On Friday 10 February 2006 19:31, Christos Zoulas wrote:
> On Feb 10,  6:41pm, jmmv84@gmail.com ("Julio M. Merino Vidal") wrote:
> -- Subject: Re: CVS commit: src/sys/dev/wscons
>
> | On Friday 10 February 2006 18:33, Christos Zoulas wrote:
> | > Module Name:	src
> | > Committed By:	christos
> | > Date:		Fri Feb 10 17:33:02 UTC 2006
> | >
> | > Modified Files:
> | > 	src/sys/dev/wscons: wsmouse.c
> | >
> | > Log Message:
> | > PR/32794: Paul Shupak: Panic in wsmouse code.
> | > Checking the number of events after you've trashed the stack is not
> | > very useful. Instead, break out of the loop if we ran out, printing a
> | > message.
> |
> | Maybe there should be assertions just before each attempt to modify
> | events[nevents] to ensure that nevents is low enough.
>
> I think it is fine for the non-loop cases (because it is obvious that
> they fit).
>
> | But... given the way the events vector is sized (to which you added a
> | comment), in which situation can it overflow?
>
> Well, we'll find out soon enough I guess. The logic is complicated with
> the repeat button mix.

But it doesn't add more events to the vector.  It only sets up a callout
which acts independently of the input function.  Note that the first
event is sent from within this function, so everything is the same as
was before (modulo the assertion).

> | > Also don't try to inject 0 events; reset our state instead. Maybe
> | > having
> |
> | Theorically the input function is called when there is, at least, an
> | event to process (this is stated in a comment at the beginning of the
> | function). Therefore, nevents == 0 cannot happen, isn't it?
>
> Then we should change it to print a diagnostic if DEBUG or DIAGNOSTIC is
> set.

I'm just trying to understand why the reporter of PR 32794 got those
assertion failures...

-- 
Julio M. Merino Vidal <jmmv84@gmail.com>
The Julipedia - http://julipedia.blogspot.com/