Re: [AD] minor fix

[Peter Wang <tjaden@xxxxxxxxxx>]

Elias Pschernig wrote:

> > > poll_joy() should read all events
[...]

> > I'm personally undecided on wether this is a good thing or not. Any 
> > thoughts?
> >    
>
> Well, Eric said this:
>
>  
>
> > > http://www.allegro.cc/forums/view_thread.php?_id=354745&;
> > >      
> > I suppose this is patch #943753 on SF's tracker.  I've attached it.
> >
> > Peter, do you have an opinion on the patch?  I remember that you > didn't like 
> > a patch that would have made the same change (read all events) to the > Xlib 
> > event loop.
> >    
>
> So I guess, nobody really knows how bad it is.
>  

I don't think the linux joystick case has much relation to the X events 
case.

I problem I have with the patch was raised by Evert: it can cause you to 
lose events.  Whether to apply the patch or not depends on how often you 
expect the joystick to be under-polled.  If it's not very often, I'd 
rather do it as it is currently done and not lose any events.

> > while (read(joy_fd[i], &e, sizeof(struct js_event)) == sizeof(struct js_event))
> >    
>
> The above means, if the joystick sends too much data - it will just hang
> the Allegro program. OTOH, if the program is too slow to receive all
> joystick input - there's a problem anyway. I don't have a joystick
> myself to test it, and at least someone finds the new behavior better -
> so I'd vote for applying it, and then see if linux programs start
> behaving bad with joystick enabled.
>  

I doubt you have to worry about that.  Amongst other reasons, joysticks 
don't generate that much data.  There is a context switch on every call 
to read() though -- you might want to try asking for more than one event 
at a time.

Peter