[PULL] xfree86: Config fixes

[Dan Nicholson]

On Mon, Feb 22, 2010 at 12:50 PM, Keith Packard <keithp at keithp.com> wrote:
> On Wed, 17 Feb 2010 17:38:11 -0800, Dan Nicholson <dbn.lists at gmail.com> wrote:
>
>>   Thomas Jaeger (1):
>>         udev: Don't filter subsystem "input"
>
> This one was merged already

That was also the "since" commit. :)

>> Dan Nicholson (2):
>>       xfree86: Fix priority ordering for ignoring input classes
>
> This one seems fine to me (at least, I think it does what the log
> message says it does).
>
> Reviewed-by: Keith Packard <keithp at keithp.com>
>
>>       config: Don't log error for ignored devices
>
> That's an ugly patch. Can you stick the error message down inside
> NewInputDeviceRequest instead of "knowing" what BadIDChoice has a magic
> meaning in several functions?

Yeah, it's not pretty. The current situation is that NIDR in xfree86
prints messages for most errors, but not all (typically allocation
failure). In the hal and dbus backends (but not udev), a message is
printed when returning !Success from NIDR.

1. If you remove the message from the config backend, adding the
device could fail silently for some errors in NIDR. That seems bad.

2. You have a !Success return that's not really an error and shouldn't
have an error message, but needs to signal back to clean up. This is
what I was trying to address.

So, it sounds like the right thing to do is pick through the NIDR
implementations, make sure they log fatal errors and remove the
messages from the config backends. Seem reasonable?

--
Dan