Re: pbbuttonsd not loading every dev event

[Kristian B. <kb...@op...>]

On Sat, 2006-03-18 at 12:51 +0100, Matthias Grimm wrote: 
> On Sat, 18 Mar 2006 05:34:27 -0500
> Kristian Benoit <kb...@op...> wrote:
> 
> > I had this problem a few times, pbbuttonsd does not recognize the eject
> > button. Plus the appletouch (absolute coordinate) events does not reset
> > the sleep timeout so I must press a key.
> > 
> > Some devices have the same vendor/product id. For example, the keyboard
> > (/dev/input/event3) has the same vendor/product as the eject key
> > (/dev/input/event4).
> > 
> > The appletouch also have the same ids as another event device too with a
> > lower number. So those where ignored by pbbuttonsd.
> 
> Oh no. :-( You can't rely on nothing nowadays. Why do we define
> standards when nobody follows them anyway? :-/
>  
> > I quikfixed my version, but I still have a strange behavior:
> 
> Could you please tell me how you did this? We need a good workaround
> for this quirk.

As my appletouch is represented by the device /dev/input/event6 and the
other with the same number is lower, I tried openning the device in
reverse order, but then the fn button did not work and the screen was
not dimming anymore. So I hardcoded the device numbers I need:

 for (n = 0; n < EVDEVCOUNT; n++) {
        sprintf(filename, "/dev/input/event%d", n);
        if ((fd = open(filename, O_RDONLY)) >= 0) {
            ioctl(fd, EVIOCGID, id);
            venprod = (gpointer) (id[ID_VENDOR]<<16 | id[ID_PRODUCT]);
            if ((g_list_find_custom (base->inputsources, venprod, cbEventDevices)) == NULL || n == 3 || n == 4 || n == 6)
                addInputSource (fd, event_handler, venprod, TRUE);
            else
                close(fd);
        } else
            fcnt++;  /* count number of event devices that could not be opened */
    }


A real solution would be to configure the event devs we want to use in
the config file...

I digged a a bit more and made this patch. It add a sub MODULE INPUT in
inputmanager.c. Its purpose is to be able to configure the input file
used by the inputmanager. So a new configuration as been added:

EventDevices = "/dev/input/event2 /dev/input/event8"

The default is the old behavior but without checking if some are the
same vendor_id/product_id.

----------------------

I fixed the bug with the appletouch not reporting it's event. The
problem is with the synaptics driver doiing a EVIOCGRAB on the event
device. This prevent the event to be propagated to /dev/input/mice which
is what one usually wants. If you configure a generic usbmouse and a
touch pad, you dont want the mouse to move while you scroll with the
touchpad. As a side effect, no one else can get those events.

So the solution is to patch the synaptics driver to no more grab the
event device and configure the usbmouse to use /dev/input/mouse?.

There is still a little problem, if your mouse is not pluged in when you
boot, you wont be able to use it without restarting X. A solution is to
always have the mouse? node. A better solution is to create a new
mouse/event aggregator like /dev/input/mice and be able to tell each
mouse/event which aggregator must receive it's events, but this is a
longer task...

Kristian