[PATCH 0/6] xserver: Memory leaks and input hotplug updates

Magnus Vigerlöf Magnus.Vigerlof at home.se
Wed Apr 11 15:15:56 PDT 2007


On Tuesday 10 April 2007 23:00, Daniel Stone wrote:
> On Wed, Mar 28, 2007 at 11:29:48PM +0200, Magnus Vigerlöf wrote:
> > Here are some changes/additions I've found needed for hotplugging
> > (both adding and removing) input devices successfully without leaking
> > resources.
>
> Hi Magnus,
> Thanks a lot!  I've pushed all the patches here except for #5, because
> that's an API change.

Excellent!

Yes, it's an API change, but from what I understand it's only used internally 
in the xserver-repository (ok, evdev have its own proto of it, don't know how 
adding a return value will affect things.. Is it in a bad way?). I only found 
these three hits in all git-repos that Xorg consist of (according to 
source-navigator):

EvdevNew: +336 driver/xf86-input-evdev/src/evdev.c
InitInput: +1023 xserver/hw/xfree86/common/xf86Init.c
NewInputDeviceRequest: +408 xserver/hw/xfree86/common/xf86Xinput.c

> > Btw, is the patch added to Bug#:10324 (x&y axis not scaled when reporting
> > extension motion events) in the pipe to be applied, or are there changes
> > needed before it can get that far?
>
> Well, that breaks ABI in that it changes DeviceIntRec, but I'm looking
> at it now.  Will probably push it into a branch along with #5 and work
> from there.

Yes, it will break the ABI, but in a 'nice' way.. As long as the two new 
fields are not accessed by the drivers it doesn't matter against what it was 
compiled. The memory is allocated/freed by the server and not by the 
individual drivers. I was kind of hoping we could hook it up with XINPUT 1.0, 
but whether that's possible or not is not something I can tell...

But it's ok for me to push it into a separate branch for the time being.

Cheers
  Magnus V



More information about the xorg mailing list