Comment 10 for bug 848237

I might clarify why this needs an X Server patch in order to function correctly:

Basically we used the shape extension in order to remove the input
area from windows once they are "minimized" such that their pixmaps
will still be valid and display in the alt-tab. However, because of
the way reparenting works in compiz, compiz needs to know when the
input shape of the client has changed so that it can adjust the parent
window shape appropriately. But, we listen for shape events inside of
unity in order to ensure that applications that change their input
shape while minimized don't end up becoming "invisible windows" with
input. However, changing the input shape with events registered would
result in cycling because we'd be acting on our own requests to change
the input shape. But core needs to have the events to change the input
shape too!

What I decided to do in the unity plugin was to disable input shape
event reporting when we actually change the input and bounding shape
of the window and then send synthetic ShapeNotify events to the frame
windows (which compiz owns) such that core would update the shape
region on the frame window correctly and have the right idea of what
the input shape was. All events which are synethetic and generated by
clients doing XSendEvent always have the "send_event" member of the
XEvent structure forced on.

Unfortunately, this works a little differently in the wire protocol
for the X Server. "SendEvent" was probably bolted on to the protocol a
little bit after X11 was formalized, so in order to avoid breaking the
protocol version the developers made it so that any request that was
sent from a SendEvent request would have the high bit (0x80) set in
the event type mask. However the protocol is unclear as to where this
high bit should actually be set, as all it says is "forced on". As
such, there are some extension libraries that set the high bit before
converting the event to wire protocol format and some extension
libraries that rely on the server to do it before sending it to the
client. For the former case, this was broken because the server
doesn't unmask the high bit before doing range checks on the event
type, which meant that for some synethetic requests (eg Shape
requests) they would fall outside of the acceptable event range set by
the X Server at runtime. (So when sending synethetic ShapeNotify
events it will return an X Error saying that the event type is out of
range even when the event type is actually valid).

I've sent the patch to fix this upstream [1] and they're going to
merge it so I'm just waiting on RAOF to backport it so that this
method works correctly.

I initially tried hacking around this by forcing the window to be
resized when it was minimized but that doesn't work correctly with
some clients (like chromium), so this is the only way to do it.

You can tell if your server is broken by running the
"test-input-remover" test in unity/tests/test-input-remover, if it
returns an X Error saying that the event type "0xcc" was a BadValue
then your server is definitely broken :) (The server should have
processed the event type 0x4c)

[1] http://lists.x.org/archives/xorg-devel/2011-September/025017.html

On Sun, Sep 18, 2011 at 11:53 AM, Sam Spilsbury <email address hidden> wrote:
> Ok, no problem, I've asked RAOF to merge the X Server patch I sent
> upstream but he's been at XDC this past week so that will probably
> happen monday.
>
> On Sat, Sep 17, 2011 at 11:05 PM, Andrea Azzarone
> <email address hidden> wrote:
>> I can confirm this bug too. It's quite annoying.
>>
>> --
>> You received this bug notification because you are a member of Unity
>> Bugs, which is subscribed to unity in Ubuntu.
>> https://bugs.launchpad.net/bugs/848237
>>
>> Title:
>>  minimizing application creates an invisible window
>>
>> Status in “unity” package in Ubuntu:
>>  Confirmed
>> Status in “unity” source package in Oneiric:
>>  Confirmed
>>
>> Bug description:
>>  If I open some applications, then I minimize one it leaves behind an
>>  invisible window until I restore the window.
>>
>>  Steps to reproduce:
>>  1. open firefox, evolution and rhythmbox
>>  2. make sure they are overlapping in some manner
>>  3. minimize rhythmbox
>>  4. try to click in evolution or firefox where rhythmbox used to be. Focus is not changed and mouse clicks do not go to the window you are clicking on
>>
>>  If I click the rhythmbox icon in the Launcher (ie, unminimize it),
>>  then mouse clicks, etc operate as expected. This is not rhythmbox
>>  specific-- eg if I minimize evolution the same things happens.
>>
>>  I did enable the copytex plugin today as mentioned in bug #771521
>>  since it is now supposed to be the default. I am up to date as of a
>>  few minutes ago.
>>
>>  WORKAROUND: don't minimize anything
>>
>>  ProblemType: Bug
>>  DistroRelease: Ubuntu 11.10
>>  Package: unity 4.14.2-0ubuntu2
>>  ProcVersionSignature: Ubuntu 3.0.0-11.17-generic 3.0.4
>>  Uname: Linux 3.0.0-11-generic x86_64
>>  .tmp.unity.support.test.0:
>>
>>  ApportVersion: 1.22.1-0ubuntu2
>>  Architecture: amd64
>>  CompizPlugins: [core,bailer,detection,composite,opengl,compiztoolbox,decor,snap,gnomecompat,grid,place,copytex,session,mousepoll,vpswitch,regex,move,imgpng,resize,wall,animation,unitymtgrabhandles,resizeinfo,workarounds,expo,ezoom,staticswitcher,fade,scale,unityshell]
>>  CompositorRunning: compiz
>>  Date: Mon Sep 12 15:03:32 2011
>>  DistUpgraded: Log time: 2011-07-22 19:38:51.178194
>>  DistroCodename: oneiric
>>  DistroVariant: ubuntu
>>  GraphicsCard:
>>   Intel Corporation Core Processor Integrated Graphics Controller [8086:0046] (rev 02) (prog-if 00 [VGA controller])
>>     Subsystem: Lenovo Device [17aa:215a]
>>  InstallationMedia: Ubuntu 11.04 "Natty Narwhal" - Release amd64 (20110425.2)
>>  MachineType: LENOVO 5129CTO
>>  ProcKernelCmdLine: BOOT_IMAGE=/boot/vmlinuz-3.0.0-11-generic root=UUID=88689ad2-74b5-442c-9290-8ae60b63e16b ro quiet splash vt.handoff=7
>>  SourcePackage: unity
>>  UnitySupportTest: Error: command ['/usr/lib/nux/unity_support_test', '-p', '-f'] failed with exit code -11:
>>  UpgradeStatus: Upgraded to oneiric on 2011-07-23 (51 days ago)
>>  dmi.bios.date: 02/24/2011
>>  dmi.bios.vendor: LENOVO
>>  dmi.bios.version: 6QET64WW (1.34 )
>>  dmi.board.name: 5129CTO
>>  dmi.board.vendor: LENOVO
>>  dmi.board.version: Not Available
>>  dmi.chassis.asset.tag: No Asset Information
>>  dmi.chassis.type: 10
>>  dmi.chassis.vendor: LENOVO
>>  dmi.chassis.version: Not Available
>>  dmi.modalias: dmi:bvnLENOVO:bvr6QET64WW(1.34):bd02/24/2011:svnLENOVO:pn5129CTO:pvrThinkPadX201s:rvnLENOVO:rn5129CTO:rvrNotAvailable:cvnLENOVO:ct10:cvrNotAvailable:
>>  dmi.product.name: 5129CTO
>>  dmi.product.version: ThinkPad X201s
>>  dmi.sys.vendor: LENOVO
>>  version.compiz: compiz 1:0.9.5.94+bzr2803-0ubuntu1
>>  version.ia32-libs: ia32-libs N/A
>>  version.libdrm2: libdrm2 2.4.26-1ubuntu1
>>  version.libgl1-mesa-dri: libgl1-mesa-dri 7.11-0ubuntu3
>>  version.libgl1-mesa-dri-experimental: libgl1-mesa-dri-experimental N/A
>>  version.libgl1-mesa-glx: libgl1-mesa-glx 7.11-0ubuntu3
>>  version.xserver-xorg: xserver-xorg 1:7.6+7ubuntu7
>>  version.xserver-xorg-input-evdev: xserver-xorg-input-evdev 1:2.6.0-1ubuntu13
>>  version.xserver-xorg-video-ati: xserver-xorg-video-ati 1:6.14.99~git20110811.g93fc084-0ubuntu1
>>  version.xserver-xorg-video-intel: xserver-xorg-video-intel 2:2.15.901-1ubuntu2
>>  version.xserver-xorg-video-nouveau: xserver-xorg-video-nouveau 1:0.0.16+git20110411+8378443-1
>>
>> To manage notifications about this bug go to:
>> https://bugs.launchpad.net/ubuntu/+source/unity/+bug/848237/+subscriptions
>>
>
>
>
> --
> Sam Spilsbury
>

--
Sam Spilsbury