cvs commit: src/sys/compat/ndis subr_ntoskrnl.c

[Maxim Sobolev]

On Fri, Jul 08, 2005 at 09:37:22PM -0600, Scott Long wrote:
> Maxim Sobolev wrote:
> >On Fri, Jul 08, 2005 at 03:24:34PM -0600, Scott Long wrote:
> >
> >>>I think the better fix is to change type of k_lock member from void* to 
> >>>the struct mtx*, I've already commited such change to the RELENG_5 few 
> >>>days ago.
> >>>
> >>>-Maxim
> >>
> >>It's a void* for portability with other OS's.  This work has already 
> >>been ported
> >>to Solaris, and maybe even other OS's by now.  Reducing the portability 
> >>is a
> >>bad idea, and will likely get you smacked around by Bill when he 
> >>notices.  I'd
> >>highly recommend not doing this again in the future.
> >
> >
> >Even if so, rik's version is no better since it adds explicit cast
> >(two of them in fact) to struct mtx*. Either way it will break
> >portability with other OSes. Therefore, they are on par, but my
> >solution is more clear wrt FreeBSD.
> >
> >-Maxim
> 
> subr_ntoskrnl.c explicitely uses the FreeBSD mtx API, so adding FreeBSD
> casts is fine.  Polluting the common headers, as you did, is not the
> same thing.

OK, you are right.

-Maxim