prefer to change the parameter's type
in the declaration to the one used after the K&R promotion.
> I'd really like for gcc to never tell me "function declaration is not a
> prototype" ever again.
I've not looked at the dix diff in details. In the X server there are
lots of pointers to functions which are used to store references to
functions with variables prototypes. These are hard to ansify properly.
--
Matthieu
More information about the xorg
mailing list