[PATCH 1/2] OS support: fix writeable client vs IgnoreClient behavior

[Jesse Barnes]

On Mon, 28 Jun 2010 18:48:51 -0700
Keith Packard <keithp at keithp.com> wrote:

> On Mon, 28 Jun 2010 17:43:28 -0700, Jesse Barnes <jbarnes at virtuousgeek.org> wrote:
> 
> > This is really a question of which behavior we want to preserve:
> >   1) clearing of IgnoredClientsWithInput across multiple IgnoreClient
> >   calls
> > or
> >   2) allowing multiple IgnoreClient and AttendClient calls to have any
> >   effect at all except when first ignoring and finally attending
> 
> I think that 2) is what we want -- permitting nesting of ignore/attend
> calls with the client blocked until all of the reasons for suspending
> the client are cleared.
> 
> Can anyone think of a reason to want the other behaviour in
> non-cooperating portions of the server?
> 
> > Here's what the counter patch looks like.  It needs to be matched with
> > a change to my DRI2 patch to unconditionally call AttendClient so the
> > counts match up.
> 
> This looks good to me. I assume the client is allocated with calloc so
> that the count starts out at zero?

Initially I just checked the alloc, and no it doesn't use calloc.  But
it does memset it to 0 in InitClient, so my initialization to 0 is
unnecessary.

> Reviewed-by: Keith Packard <keithp at keithp.com>
 
Thanks.  Do you want to apply it with that change or shall I send an
update?

Thanks,
-- 
Jesse Barnes, Intel Open Source Technology Center