Re: Re: Stress testing KWin's screen handling

2014-11-27 Thread Martin Gräßlin
On Thursday 27 November 2014 12:46:30 Aleix Pol wrote:
> On Thu, Nov 27, 2014 at 9:03 AM, Martin Gräßlin  wrote:
> > Martin GräßlinOn Tuesday 25 November 2014 13:17:28  wrote:
> > > Hi all,
> > > I hope that it doesn't fail for others and we can get the changes in to
> > > improve the situation.
> > 
> > Given the overall positive feedback I just decided to push the change to
> > master. Please continue to test and best report bugs :-)
> 
> Thanks! I'm quite happy so far, it feels much better than KDE4's KWin.

wohoo - that's nothing I aimed for ;-) We should have kicked out 
QDesktopWidget a lng time ago if that's the result.

Cheers
Martin

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Stress testing KWin's screen handling

2014-11-27 Thread Lukáš Tinkl

Dne 27.11.2014 v 12:46 Aleix Pol napsal(a):

On Thu, Nov 27, 2014 at 9:03 AM, Martin Gräßlin mailto:mgraess...@kde.org>> wrote:

Martin GräßlinOn Tuesday 25 November 2014 13:17:28  wrote:
 > Hi all,
> I hope that it doesn't fail for others and we can get the changes in to
> improve the situation.

Given the overall positive feedback I just decided to push the change to
master. Please continue to test and best report bugs :-)


Thanks! I'm quite happy so far, it feels much better than KDE4's KWin.

Aleix



Indeed, thanks a lot! Together with dvratil's changes in kscreen, works 
like a charm :)

--
Lukáš Tinkl 
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Stress testing KWin's screen handling

2014-11-27 Thread Aleix Pol
On Thu, Nov 27, 2014 at 9:03 AM, Martin Gräßlin  wrote:

> Martin GräßlinOn Tuesday 25 November 2014 13:17:28  wrote:
> > Hi all,
> > I hope that it doesn't fail for others and we can get the changes in to
> > improve the situation.
>
> Given the overall positive feedback I just decided to push the change to
> master. Please continue to test and best report bugs :-)
>

Thanks! I'm quite happy so far, it feels much better than KDE4's KWin.

Aleix
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Stress testing KWin's screen handling

2014-11-27 Thread Martin Gräßlin
Martin GräßlinOn Tuesday 25 November 2014 13:17:28  wrote:
> Hi all,
> I hope that it doesn't fail for others and we can get the changes in to
> improve the situation.

Given the overall positive feedback I just decided to push the change to 
master. Please continue to test and best report bugs :-)

Cheers
Martin

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Stress testing KWin's screen handling

2014-11-26 Thread Luca Beltrame
In data martedì 25 novembre 2014 13:17:28, Martin Gräßlin ha scritto:

> I hope that it doesn't fail for others and we can get the changes in to
> improve the situation.

I tested this today with a projector: unplugging it from my laptop did not 
cause rendering to go haywire like it used to do. 

-- 
Luca Beltrame - KDE Forums team
KDE Science supporter
GPG key ID: 6E1A4E79

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Stress testing KWin's screen handling

2014-11-26 Thread Daniel Vrátil
On Wednesday 26 of November 2014 11:48:28 Martin Gräßlin wrote:
> On Wednesday 26 November 2014 11:39:18 Daniel Vrátil wrote:
> > > ok, that still sounds like a rendering error. A few questions:
> > > * does qdus.org.kde.KWin /KWin supportInformation report correct screen
> > > information?
> > 
> > Yes
> > 
> > > * does xrandr report correct screen information?
> > 
> > Yes
> > 
> > > * does restarting compositing fix it?
> > 
> > Yes
> 
> good, so we have the correct information just render in a wrong way. That
> significantly reduces the problem scope.
> 
> > > For three screens I'm completely out of testing possibilities. I don't
> > > have
> > > three screens and even if I had I would not be able to connect them.
> > 
> > I managed to reproduce this when I unplugged the 3rd screen, so I don't
> > think this is strictly related to having ridiculous amount of screens, but
> > should be reproducible with two screens too.
> > 
> > If there's any debug info I can provide, just let me know. I really want
> > this fixed, but KWin internals are beyond my understanding :-)
> 
> How was the layout of the screens and which one did you unplug (e.g. right
> most, middle or left most)?

There were three screens next to each other, I unplugged the right-most one. 

I now realize that I can reliably reproduce this by placing a maximized window 
(like KMail) on that screen. Once unplugged, the KMail window (the actual 
window with content) is moved one screen left, to the one that is now the 
right-most one, while KMail window decorations are moved all the way to the 
left-most screen.

> Are you using the OpenGL or XRender compositor
> (just asking as the xrender could still be wrong as I didn't adjust it)?

I'm using OpenGL 2.0 with GLX interface.

> 
> Cheers
> Martin

-- 
Daniel Vrátil | dvra...@redhat.com | dvratil on #kde-devel, #kontact, #akonadi
Software Engineer - KDE Desktop Team, Red Hat Inc.

GPG Key: 0xC59D614F6F4AE348
Fingerprint: 4EC1 86E3 C54E 0B39 5FDD B5FB C59D 614F 6F4A E348

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Re: Stress testing KWin's screen handling

2014-11-26 Thread Martin Gräßlin
On Wednesday 26 November 2014 11:39:18 Daniel Vrátil wrote:
> > ok, that still sounds like a rendering error. A few questions:
> > * does qdus.org.kde.KWin /KWin supportInformation report correct screen
> > information?
> 
> Yes
> 
> > * does xrandr report correct screen information?
> 
> Yes
> 
> > * does restarting compositing fix it?
> 
> Yes

good, so we have the correct information just render in a wrong way. That 
significantly reduces the problem scope.

> 
> > For three screens I'm completely out of testing possibilities. I don't
> > have
> > three screens and even if I had I would not be able to connect them.
> 
> I managed to reproduce this when I unplugged the 3rd screen, so I don't
> think this is strictly related to having ridiculous amount of screens, but
> should be reproducible with two screens too.
> 
> If there's any debug info I can provide, just let me know. I really want
> this fixed, but KWin internals are beyond my understanding :-)

How was the layout of the screens and which one did you unplug (e.g. right 
most, middle or left most)? Are you using the OpenGL or XRender compositor 
(just asking as the xrender could still be wrong as I didn't adjust it)?

Cheers
Martin

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Stress testing KWin's screen handling

2014-11-26 Thread Daniel Vrátil
On Wednesday 26 of November 2014 08:19:26 Martin Gräßlin wrote:
> On Tuesday 25 November 2014 16:28:16 Daniel Vrátil wrote:
> > On Tuesday 25 of November 2014 13:17:28 Martin Gräßlin wrote:
> > > Hi all,
> > > 
> > > I spent some time on screen management in KWin today and got it to the
> > > point where it doesn't fail any more no matter what I try. So please
> > > everyone using multiple screens and especially dynamically plug in and
> > > out, please give a try to the patch set in [1]. Please ensure to have
> > > latest master as it contains a crash fix for a crash triggered by the
> > > patch set.
> > > 
> > > Short summary of the changes in the patch set:
> > > * uses XRandR instead of QDesktopWidget
> > > * uses KWin internal information about overall screen geometry instead
> > > of
> > > relying on the information in the X11 screen structure.
> > > 
> > > The second part is the code I added today. My testing showed that
> > > unplugging a screen gives us proper XRandR events so KWin's internal is
> > > up to date, but the X11 screen information is wrong. So when we
> > > partially
> > > used the one and partially the other the rendering was just horribly
> > > broken. Now it's all based on the KWin internal information and I
> > > couldn't get the rendering broken any more.
> > > 
> > > When changing screens please be patient. It takes time to settle the
> > > changes. Especially plasmashell takes quite some time on my system to
> > > render correctly again.
> > 
> > Coincidentally, I just merged my KScreen redesign, which should make this
> > faster.
> 
> sounds like I need to trigger kdesrc-build ;-)
> 
> > > I hope that it doesn't fail for others and we can get the changes in to
> > > improve the situation.
> > 
> > So far it's much better than before, but still it sometimes happens, that
> > after screen reshuffle, window decorations get detached from the windows
> > and moved elsewhere. It just happened to me, after plugging in the 3rd
> > screen: http://pub.dvratil.cz/kwin-bug.ogv, but I'm not able to reliably
> > reproduce this.
> 
> ok, that still sounds like a rendering error. A few questions:
> * does qdus.org.kde.KWin /KWin supportInformation report correct screen
> information?

Yes

> * does xrandr report correct screen information?

Yes

> * does restarting compositing fix it?

Yes

> 
> For three screens I'm completely out of testing possibilities. I don't have
> three screens and even if I had I would not be able to connect them.

I managed to reproduce this when I unplugged the 3rd screen, so I don't think 
this is strictly related to having ridiculous amount of screens, but should be 
reproducible with two screens too.

If there's any debug info I can provide, just let me know. I really want this 
fixed, but KWin internals are beyond my understanding :-)

> A kingdom, a kingdom for unit testing xrandr.

Amen to that


Dan

> 
> Cheers
> Martin

-- 
Daniel Vrátil | dvra...@redhat.com | dvratil on #kde-devel, #kontact, #akonadi
Software Engineer - KDE Desktop Team, Red Hat Inc.

GPG Key: 0xC59D614F6F4AE348
Fingerprint: 4EC1 86E3 C54E 0B39 5FDD B5FB C59D 614F 6F4A E348

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Re: Stress testing KWin's screen handling

2014-11-25 Thread Martin Gräßlin
On Tuesday 25 November 2014 16:28:16 Daniel Vrátil wrote:
> On Tuesday 25 of November 2014 13:17:28 Martin Gräßlin wrote:
> > Hi all,
> > 
> > I spent some time on screen management in KWin today and got it to the
> > point where it doesn't fail any more no matter what I try. So please
> > everyone using multiple screens and especially dynamically plug in and
> > out, please give a try to the patch set in [1]. Please ensure to have
> > latest master as it contains a crash fix for a crash triggered by the
> > patch set.
> > 
> > Short summary of the changes in the patch set:
> > * uses XRandR instead of QDesktopWidget
> > * uses KWin internal information about overall screen geometry instead of
> > relying on the information in the X11 screen structure.
> > 
> > The second part is the code I added today. My testing showed that
> > unplugging a screen gives us proper XRandR events so KWin's internal is
> > up to date, but the X11 screen information is wrong. So when we partially
> > used the one and partially the other the rendering was just horribly
> > broken. Now it's all based on the KWin internal information and I
> > couldn't get the rendering broken any more.
> > 
> > When changing screens please be patient. It takes time to settle the
> > changes. Especially plasmashell takes quite some time on my system to
> > render correctly again.
> 
> Coincidentally, I just merged my KScreen redesign, which should make this
> faster.

sounds like I need to trigger kdesrc-build ;-)

> 
> > I hope that it doesn't fail for others and we can get the changes in to
> > improve the situation.
> 
> So far it's much better than before, but still it sometimes happens, that
> after screen reshuffle, window decorations get detached from the windows and
> moved elsewhere. It just happened to me, after plugging in the 3rd screen:
> http://pub.dvratil.cz/kwin-bug.ogv, but I'm not able to reliably reproduce
> this.

ok, that still sounds like a rendering error. A few questions:
* does qdus.org.kde.KWin /KWin supportInformation report correct screen 
information?
* does xrandr report correct screen information?
* does restarting compositing fix it?

For three screens I'm completely out of testing possibilities. I don't have 
three screens and even if I had I would not be able to connect them.

A kingdom, a kingdom for unit testing xrandr.

Cheers
Martin

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Stress testing KWin's screen handling

2014-11-25 Thread Daniel Vrátil
On Tuesday 25 of November 2014 13:17:28 Martin Gräßlin wrote:
> Hi all,
> 
> I spent some time on screen management in KWin today and got it to the point
> where it doesn't fail any more no matter what I try. So please everyone
> using multiple screens and especially dynamically plug in and out, please
> give a try to the patch set in [1]. Please ensure to have latest master as
> it contains a crash fix for a crash triggered by the patch set.
> 
> Short summary of the changes in the patch set:
> * uses XRandR instead of QDesktopWidget
> * uses KWin internal information about overall screen geometry instead of
> relying on the information in the X11 screen structure.
> 
> The second part is the code I added today. My testing showed that unplugging
> a screen gives us proper XRandR events so KWin's internal is up to date,
> but the X11 screen information is wrong. So when we partially used the one
> and partially the other the rendering was just horribly broken. Now it's
> all based on the KWin internal information and I couldn't get the rendering
> broken any more.
> 
> When changing screens please be patient. It takes time to settle the
> changes. Especially plasmashell takes quite some time on my system to
> render correctly again.

Coincidentally, I just merged my KScreen redesign, which should make this 
faster.

> 
> I hope that it doesn't fail for others and we can get the changes in to
> improve the situation.

So far it's much better than before, but still it sometimes happens, that 
after screen reshuffle, window decorations get detached from the windows and 
moved elsewhere. It just happened to me, after plugging in the 3rd screen:
http://pub.dvratil.cz/kwin-bug.ogv, but I'm not able to reliably reproduce 
this.

Dan


> 
> Cheers
> Martin
> 
> [1] https://git.reviewboard.kde.org/r/117614/

-- 
Daniel Vrátil | dvra...@redhat.com | dvratil on #kde-devel, #kontact, #akonadi
Software Engineer - KDE Desktop Team, Red Hat Inc.

GPG Key: 0xC59D614F6F4AE348
Fingerprint: 4EC1 86E3 C54E 0B39 5FDD B5FB C59D 614F 6F4A E348

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Re: Stress testing KWin's screen handling

2014-11-25 Thread Aleix Pol
On Tue, Nov 25, 2014 at 1:17 PM, Martin Gräßlin  wrote:

> Hi all,
>
> I spent some time on screen management in KWin today and got it to the
> point
> where it doesn't fail any more no matter what I try. So please everyone
> using
> multiple screens and especially dynamically plug in and out, please give a
> try
> to the patch set in [1]. Please ensure to have latest master as it
> contains a
> crash fix for a crash triggered by the patch set.
>
> Short summary of the changes in the patch set:
> * uses XRandR instead of QDesktopWidget
> * uses KWin internal information about overall screen geometry instead of
> relying on the information in the X11 screen structure.
>
> The second part is the code I added today. My testing showed that
> unplugging a
> screen gives us proper XRandR events so KWin's internal is up to date, but
> the
> X11 screen information is wrong. So when we partially used the one and
> partially the other the rendering was just horribly broken. Now it's all
> based
> on the KWin internal information and I couldn't get the rendering broken
> any
> more.
>
> When changing screens please be patient. It takes time to settle the
> changes.
> Especially plasmashell takes quite some time on my system to render
> correctly
> again.
>
> I hope that it doesn't fail for others and we can get the changes in to
> improve the situation.
>
> Cheers
> Martin
>
> [1] https://git.reviewboard.kde.org/r/117614/


Hi Martin,
I just applied your patch, seems to work so far. I'll tell you if it breaks
:D.

Aleix
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel


Stress testing KWin's screen handling

2014-11-25 Thread Martin Gräßlin
Hi all,

I spent some time on screen management in KWin today and got it to the point 
where it doesn't fail any more no matter what I try. So please everyone using 
multiple screens and especially dynamically plug in and out, please give a try 
to the patch set in [1]. Please ensure to have latest master as it contains a 
crash fix for a crash triggered by the patch set.

Short summary of the changes in the patch set:
* uses XRandR instead of QDesktopWidget
* uses KWin internal information about overall screen geometry instead of 
relying on the information in the X11 screen structure.

The second part is the code I added today. My testing showed that unplugging a 
screen gives us proper XRandR events so KWin's internal is up to date, but the 
X11 screen information is wrong. So when we partially used the one and 
partially the other the rendering was just horribly broken. Now it's all based 
on the KWin internal information and I couldn't get the rendering broken any 
more.

When changing screens please be patient. It takes time to settle the changes. 
Especially plasmashell takes quite some time on my system to render correctly 
again.

I hope that it doesn't fail for others and we can get the changes in to 
improve the situation.

Cheers
Martin

[1] https://git.reviewboard.kde.org/r/117614/

signature.asc
Description: This is a digitally signed message part.
___
Plasma-devel mailing list
Plasma-devel@kde.org
https://mail.kde.org/mailman/listinfo/plasma-devel