Merge lp:~smspillaz/compiz-core/compiz-core.work_923683 into lp:compiz-core

Proposed by Sam Spilsbury
Status: Superseded
Proposed branch: lp:~smspillaz/compiz-core/compiz-core.work_923683
Merge into: lp:compiz-core
Diff against target: 1420 lines (+269/-571)
10 files modified
include/core/window.h (+2/-2)
plugins/composite/src/window.cpp (+16/-16)
plugins/decor/src/decor.cpp (+15/-15)
plugins/move/src/move.cpp (+12/-12)
plugins/opengl/src/window.cpp (+2/-2)
src/event.cpp (+1/-1)
src/privatewindow.h (+7/-8)
src/screen.cpp (+2/-0)
src/window.cpp (+208/-511)
src/windowgeometry.cpp (+4/-4)
To merge this branch: bzr merge lp:~smspillaz/compiz-core/compiz-core.work_923683
Reviewer Review Type Date Requested Status
Daniel van Vugt Needs Resubmitting
Sam Spilsbury Pending
Alan Griffiths Pending
Review via email: mp+100729@code.launchpad.net

This proposal supersedes a proposal from 2012-04-03.

This proposal has been superseded by a proposal from 2012-04-04.

Description of the change

Resubmitted now that https://code.launchpad.net/~smspillaz/compiz-core/compiz-core.fix_969108.2/+merge/100441 and https://code.launchpad.net/~smspillaz/compiz-core/compiz-core.fix_969101/+merge/100270 are active

Note: you'll get funny behaviour around window edges unless you use lp:~smspillaz/compiz-snap-plugin/compiz-snap-plugin.work_923683 . The snap plugin was using the interfaces incorrectly.

Always use the asynchronous codepaths in core. This commit changes the following:

 * CompWindow::move is now just a wrapper around CompWindow::configureXWindow
 * CompWindow::configureXWindow will always call through to moveNotify
 * moveNotify is only ever called pre-request to the server in the case of
   managed windows (SubstructureRedirectMask guaruntees they will end up in
   the right place) and post ConfigureNotify for override redirect windows
 * resizeNotify is always called post ConfigureNotify regardless of whether
   or not the window is managed or not
 * composite and opengl now use the geometry last sent to the server in order
   to compute display matrices and window positions as well as damage regions
 * decor now also does the above
 * const correctness in include/core/window.h
 * removed priv->width and priv->height , which was just redundant data used for
   window shading and honestly, is a ticking time bomb for future maintainers.
 * CompWindow::syncPosition is now deprecated and a no-op
 * Removed the majority of PrivateWindow::updateFrameWindow - this function had a lot
   of redundant code which PrivateWindow::reconfigureXWindow had anyways - the big work
   there was to send synthetic ConfigureNotify events to ourselves to notify about windows
   moving within frame windows that don't move themselves. Its safer to use reconfigureXWindow
   in this case and probably more sane too. We should look into removing the other update* functions.

Caveats:

 0) There are no tests yet
 1) Window movement performance will regress with this branch. This is *probably* due to the fact that for every slight movement, we send a full ConfigureWindow request to the X Server and it has to post back to us a full ConfigureNotify, for pretty much every pixel the window moves. In the case that we're not doing compositing thats fine, but in every other case its *super* chatty and largely unnecessary. This can probably be optimized a lot, but that will require work to gather up requests that we want to send to the X server - combine them (eg, merging the stack of requests based on the change mask and using the latest values where appropriate). Thats a lot of work and probably not appropriate in this branch
 2) The diff is a little big
 3) No tests
 4) Window shading is more than likely going to break again, mostly because I haven't tested it yet, but also because shading does some strange things and manipulated geometry values in strange ways
 5) No tests

Testing plan?

I would imagine here that the PendingConfigureRequests object can probably be split out into something testable, as can the logic to strip flags from frame window geometry and client window geometry updates.

To post a comment you must log in.
Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal
Download full text (10.3 KiB)

The diff is big and scary, but its not too complicated. I'll explain each separate part

1 === modified file 'include/core/window.h'
2 --- include/core/window.h 2012-01-20 15:20:44 +0000
3 +++ include/core/window.h 2012-01-31 17:12:59 +0000
4 @@ -398,9 +398,9 @@
5
6 bool hasUnmapReference ();
7
8 - bool resize (XWindowAttributes);
9 + bool resize (const XWindowAttributes &);
10
11 - bool resize (Geometry);
12 + bool resize (const Geometry &);
13
14 bool resize (int x, int y, int width, int height,
15 int border = 0);

const correctness

31 int x1, x2, y1, y2;
32
33 - CompWindow::Geometry geom = priv->window->geometry ();
34 - CompWindowExtents output = priv->window->output ();
35 + const CompWindow::Geometry &geom = priv->window->serverGeometry ();
36 + const CompWindowExtents &output = priv->window->output ();

more const correctness

35 + const CompWindow::Geometry &geom = priv->window->serverGeometry ();
36 + const CompWindowExtents &output = priv->window->output ();

Lots of this - in paint code we want to use serverFoo because it was the last thing sent to the server. And this is always guarunteed to come back to us.

166 - unsigned int width = window->size ().width ();
167 - unsigned int height = window->size ().height ();
168 + unsigned int width = window->geometry ().width ();
169 + unsigned int height = window->geometry ().height ();

CompWindow::size was just returning CompSize (priv->width, priv->height) which was just redundant data which should be removed. Replaced it with geometry () for clarity

299 + void move (int dx, int dy, bool sync);
300 + bool resize (int dx, int dy, int dwidth, int dheight, int dborder);
301 + bool resize (const CompWindow::Geometry &g);
302 + bool resize (const XWindowAttributes &attrib);
303 +

Added PrivateWindow::move and PrivateWindow::resize - to be called whenever a new position / size comes back from the server. (cut'n'paste code from CompWindow::move and CompWindow::resize

335 -
336 - gettimeofday (&lastConfigureRequest, NULL);
337 - /* Flush any changes made to serverFrameGeometry or serverGeometry to the server
338 - * since there is a race condition where geometries will go out-of-sync with
339 - * window movement */
340 -
341 - window->syncPosition ();
342 -
343 - if (serverInput.left || serverInput.right || serverInput.top || serverInput.bottom)
344 - {
345 - int bw = serverGeometry.border () * 2;
346 -
347 - xwc.x = serverGeometry.x () - serverInput.left;
348 - xwc.y = serverGeometry.y () - serverInput.top;
349 - xwc.width = serverGeometry.width () + serverInput.left + serverInput.right + bw;
350 - if (shaded)
351 - xwc.height = serverInput.top + serverInput.bottom + bw;
352 - else
353 - xwc.height = serverGeometry.height () + serverInput.top + serverInput.bottom + bw;
354 -
355 - if (shaded)
356 - height = serverInput.top + serverInput.bottom;
357 -
358 - if (serverFrameGeometry.x () == xwc.x)
359 - valueMask &= ~(CWX);
360 - else
361 - serverFrameGeometry.setX (xwc.x);
362 -
*snip*
616 - XMoveResizeWindow (screen->dpy (), id, 0, 0,
617 - serverGeometry.width (), serverGeometry.height ());
618 - window->sendConfigureNotify ();
619 - window->windowNotify (CompWindowNotifyFrameUpdate...

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

824 -
825 - if (!pendingConfigures.pending ())
826 - {
827 - /* Tell plugins its ok to start doing stupid things again but
828 - * obviously FIXME */
829 - CompOption::Vector options;
830 - CompOption::Value v;
831 -
832 - options.push_back (CompOption ("window", CompOption::TypeInt));
833 - v.set ((int) id);
834 - options.back ().set (v);
835 - options.push_back (CompOption ("active", CompOption::TypeInt));
836 - v.set ((int) 0);
837 - options.back ().set (v);
838 -
839 - /* Notify other plugins that it is unsafe to change geometry or serverGeometry
840 - * FIXME: That API should not be accessible to plugins, this is a hack to avoid
841 - * breaking ABI */
842 -
843 - screen->handleCompizEvent ("core", "lock_position", options);
844 - }

DIE

924 -bool
925 -PrivateWindow::checkClear ()
926 -{
927 - if (pendingConfigures.pending ())
928 - {
929 - /* FIXME: This is a hack to avoid performance regressions
930 - * and must be removed in 0.9.6 */
931 - compLogMessage ("core", CompLogLevelWarn, "failed to receive ConfigureNotify event on 0x%x\n",
932 - id);
933 - pendingConfigures.dump ();
934 - pendingConfigures.clear ();
935 - }
936 -
937 - return false;
938 -}
939 -
940 void
941 compiz::X11::PendingEventQueue::add (PendingEvent::Ptr p)
942 {
943 @@ -2454,21 +2148,6 @@
944 mValueMask (valueMask),
945 mXwc (*xwc)
946 {
947 - CompOption::Vector options;
948 - CompOption::Value v;
949 -
950 - options.push_back (CompOption ("window", CompOption::TypeInt));
951 - v.set ((int) w);
952 - options.back ().set (v);
953 - options.push_back (CompOption ("active", CompOption::TypeInt));
954 - v.set ((int) 1);
955 - options.back ().set (v);
956 -
957 - /* Notify other plugins that it is unsafe to change geometry or serverGeometry
958 - * FIXME: That API should not be accessible to plugins, this is a hack to avoid
959 - * breaking ABI */
960 -
961 - screen->handleCompizEvent ("core", "lock_position", options);
962 }

DIE

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

GOOD IN THEORY:

* Most of the description sounds like steps in the right direction.

* Simplified: Replacing 559 lines of code with 247.

UNSURE ABOUT THE THEORY:

* "composite and opengl now use the geometry last sent to the server"
This should give maximum performance, sure, but what if (for some reason) the server rejects the configure request as being somehow invalid? Would you ever encounter that, like moving a window to an invalid location?
For stability and correctness (which should be the primary goal right now) don't we want "composite and opengl now use the geometry last received from the server"?

* Caveat #1 is not really a caveat. It's the right (intended) design. But if you want to see a (buggy) example of how to solve the chattiness look at the timer code here:
https://code.launchpad.net/~vanvugt/compiz-core/fix-764330-trunk/+merge/86497
This was slightly buggy though as it broke keyboatrd-initiated moves (incorrect pointer warp). Simple to fix, probably.

* Still no XFlush's to ensure XConfigureWindow happens immediately. Or are there adequate XSyncs already?

BAD IN THEORY:

* The diff is too big for a mere mortal (who hasn't worked on all the code before) to fully load into their mental interpreter and evaluate its theoretical correctness. Please read that sentence again; it's important.

GOOD IN PRACTICE:

* Window movement is smooth and predictable, but not really better than in oneiric.

BAD IN PRACTICE:

* The bug introduced is worse than the bug being fixed; Window contents offset from the frame/decorations whenever a window is created or maximized.

* All that code, but we still haven't reduced the lag below oneiric-levels. Still not as good as Gnome Shell.

CONCLUSION:

The theory is mostly good but there is at least one serious bug that needs fixing. We probably shouldn't risk 0.9.7.0 any more with large changes like this. We should aim to get this code into a later release when it's more stable, not 0.9.7.0.

review: Needs Fixing
Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal
Download full text (3.2 KiB)

> GOOD IN THEORY:
>
> * Most of the description sounds like steps in the right direction.
>
> * Simplified: Replacing 559 lines of code with 247.
>
>
> UNSURE ABOUT THE THEORY:
>
> * "composite and opengl now use the geometry last sent to the server"
> This should give maximum performance, sure, but what if (for some reason) the
> server rejects the configure request as being somehow invalid? Would you ever
> encounter that, like moving a window to an invalid location?
> For stability and correctness (which should be the primary goal right now)
> don't we want "composite and opengl now use the geometry last received from
> the server"?

See http://tronche.com/gui/x/xlib/window/XConfigureWindow.html

Since we have SubstructureRedirectMask, it is guarunteed that any ConfigureWindow request we make on a child of the root window is guarunteed to end up in that position (hence the event tracker).

The only error cases I can think of is where you ask for a window to be of a size < 0 or > MAXSHORT. Ideally we should check for that case.

>
> * Caveat #1 is not really a caveat. It's the right (intended) design. But if
> you want to see a (buggy) example of how to solve the chattiness look at the
> timer code here:
> https://code.launchpad.net/~vanvugt/compiz-core/fix-764330-trunk/+merge/86497
> This was slightly buggy though as it broke keyboatrd-initiated moves
> (incorrect pointer warp). Simple to fix, probably.
>

There are some optimizations that can be done here - for example, if a window is grabbed, you can withold sending positions to the server until it is ungrabbed, or there is some other reason why you would want to do that. When that happens you can send it all at once.

> * Still no XFlush's to ensure XConfigureWindow happens immediately. Or are
> there adequate XSyncs already?

The GLib mainloop work introduced an XFlush as soon as we have finished dispatching the X event source.

>
>
> BAD IN THEORY:
>
> * The diff is too big for a mere mortal (who hasn't worked on all the code
> before) to fully load into their mental interpreter and evaluate its
> theoretical correctness. Please read that sentence again; it's important.

Indeed. The good thing is that its mostly all deletion and consolidation.

>
>
> GOOD IN PRACTICE:
>
> * Window movement is smooth and predictable, but not really better than in
> oneiric.
>
>
> BAD IN PRACTICE:
>
> * The bug introduced is worse than the bug being fixed; Window contents offset
> from the frame/decorations whenever a window is created or maximized.

Alright. I haven't seen this yet, but I have an idea of what might cause it. Will have a look into it when I get some time.

>
> * All that code, but we still haven't reduced the lag below oneiric-levels.
> Still not as good as Gnome Shell.

This will be fixed once the chattyness goes away. (as above). Watch your X.org CPU usage as you move windows - it skyrockets because at the moment we flood it with requests.

>
>
> CONCLUSION:
>
> The theory is mostly good but there is at least one serious bug that needs
> fixing. We probably shouldn't risk 0.9.7.0 any more with large changes like
> this. We should aim to get this code into a later r...

Read more...

Revision history for this message
Alan Griffiths (alan-griffiths) wrote : Posted in a previous version of this proposal

OK, I'm still wading through the changes. But I feel like a rant...

1307 CompSize
1308 CompWindow::size () const
1309 {
1310 - return CompSize (priv->width + priv->geometry.border () * 2,
1311 - priv->height + priv->geometry.border () * 2);
1312 + return CompSize (priv->geometry.width () + priv->geometry.border () * 2,
1313 + priv->geometry.height () + priv->geometry.border () * 2);
1314 }

As a change this doesn't look too bad - but it ignores a horrid design!

1. Chained operations like "priv->geometry.width ()" imply too muck knowledge of the details of "priv".
   That is "priv->width ()" would reflect less coupling.

2. You're tracing the path "priv->geometry..." many times, which suggests that the logic belongs in "geometry".
   "return priv->geometry.size ()" or "return CompSize(priv->geometry)" would reflect better coherence.

So, assuming (because borders may be optional?) that there's an unambiguous mapping from CompWindow::Geometry:

a. Add a constructor to CompSize: "explicit CompSize(CompWindow::Geometry const& g) : mWidth(g.width () + g.border () * 2) ..."
b. Add an inline method "CompSize PrivateWindow::size () const { return CompSize(priv->geometry); }"
c. Rewrite the above as "CompSize CompWindow::size () const { "return priv->size ()"; }"

Revision history for this message
Alan Griffiths (alan-griffiths) wrote : Posted in a previous version of this proposal

264 - /* been shaded */
265 + /* been shaded
266 if (w->priv->height == 0)
267 {
268 if (w->id () == priv->activeWindow)
269 w->moveInputFocusTo ();
270 - }
271 + }*/

Comments are not for storing old versions of the code - that's what we use bzr for. ;)

review: Needs Fixing
Revision history for this message
Alan Griffiths (alan-griffiths) wrote : Posted in a previous version of this proposal

There's a lot of "<blah>.width () + <blah>.border () * 2" and "<blah>.height () + <blah>.border () * 2" around. Surely CompWindow::Geometry could have a "widthWithBorders" method - or maybe a free function?

I'm tempted by

template<Typename T>
inline int heightWithBorders(T const& blah) { return blah.height () + blah.border () * 2; }

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

> OK, I'm still wading through the changes. But I feel like a rant...
>
> 1307 CompSize
> 1308 CompWindow::size () const
> 1309 {
> 1310 - return CompSize (priv->width + priv->geometry.border () * 2,
> 1311 - priv->height + priv->geometry.border () * 2);
> 1312 + return CompSize (priv->geometry.width () + priv->geometry.border ()
> * 2,
> 1313 + priv->geometry.height () + priv->geometry.border () * 2);
> 1314 }
>
> As a change this doesn't look too bad - but it ignores a horrid design!
>
> 1. Chained operations like "priv->geometry.width ()" imply too muck knowledge
> of the details of "priv".
> That is "priv->width ()" would reflect less coupling.

Probably, although priv is just a private member with implementation details, I don't really see this as a large design problem.

>
> 2. You're tracing the path "priv->geometry..." many times, which suggests that
> the logic belongs in "geometry".
> "return priv->geometry.size ()" or "return CompSize(priv->geometry)" would
> reflect better coherence.
>
> So, assuming (because borders may be optional?) that there's an unambiguous
> mapping from CompWindow::Geometry:
>
> a. Add a constructor to CompSize: "explicit CompSize(CompWindow::Geometry
> const& g) : mWidth(g.width () + g.border () * 2) ..."
> b. Add an inline method "CompSize PrivateWindow::size () const { return
> CompSize(priv->geometry); }"
> c. Rewrite the above as "CompSize CompWindow::size () const { "return
> priv->size ()"; }"

+1 for all three

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

> 264 - /* been shaded */
> 265 + /* been shaded
> 266 if (w->priv->height == 0)
> 267 {
> 268 if (w->id () == priv->activeWindow)
> 269 w->moveInputFocusTo ();
> 270 - }
> 271 + }*/
>
> Comments are not for storing old versions of the code - that's what we use bzr
> for. ;)

Indeed, this is a small portion of the code and I'm not yet sure what to do with it. When I get around to revisiting this section (when I actually get time to look at this again, wanted to get it up for review early) I'll see what can be done.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

> There's a lot of "<blah>.width () + <blah>.border () * 2" and "<blah>.height
> () + <blah>.border () * 2" around. Surely CompWindow::Geometry could have a
> "widthWithBorders" method - or maybe a free function?
>
> I'm tempted by
>
> template<Typename T>
> inline int heightWithBorders(T const& blah) { return blah.height () +
> blah.border () * 2; }

I'm not really sure ?

The complexity here comes from a legacy part of the X Server coming into play - a window could have fixed dimentions, but also specify a "border" which would be exclusive of the fixed dimentions, so you'd have to take this into account for all positioning operations (Xterm comes to mind here).

I definitely agree that geom.width () + geom.border () * 2 feels fragile and indeed, that has tripped me up many times before. Maybe a default parameter with a bitflag makes sense here, eg IncludeBorder , IncludeBorderFirst , IncludeBorderSecond (as it is on both sides)

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

I'll note that the above isn't *really* all that relevant to this merge though, but good things to keep in mind in any case. I'd rather not see the diff get any better except for adding tests.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Please resubmit for target branch lp:compiz-core (0.9.7)

review: Needs Resubmitting
Revision history for this message
Alan Griffiths (alan-griffiths) wrote : Posted in a previous version of this proposal

A lot of the above is about making the code better (which the proposal doesn't attempt). However, in a previous version I pointed out that:

> 264 - /* been shaded */
> 265 + /* been shaded
> 266 if (w->priv->height == 0)
> 267 {
> 268 if (w->id () == priv->activeWindow)
> 269 w->moveInputFocusTo ();
> 270 - }
> 271 + }*/
>
> Comments are not for storing old versions of the code - that's what we use bzr for. ;)

I still think that needs fixing.

review: Needs Fixing
Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Alright, I've updated this so that there's no distorted windows on decoration size change. Was (ironically) a race condition.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

For the sake of not making this diff any bigger, I'm not going to introduce unit tests here.

As for testing this I'd say the following is appropriate:

 * Add testcase for PendingConfigureEvent
 * Add testcase for rectsToRegion (serverGeometry should really be DI'd here)

Notes for the future:

 * As alan has said - the whole priv->geometry ().width () + priv->geometry ().border () * 2 is /really/ awkward and error prone, but we need it to support windows like xterm that still use this (stupid, legacy) attribute on their windows. I would say that the role of compiz::window::Geometry should thus be expanded somewhat
   - it should also encapsulate priv->region and really, geometry::x and friends should be made with reference to that. Ideally we'll store two regions, one with borders and one without. That sucks, since its a little memory hungry, but at least the cost is only really born whenever those regions need to be updated (::translate on a region with one rectangle is basically free, slightly more complicated on regions with multiple rects (eg, shaped window). The default should be to return the size /without/ borders, and have a default-parameter enum to get the size with borders.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Unfortunately still fails basic testing. The same bug remains;

* The bug introduced is worse than the bug being fixed; Window contents offset from the frame/decorations whenever a window is created or maximized.

review: Needs Fixing
Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Also... I would have thought/hoped that a simplified/stable solution would eliminate "serverGeometry" completely.

        /**
        * Geometry retrieved from the
         * last ConfigureNotify event received
         */
        Geometry & geometry () const;

        /**
         * Geometry last sent to the server
         */
        Geometry & serverGeometry () const;

I understand why we have, and why we might want, serverGeometry. However it is an optimization which only makes sense to attempt if the code is stable and bug-free to begin with.

While serious bugs remain, I think the first goal should be to simplify the logic down to just using "geometry" and remove or stub "serverGeometry".

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

> Also... I would have thought/hoped that a simplified/stable solution would
> eliminate "serverGeometry" completely.
>
> /**
> * Geometry retrieved from the
> * last ConfigureNotify event received
> */
> Geometry & geometry () const;
>
> /**
> * Geometry last sent to the server
> */
> Geometry & serverGeometry () const;
>
> I understand why we have, and why we might want, serverGeometry. However it is
> an optimization which only makes sense to attempt if the code is stable and
> bug-free to begin with.
>
> While serious bugs remain, I think the first goal should be to simplify the
> logic down to just using "geometry" and remove or stub "serverGeometry".

There are some usescases for when we need to know what the last /received/ geometry from the server is. That's why the whole geometry/serverGeometry thing exists.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Also, I'm not seeing this offset problem - could you give me some steps to reproduce it ?

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

> > Also... I would have thought/hoped that a simplified/stable solution would
> > eliminate "serverGeometry" completely.
> >
> > /**
> > * Geometry retrieved from the
> > * last ConfigureNotify event received
> > */
> > Geometry & geometry () const;
> >
> > /**
> > * Geometry last sent to the server
> > */
> > Geometry & serverGeometry () const;
> >
> > I understand why we have, and why we might want, serverGeometry. However it
> is
> > an optimization which only makes sense to attempt if the code is stable and
> > bug-free to begin with.
> >
> > While serious bugs remain, I think the first goal should be to simplify the
> > logic down to just using "geometry" and remove or stub "serverGeometry".
>
>
> There are some usescases for when we need to know what the last /received/
> geometry from the server is. That's why the whole geometry/serverGeometry
> thing exists.

Hmm, we could make it so that serverGeometry is returned for managed windows and geometry is returned for override redirect windows on the public API. That of course means that we have to break the public API and update all the plugins.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

The practical problem:

Just start compiz --replace composite opengl move resize decor
and gtk-window-decorator --replace
Then every window will have its contents horizontally shifted around 15 pixels from the correct location relative to its decorations. The shift occurs whenever a window is created, maximized or restored. The shift goes away (corrects itself) when you start moving the window. On a positive note, the lag *appears* totally fixed and dragging windows is performing as well as Gnome Shell now.

The theoretical problem:

You can ignore my comments about geometry/serverGeometry for now. They are insignificant compared to the above bug. And besides, the use of both geometry and serverGeometry now appears to be yielding the desired reduction in lag.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Screenshot of the bug introduced by this branch:
https://launchpadlibrarian.net/92172452/shift.png

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Alright, I'll have another look into it ? (Not seeing it here :( )

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

And it gets worse:
* The usual lag has come back. I have no idea how or why it was fixed when I ran it the first time.
* When dragging windows around, the area of desktop recently exposed flashes white.

So there is no improvement in performance and 2 serious regressions introduced :(

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

    //XSynchronize (dpy, TRUE);

:) Haven't merged your other branch

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Yikes. That would be another regression to be missing:
src/screen.cpp: XSynchronize (dpy, synchronousX ? True : False);

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Perhaps remember to pull from trunk more often.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

where trunk == lp:compiz-core

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Why doesn't the removal of XSynchronize show up in the below diff? It's obvious if I merge the branch myself. Maybe LaunchPad needs a kick to update the below diff?

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

I think the horizontal shift bug is coming from upstream lp:compiz-core, and is exposed moreso by this branch. I can reproduce the same kind of horizontal shift jitter using just lp:compiz-core and resizing a window rapidly.

Revision history for this message
Alan Griffiths (alan-griffiths) wrote : Posted in a previous version of this proposal

> The default should be to return the size /without/ borders,
> and have a default-parameter enum to get the size with borders.

There are very few designs where default parameters are better than multiple functions. This isn't one.

auto sizeWithoutBorders = geometry.size();
auto sizeWithBorders = geometry.sizeWithBorders(); // not geometry.size(::compiz::geometry::WithBorders);

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

OK, the horizontal shifting bug does not seem to be caused by this branch. Just made worse by this branch.

Sam, please review that nasty issue first: bug 928173.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

This branch will incidentally help with a lot of the resizing issues, so I think it should be merged first.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

I would prefer not to approve this branch so long as the offset/shift bug is as bad as it is. If absolutely necessary, we could release without a fix for bug 928173 because it is hidden by the default Ubuntu config. But introducing this branch makes the bug unacceptably worse, as the screenshot showed.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

To clarify, "unacceptably worse" means that the bug will no longer be hidden in Ubuntu and will occur on every new window that opens. At least on my machine (and presumably others).

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

.... I'm still not even seeing this issue.

I can look into resizing tonight if you really think that this is a blocker.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

I think I may have confused the situation by suggesting the regression introduced by this branch is bug 928173. The symptoms are actually slightly different. It's only a theory that it's the same bug, because the size of the horizontal offset looks similar.

The bug I see with this branch would be worthy of a new bug report (a critical regression) that I don't want to log. And we won't have to log that bug so long as it's fixed before this branch is merged.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

On Wed, 8 Feb 2012, Daniel van Vugt wrote:

> I think I may have confused the situation by suggesting the regression introduced by this branch is bug 928173. The symptoms are actually slightly different. It's only a theory that it's the same bug, because the size of the horizontal offset looks similar.
>
> The bug I see with this branch would be worthy of a new bug report (a critical regression) that I don't want to log. And we won't have to log that bug so long as it's fixed before this branch is merged.
> --
> https://code.launchpad.net/~smspillaz/compiz-core/compiz-core.work_923683/+merge/91654
> You are the owner of lp:~smspillaz/compiz-core/compiz-core.work_923683.
>

Alright.

I am looking into the decor plugin for a way to resolve this.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Despite the success of the fix for #928173: lp:~smspillaz/compiz-core/compiz-core.decor_928173
the offset bug introduced by this branch persists.

This confirms that the horizontal offset/shift problems introduced (or exposed) by this branch are certainly not the same as bug 928173.

review: Needs Fixing
Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

I still can't even observe this problem that you're referring to :(

Can you specify, exactly what windows this occurs on and /exact/ steps to reproduce it ?

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Steps to reproduce the issue:

1. Start with an exact clone of lp:compiz-core
2. Merge in lp:~smspillaz/compiz-core/compiz-core.work_923683
3. Build and install it.
4. Run compiz --replace composite opengl move resize decor
5. Run gtk-window-decorator --replace

Now every window will be corrupted. Not just the existing ones, but any new ones you open too. The corruption persists during window resizing, but vanishes as soon as you move a window.

Here is a new screenshot:
https://launchpadlibrarian.net/92390683/shifted2.png

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Is the window itself shifted (eg are the input regions correctly lined up) or is it just the image that is shifted ?

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

The window itself is shifted. The input regions correctly line up with the image still.

However, it looks like the damage events are not shifted. This causes some redraw problems.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Can you post the xwininfo -all of the window, xwininfo -all -id of the parent and xwininfo -all -parent of the parent of that ?

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Attachments sent via email. This is interesting though;

$ diff correct.xwininfo shifted.xwininfo
11c11
< 0x2c03ca7 (has no name): () 1x1+-1+-1 +45+69
---
> 0x2c03ca7 (has no name): () 1x1+-1+-1 +34+69
13c13
< Absolute upper-left X: 46
---
> Absolute upper-left X: 35
31,32c31,32
< Corners: +46+70 -1232+70 -1232-720 +46-720
< -geometry 80x24+35+41
---
> Corners: +35+70 -1243+70 -1243-720 +35-720
> -geometry 80x24+34+41

It looks like geometry/serverGeometry might be out of sync.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

And why do I have so many 1x1 pixel windows?! I swear occasionally I see them on screen too.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

On Sun, 12 Feb 2012, Daniel van Vugt wrote:

> And why do I have so many 1x1 pixel windows?!

Applications use them as a means of doing IPC. Gtk+ is a serial offender
here.

> I swear occasionally I see them on screen too.

Under what circumstances?

> --
> https://code.launchpad.net/~smspillaz/compiz-core/compiz-core.work_923683/+merge/91654
> You are the owner of lp:~smspillaz/compiz-core/compiz-core.work_923683.
>

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

On Sun, 12 Feb 2012, Daniel van Vugt wrote:

The frame window is in the wrong position. This means that serverGeometry
isn't the last thing sent to the server. serverGeometry is still writable
in this branch, so it could be $evilplugin doing the wrong thing here.

I don't want to spend any more time on this until wednesday. So I'll pick
it up then.

(compiz run with --debug will have some logs which can confirm this btw)

Incidentally, you don't see any of those "unmatched ConfigureNotify"
warnings do you? They're all gone here but they indicate the first sign of
trouble.

> Attachments sent via email. This is interesting though;
>
> $ diff correct.xwininfo shifted.xwininfo
> 11c11
> < 0x2c03ca7 (has no name): () 1x1+-1+-1 +45+69
> ---
>> 0x2c03ca7 (has no name): () 1x1+-1+-1 +34+69
> 13c13
> < Absolute upper-left X: 46
> ---
>> Absolute upper-left X: 35
> 31,32c31,32
> < Corners: +46+70 -1232+70 -1232-720 +46-720
> < -geometry 80x24+35+41
> ---
>> Corners: +35+70 -1243+70 -1243-720 +35-720
>> -geometry 80x24+34+41
>
> It looks like geometry/serverGeometry might be out of sync.
> --
> https://code.launchpad.net/~smspillaz/compiz-core/compiz-core.work_923683/+merge/91654
> You are the owner of lp:~smspillaz/compiz-core/compiz-core.work_923683.
>

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

It's random and very rare. But sometimes I see 1-pixel windows (with shadow). Can't ever seem to get xwininfo for them. Sometimes I also get very large white anonymous windows blocking my view. But it's all very hard to reproduce. And off-topic.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Correct. I did stop getting "Warn: failed to receive ConfigureNotify event on ..." when using this branch.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Well, you should stop getting "failed to receive" since I've removed the timeout which checks for unmatched events :) (A necessary hack coming up to the oneiric release).

I'm more interested in warnings that say "unmatched ConfigureNotify"

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Could you have a look again to see if the issue is still occurring for you ? I've just re-synced with trunk, so it may be fixed.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

OK, I found the issue. Its actually a bug that's been in the code for quite some time which would cause the client to not move within the frame when the frame was updated until the client position itself was updated.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

There's still one more issue here though to do with override redirect windows not getting their paint regions updated, so that needs to be looked into as well. And tests.

review: Needs Fixing
Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

^ Those fix the offset issue confirmed here (was finally able to reproduce it with qtcreator)

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

1. Confirmed the shifted window regression is now fixed. Seems to work without any obvious bugs now.

2. I'm not entirely sure about replacing geometry with serverGeometry. I thought it would be a good idea to do the opposite. There would be a slight performance penalty, but using just "geometry" would guarantee that compiz always agrees with the actual server geometry, instead of guessing/assuming that serverGeometry is accepted by the server (which does not always seem to be true, hence bug 886192).

3. There are two instances of double semicolons ";;" in this proposal.

4. I suspect this proposal will conflict just slightly with the proposal for bug 940139. But it should be minor.

5. NEW REGRESSION:
375 - XSynchronize (dpy, synchronousX ? True : False);
376 +// priv->connection = XGetXCBConnection (priv->dpy);

6. Why always two spaces before "* 2" ?

review: Needs Fixing
Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

> 2. I'm not entirely sure about replacing geometry with serverGeometry. I
> thought it would be a good idea to do the opposite. There would be a slight
> performance penalty, but using just "geometry" would guarantee that compiz
> always agrees with the actual server geometry, instead of guessing/assuming
> that serverGeometry is accepted by the server (which does not always seem to
> be true, hence bug 886192).
>

serverGeometry will always be "accepted" by the server as long as the window is managed and the requested window geometry will not generate a BadValue error (eg, 0 < 1 < MAXINT) (eg, it is not override redirect and it is a child of a window for which we have a SubstructureRedirectMask.

Bug 886192 is not an example of this. In fact, the behaviour exhibited by bug 886192 is primarily /because/ we use the geometry last received from the server rather than geometry last sent, and the latency of which explains the reason why the window movement lags the cursor, because the server is "catching up".

The only case where you can't guaruntee that a similar update for geometry is going to be delivered by the server for serverGeometry as stored on XConfigureWindow is either in the case that A) Another client has SubstructureRedirectMask on the root window or a parent window owned by compiz and in that case compiz shouldn't even touch that window at all or B) override redirect windows, and as you can see in the code, we /always/ use geometry for override redirect windows in placement sensitive operations. Incidentally, override redirect windows are why getting window stacking right is such a nightmare, but thats a topic for another day.

> 3. There are two instances of double semicolons ";;" in this proposal.

Fix
>
> 4. I suspect this proposal will conflict just slightly with the proposal for
> bug 940139. But it should be minor.

Yes, fixable

>
> 5. NEW REGRESSION:
> 375 - XSynchronize (dpy, synchronousX ? True : False);
> 376 +// priv->connection = XGetXCBConnection (priv->dpy);

Please elaborate further on why this is a regression.

>
> 6. Why always two spaces before "* 2" ?

Most likely copy-and-paste errors

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

#5 is a regression because it will make "--sync" be ignored. Line 375 is important and should not be deleted.

review: Needs Fixing
Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

On Mon, 27 Feb 2012, Daniel van Vugt wrote:

> Review: Needs Fixing
>
> #5 is a regression because it will make "--sync" be ignored. Line 375 is important and should not be deleted.

Thank you. Fixing.

> --
> https://code.launchpad.net/~smspillaz/compiz-core/compiz-core.work_923683/+merge/94683
> You are the owner of lp:~smspillaz/compiz-core/compiz-core.work_923683.
>

Revision history for this message
Mikkel Kamstrup Erlandsen (kamstrup) wrote : Posted in a previous version of this proposal

> > #5 is a regression because it will make "--sync" be ignored. Line 375 is
> important and should not be deleted.
>
> Thank you. Fixing.

Be sure to add a comment there explaining why that call needs to be there - this review demonstrates that it should have been there in the first place :-)

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Comments are good. Though the string "synchronousX" is unique so its use should be quite clear already without a comment.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Looks like a mistake:

710 - if (x2 > priv->width)
711 - x2 = priv->width;
712 - if (y2 > priv->height)
713 - y2 = priv->height;
714 + if (x2 > priv->serverGeometry.width ())
715 + x2 = priv->serverGeometry.height (); <====== width?
716 + if (y2 > priv->serverGeometry.width ()) <====== height?
717 + y2 = priv->serverGeometry.height ();

review: Needs Fixing
Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Also, there seems to be a HUGE performance regression when moving windows with this new branch. Try running under callgrind. With lp:compiz-core it's nice and fast. However with this branch it is painfully slow moving windows.

About 75% of the time is spent in:
  775,164,942 /home/dan/bzr/compiz-core/sam/plugins/move/src/move.cpp:moveHandleMotionEvent(CompScreen*, int, int) [/home/dan/sam/lib/compiz/libmove.so]
  772,734,309 /home/dan/bzr/compiz-core/sam/src/window.cpp:CompWindow::configureXWindow(unsigned int, XWindowChanges*) [/home/dan/sam/lib/libcompiz_core.so.0.9.7.0]
  771,062,144 /home/dan/bzr/compiz-core/sam/src/window.cpp:CompWindow::move(int, int, bool) [/home/dan/sam/lib/libcompiz_core.so.0.9.7.0]
  636,443,803 /home/dan/bzr/compiz-core/sam/src/window.cpp:CompWindow::moveNotify(int, int, bool) [/home/dan/sam/lib/libcompiz_core.so.0.9.7.0]
  636,266,731 /home/dan/bzr/compiz-core/sam/plugins/decor/src/decor.cpp:DecorWindow::moveNotify(int, int, bool) [/home/dan/sam/lib/compiz/libdecor.so]

review: Needs Fixing
Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

The main paths causing this massive performance regression are:

70% of the time:
CompWindow::move
DecorWindow::moveNotify
DecorWindow::computeShadowRegion

18% of the time:
CompWindow::move
CompWindow::configureXWindow
PrivateWindow::reconfigureXWindow

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Hmm, I would imagine this is the throttling of motion events that we mentioned earlier. I think in lp:compiz-core we would have just spent all that time in CompWindow::move although I'll conceed that configureXWindow is a bit more of an ... involved function (although I haven't noticed any *noticable* performance problems).

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

I also recommend "kcachegrind" to display callgrind output in a pretty graphical way.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Try using valgrind/callgrind to simulate a slow machine. You will see a monumental difference in performance when dragging windows. Even my i7 (running callgrind) can't keep up with dragging windows using this branch. But it is almost perfectly fluid without this branch.

Maybe it is the fact that all motion events are getting though now. But regardless, people with slow-to-regular machines will be noticeably impacted. We need a fix before accepting this.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

From what I can tell, the problem is that moveNotify is called much more often with this branch. However moveNotify is far too expensive to call so often, mostly due to DecorWindow::moveNotify.

Does DecorWindow::moveNotify really need to computeShadowRegion on EVERY window whenever a single window moves?

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

One small caveat is that I haven't merged lp:compiz-core into this one today, which includes the region fixes.

> Does DecorWindow::moveNotify really need to computeShadowRegion on EVERY window whenever a single window moves?

Yes it does. There are some optimizations we can do there to make it only compute regions on certain windows. I've done them before, so its not too hard. Not for this branch.

> From what I can tell, the problem is that moveNotify is called much more often with this branch. However moveNotify is far too expensive to call so often, mostly due to DecorWindow::moveNotify.

So I think there are three optimizations which can be applied here, all of which are quite involved, so I'll need to put them in separate branches.

Firstly, we need to throttle motion events, so they aren't dispatched spuriously. That's easy enough to do.

The second which is a bit more complicated is to also throttle dispatch of serverGeometry updates / moveNotify / XConfigureWindow as well. There are some plugins which need immediate move notify updates but probably not immediate move notify updates 1000 times a second.

Another optimization is to only call XConfigureWindow when we actually need to update the server side position and not necessarily all the time. This does mean that serverGeometry will be "ahead" of whats being sent to the server, but that's OK since we can just flush the changes on demand (and besides, its meant to be ahead, which is why its there).

Incidentally, I'm not able to get such high readings in callgrind for window movement, but I guess this is a case-by-case thing.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Window movement accounts for all my CPU because that's all I am testing since I noticed the problem;
  Start compiz
  Move windows around lots
  Stop compiz

With this branch I see movement account for 60-80% of compiz' time and is incredibly slow and stuttering. Without this branch, it only uses maximum 20% of compiz' time and is visually much smoother.

Revision history for this message
Daniel van Vugt (vanvugt) : Posted in a previous version of this proposal
review: Needs Fixing
Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

I merged lp:compiz-core from today. Could you give it another try and let me know how it goes ?

Also some thoughts on those optimizations would be nice :)

Currently I'm testing using the entire unity stack, but I can reduce it down to just compiz and a basic set of plugins.

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

I was already testing this branch merged with the latest lp:compiz-core (including yesterday's optimizations) all along. Which makes the regression from this branch more worrying.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

tOn Tue, 28 Feb 2012, Daniel van Vugt wrote:

> I was already testing this branch merged with the latest lp:compiz-core (including yesterday's optimizations) all along. Which makes the regression from this branch more worrying.
> --
> https://code.launchpad.net/~smspillaz/compiz-core/compiz-core.work_923683/+merge/94923
> You are the owner of lp:~smspillaz/compiz-core/compiz-core.work_923683.
>

OK, well, I know there are some code paths which can be slow, so lets look
at optimizing. Do you have any thoughts about the ideas I raised ?

Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

I don't have the encyclopaedic knowledge of CompWindow or the time to research your suggestions right now. So no comment.

But I'm happy to test/profile any further additions to this proposal... tomorrow.

(Daniel logs off)

review: Needs Fixing
Revision history for this message
Alan Griffiths (alan-griffiths) wrote : Posted in a previous version of this proposal

I don't see anything obviously wrong, but would like to hear the results of Daniel's profiling.

review: Abstain
Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

Tested revision 2994. Moving windows is still so slow it's unusable (under valgrind). We should fix this because it's a good indication that the same would occur in the wild on a slow machine.

compiz is spending 70% of it's CPU in or below moveHandleMotionEvent. The two main CPU hogs are:

(73% of the 70%)
moveHandleMotionEvent
CompWindow::move
DecorWindow::moveNotify
DecorWindow::computeShadowRegion

(16% of the 70%)
moveHandleMotionEvent
CompWindow::move
CompWindow::configureXWindow
PrivateWindow::reconfigureXWindow

review: Needs Fixing
Revision history for this message
Daniel van Vugt (vanvugt) wrote : Posted in a previous version of this proposal

And no, combining this branch with lp:~smspillaz/compiz-core/compiz-core.optimization-inlining
does not solve or change the performance problem.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

Indeed, that's not meant to completely fix it, but it does fix some hotspots and other inefficiencies I observed :)

Revision history for this message
Alan Griffiths (alan-griffiths) wrote : Posted in a previous version of this proposal

I'm concerned that the code still looks "fragile" (to use Sam's word). For example:

50 + y1 = geom.height () + geom.border ();
...
59 + y2 = geom.height () - geom.border ();

It isn't explicit why border is added in one case and subtracted in the other. I suspect that there's an abstraction missing (or at least suitably named functions that derive the appropriate result from geom). Something like "internalHeight = internalHeight(geom)" is much easier to follow.

I've not tested the latest version enough to be confident in it. Will aim to do so on Thursday.

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal

> I'm concerned that the code still looks "fragile" (to use Sam's word). For
> example:
>
> 50 + y1 = geom.height () + geom.border ();
> ...
> 59 + y2 = geom.height () - geom.border ();
>
> It isn't explicit why border is added in one case and subtracted in the other.
> I suspect that there's an abstraction missing (or at least suitably named
> functions that derive the appropriate result from geom). Something like
> "internalHeight = internalHeight(geom)" is much easier to follow.
>
> I've not tested the latest version enough to be confident in it. Will aim to
> do so on Thursday.

Ack, I'll add something like that to compiz::window::Geometry

Revision history for this message
Sam Spilsbury (smspillaz) wrote : Posted in a previous version of this proposal
Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Tested this branch by itself and merged with lp:~smspillaz/compiz-core/compiz-core.fix_969108.2

By itself, this branch still makes window movement unacceptably slow. With the other branch designed to make movement more efficient, windows now often don't move at all when dragged.

Unfortunately performance is getting worse, not better.

review: Needs Fixing
Revision history for this message
Sam Spilsbury (smspillaz) wrote :

Its probably a small typo during refactoring. Please don't panic

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

I got this branch performing somewhat better by combining it with:
https://code.launchpad.net/~vanvugt/compiz-core/lastMotionTime/+merge/100742

However I would say still noticeably slower than lp:compiz-core. :(

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Confirmed the performance problems with this branch do seem to be fixed by:
https://code.launchpad.net/~smspillaz/compiz-core/compiz-core.fix_969101/+merge/100270

Revision history for this message
Daniel van Vugt (vanvugt) wrote :

Please resubmit with prerequisite:
lp:~smspillaz/compiz-core/compiz-core.fix_969101

review: Needs Resubmitting
Revision history for this message
Daniel van Vugt (vanvugt) wrote :

This branch does appear to fix bug 862430.

2997. By Sam Spilsbury

Fix LP #932520 - Update regions correctly when sending geometry updates

Use information last sent to server to determine window regions and shape
regions, update regions for resizing when we send the information to the server
as the shape rectangles will be correct anyways (even though it costs us a
round trip).

2998. By Sam Spilsbury

Replace instances of geometry with serverGeoemtry where appropriate

2999. By Sam Spilsbury

Merge lp:compiz-core

3000. By Sam Spilsbury

Fix typoo

3001. By Sam Spilsbury

Merge trunk

3002. By Sam Spilsbury

Remove abi breaks

3003. By Sam Spilsbury

Don't break abi

3004. By Sam Spilsbury

Use geometry last sent to server to calculate damage rects

3005. By Sam Spilsbury

Use server side rects for calculating clip regions

3006. By Sam Spilsbury

Use serverGeometry

3007. By Sam Spilsbury

Damage the screen when windows are resized

3008. By Sam Spilsbury

Style issue

3009. By Sam Spilsbury

Merged trunk and fixed a number of problems with resize synchronization

3010. By Sam Spilsbury

Send resize notification immediately for windows that aren't override redirect

3011. By Sam Spilsbury

Fix jumping around on resize because we were rebinding way too much at the wrong time.

Only rebind on paints.

3012. By Sam Spilsbury

Don't not update decorations when we're grabbed for resizing

3013. By Sam Spilsbury

Null check

3014. By Sam Spilsbury

resize: unconditionally finish the resize operation when releasing the button

3015. By Sam Spilsbury

Merge trunk

3016. By Sam Spilsbury

Don't send a moveNotify and resizeNotify if the window was moved and resized

3017. By Sam Spilsbury

Cleanup wobbly model update code

3018. By Sam Spilsbury

Ensure that override redirect windows get their regions set correctly.
Also they can't have input extents.

3019. By Sam Spilsbury

Force all override redirect windows to use geometry last received from
server. Force all managed windows to use geometry last sent.

The difference between geometry and server* is now moot and can be
removed in a future release

3020. By Sam Spilsbury

Clean some cruft

3021. By Sam Spilsbury

Update decoration matrix of shaded windows

3022. By Sam Spilsbury

Send a resizeNotify to indicate to plugins when map operations have completed

3023. By Sam Spilsbury

Merge lp:compiz-core

3024. By Sam Spilsbury

Don't notify plugins of unreparent in the right place

3025. By Sam Spilsbury

Merged lp:~vanvugt/compiz-core/compiz-core/fix-981703-properly

3026. By Sam Spilsbury

updateState is bitwise, so never clobber it

3027. By Sam Spilsbury

Base decision to send notifications to plugins on absolute position
changes

3028. By Sam Spilsbury

Don't set height = 0 if the window still has a pixmap

3029. By Sam Spilsbury

Remove some unnecessary cahgnes in the wobbly plugin

3030. By Sam Spilsbury

Merge lp:compiz-core

3031. By Sam Spilsbury

Now that geometry and serverGeometry both return the same thing, the changes
in client code don't make any sense to keep

3032. By Sam Spilsbury

Fix logic errors

3033. By Sam Spilsbury

!= 0 not necessary

3034. By Sam Spilsbury

Merge lp:compiz-core

Unmerged revisions

3034. By Sam Spilsbury

Merge lp:compiz-core

3033. By Sam Spilsbury

!= 0 not necessary

3032. By Sam Spilsbury

Fix logic errors

3031. By Sam Spilsbury

Now that geometry and serverGeometry both return the same thing, the changes
in client code don't make any sense to keep

3030. By Sam Spilsbury

Merge lp:compiz-core

3029. By Sam Spilsbury

Remove some unnecessary cahgnes in the wobbly plugin

3028. By Sam Spilsbury

Don't set height = 0 if the window still has a pixmap

3027. By Sam Spilsbury

Base decision to send notifications to plugins on absolute position
changes

3026. By Sam Spilsbury

updateState is bitwise, so never clobber it

3025. By Sam Spilsbury

Merged lp:~vanvugt/compiz-core/compiz-core/fix-981703-properly

Preview Diff

[H/L] Next/Prev Comment, [J/K] Next/Prev File, [N/P] Next/Prev Hunk
1=== modified file 'include/core/window.h'
2--- include/core/window.h 2012-03-22 17:00:51 +0000
3+++ include/core/window.h 2012-04-04 05:26:30 +0000
4@@ -410,9 +410,9 @@
5
6 bool hasUnmapReference ();
7
8- bool resize (XWindowAttributes);
9+ bool resize (const XWindowAttributes &);
10
11- bool resize (Geometry);
12+ bool resize (const Geometry &);
13
14 bool resize (int x, int y, int width, int height,
15 int border = 0);
16
17=== modified file 'plugins/composite/src/window.cpp'
18--- plugins/composite/src/window.cpp 2012-03-22 17:00:51 +0000
19+++ plugins/composite/src/window.cpp 2012-04-04 05:26:30 +0000
20@@ -250,7 +250,7 @@
21
22 if (x2 > x1 && y2 > y1)
23 {
24- CompWindow::Geometry geom = priv->window->geometry ();
25+ const CompWindow::Geometry &geom = priv->window->serverGeometry ();
26
27 x1 += geom.x () + geom.border ();
28 y1 += geom.y () + geom.border ();
29@@ -272,20 +272,20 @@
30 {
31 int x1, x2, y1, y2;
32
33- CompWindow::Geometry geom = priv->window->geometry ();
34- CompWindowExtents output = priv->window->output ();
35+ const CompWindow::Geometry &geom = priv->window->serverGeometry ();
36+ const CompWindowExtents &output = priv->window->output ();
37
38 /* top */
39 x1 = -output.left - geom.border ();
40 y1 = -output.top - geom.border ();
41- x2 = priv->window->size ().width () + output.right - geom.border ();
42+ x2 = geom.width () + geom.border () + output.right;
43 y2 = -geom.border ();
44
45 if (x1 < x2 && y1 < y2)
46 addDamageRect (CompRect (x1, y1, x2 - x1, y2 - y1));
47
48 /* bottom */
49- y1 = priv->window->size ().height () - geom.border ();
50+ y1 = geom.height () + geom.border ();
51 y2 = y1 + output.bottom - geom.border ();
52
53 if (x1 < x2 && y1 < y2)
54@@ -295,13 +295,13 @@
55 x1 = -output.left - geom.border ();
56 y1 = -geom.border ();
57 x2 = -geom.border ();
58- y2 = priv->window->size ().height () - geom.border ();
59+ y2 = geom.height () - geom.border ();
60
61 if (x1 < x2 && y1 < y2)
62 addDamageRect (CompRect (x1, y1, x2 - x1, y2 - y1));
63
64 /* right */
65- x1 = priv->window->size ().width () - geom.border ();
66+ x1 = geom.width () - geom.border ();
67 x2 = x1 + output.right - geom.border ();
68
69 if (x1 < x2 && y1 < y2)
70@@ -322,7 +322,7 @@
71 x = rect.x ();
72 y = rect.y ();
73
74- CompWindow::Geometry geom = priv->window->geometry ();
75+ const CompWindow::Geometry &geom = priv->window->serverGeometry ();
76 x += geom.x () + geom.border ();
77 y += geom.y () + geom.border ();
78
79@@ -341,16 +341,16 @@
80 if (priv->window->shaded () || force ||
81 (priv->window->isViewable ()))
82 {
83- int border = priv->window->geometry ().border ();
84+ int border = priv->window->serverGeometry ().border ();
85
86 int x1 = -MAX (priv->window->output ().left,
87 priv->window->input ().left) - border;
88 int y1 = -MAX (priv->window->output ().top,
89 priv->window->input ().top) - border;
90- int x2 = priv->window->size ().width () +
91+ int x2 = priv->window->serverGeometry ().width () +
92 MAX (priv->window->output ().right,
93 priv->window->input ().right) ;
94- int y2 = priv->window->size ().height () +
95+ int y2 = priv->window->serverGeometry ().height () +
96 MAX (priv->window->output ().bottom,
97 priv->window->input ().bottom) ;
98 CompRect r (x1, y1, x2 - x1, y2 - y1);
99@@ -410,7 +410,7 @@
100
101 if (!w->damageRect (initial, CompRect (x, y, width, height)))
102 {
103- CompWindow::Geometry geom = w->priv->window->geometry ();
104+ CompWindow::Geometry geom = w->priv->window->serverGeometry ();
105
106 x += geom.x () + geom.border ();
107 y += geom.y () + geom.border ();
108@@ -622,14 +622,14 @@
109 {
110 int x, y, x1, x2, y1, y2;
111
112- x = window->geometry ().x ();
113- y = window->geometry ().y ();
114+ x = window->serverGeometry ().x ();
115+ y = window->serverGeometry ().y ();
116
117 x1 = x - window->output ().left - dx;
118 y1 = y - window->output ().top - dy;
119- x2 = x + window->size ().width () +
120+ x2 = x + window->serverGeometry ().width () +
121 window->output ().right - dx;
122- y2 = y + window->size ().height () +
123+ y2 = y + window->serverGeometry ().height () +
124 window->output ().bottom - dy;
125
126 cScreen->damageRegion (CompRegion (CompRect (x1, y1, x2 - x1, y2 - y1)));
127
128=== modified file 'plugins/decor/src/decor.cpp'
129--- plugins/decor/src/decor.cpp 2012-03-22 17:00:51 +0000
130+++ plugins/decor/src/decor.cpp 2012-04-04 05:26:30 +0000
131@@ -91,7 +91,7 @@
132 void
133 DecorWindow::computeShadowRegion ()
134 {
135- shadowRegion = CompRegion (window->outputRect ());
136+ shadowRegion = CompRegion (window->serverOutputRect ());
137
138 if (window->type () == CompWindowTypeDropdownMenuMask ||
139 window->type () == CompWindowTypePopupMenuMask)
140@@ -141,11 +141,11 @@
141 if (window->type () == CompWindowTypeDropdownMenuMask &&
142 shadowRegion == CompRegionRef (window->outputRect ().region ()))
143 {
144- CompRect area (window->outputRect ().x1 (),
145- window->outputRect ().y1 (),
146- window->outputRect ().width (),
147- window->inputRect ().y1 () -
148- window->outputRect ().y1 ());
149+ CompRect area (window->serverOutputRect ().x1 (),
150+ window->serverOutputRect ().y1 (),
151+ window->serverOutputRect ().width (),
152+ window->serverInputRect ().y1 () -
153+ window->serverOutputRect ().y1 ());
154
155 shadowRegion -= area;
156 }
157@@ -1072,8 +1072,8 @@
158 for (i = 0; i < wd->nQuad; i++)
159 {
160 int x, y;
161- unsigned int width = window->size ().width ();
162- unsigned int height = window->size ().height ();
163+ unsigned int width = window->geometry ().width ();
164+ unsigned int height = window->geometry ().height ();
165
166 if (window->shaded ())
167 height = 0;
168@@ -1082,8 +1082,8 @@
169 &x1, &y1, &x2, &y2, &sx, &sy);
170
171 /* Translate by x and y points of this window */
172- x = window->geometry ().x ();
173- y = window->geometry ().y ();
174+ x = window->serverGeometry ().x ();
175+ y = window->serverGeometry ().y ();
176
177 wd->quad[i].box.x1 = x1 + x;
178 wd->quad[i].box.y1 = y1 + y;
179@@ -1111,8 +1111,8 @@
180 bool
181 DecorWindow::checkSize (const Decoration::Ptr &decoration)
182 {
183- return (decoration->minWidth <= (int) window->size ().width () &&
184- decoration->minHeight <= (int) window->size ().height ());
185+ return (decoration->minWidth <= (int) window->geometry ().width () &&
186+ decoration->minHeight <= (int) window->geometry ().height ());
187 }
188
189 /*
190@@ -2124,8 +2124,8 @@
191 {
192 int x, y;
193
194- x = window->geometry (). x ();
195- y = window->geometry (). y ();
196+ x = window->serverGeometry (). x ();
197+ y = window->serverGeometry (). y ();
198
199 region += frameRegion.translated (x - wd->decor->input.left,
200 y - wd->decor->input.top);
201@@ -2146,7 +2146,7 @@
202 void
203 DecorWindow::updateWindowRegions ()
204 {
205- const CompRect &input (window->inputRect ());
206+ const CompRect &input (window->serverInputRect ());
207
208 if (regions.size () != gWindow->textures ().size ())
209 regions.resize (gWindow->textures ().size ());
210
211=== modified file 'plugins/move/src/move.cpp'
212--- plugins/move/src/move.cpp 2012-02-16 05:31:28 +0000
213+++ plugins/move/src/move.cpp 2012-04-04 05:26:30 +0000
214@@ -138,8 +138,8 @@
215 {
216 int xRoot, yRoot;
217
218- xRoot = w->geometry ().x () + (w->size ().width () / 2);
219- yRoot = w->geometry ().y () + (w->size ().height () / 2);
220+ xRoot = w->serverGeometry ().x () + (w->size ().width () / 2);
221+ yRoot = w->serverGeometry ().y () + (w->size ().height () / 2);
222
223 s->warpPointer (xRoot - pointerX, yRoot - pointerY);
224 }
225@@ -169,8 +169,8 @@
226 if (ms->w)
227 {
228 if (state & CompAction::StateCancel)
229- ms->w->move (ms->savedX - ms->w->geometry ().x (),
230- ms->savedY - ms->w->geometry ().y (), false);
231+ ms->w->move (ms->savedX - ms->w->serverGeometry ().x (),
232+ ms->savedY - ms->w->serverGeometry ().y (), false);
233
234 ms->w->syncPosition ();
235
236@@ -314,12 +314,12 @@
237
238 w = ms->w;
239
240- wX = w->geometry ().x ();
241- wY = w->geometry ().y ();
242- wWidth = w->geometry ().width () +
243- w->geometry ().border () * 2;
244- wHeight = w->geometry ().height () +
245- w->geometry ().border () * 2;
246+ wX = w->serverGeometry ().x ();
247+ wY = w->serverGeometry ().y ();
248+ wWidth = w->serverGeometry ().width () +
249+ w->serverGeometry ().border () * 2;
250+ wHeight = w->serverGeometry ().height () +
251+ w->serverGeometry ().border () * 2;
252
253 ms->x += xRoot - lastPointerX;
254 ms->y += yRoot - lastPointerY;
255@@ -484,8 +484,8 @@
256
257 if (dx || dy)
258 {
259- w->move (wX + dx - w->geometry ().x (),
260- wY + dy - w->geometry ().y (), false);
261+ w->move (wX + dx - w->serverGeometry ().x (),
262+ wY + dy - w->serverGeometry ().y (), false);
263
264 if (!ms->optionGetLazyPositioning ())
265 w->syncPosition ();
266
267=== modified file 'plugins/opengl/src/window.cpp'
268--- plugins/opengl/src/window.cpp 2012-03-22 17:00:51 +0000
269+++ plugins/opengl/src/window.cpp 2012-04-04 05:26:30 +0000
270@@ -76,7 +76,7 @@
271 void
272 PrivateGLWindow::setWindowMatrix ()
273 {
274- CompRect input (window->inputRect ());
275+ CompRect input (window->serverInputRect ());
276
277 if (textures.size () != matrices.size ())
278 matrices.resize (textures.size ());
279@@ -344,7 +344,7 @@
280 void
281 PrivateGLWindow::updateWindowRegions ()
282 {
283- CompRect input (window->inputRect ());
284+ CompRect input (window->serverInputRect ());
285
286 if (regions.size () != textures.size ())
287 regions.resize (textures.size ());
288
289=== modified file 'src/event.cpp'
290--- src/event.cpp 2012-03-24 09:49:30 +0000
291+++ src/event.cpp 2012-04-04 05:26:30 +0000
292@@ -1251,7 +1251,7 @@
293 }
294
295 /* been shaded */
296- if (w->priv->height == 0)
297+ if (w->shaded ())
298 {
299 if (w->id () == priv->activeWindow)
300 w->moveInputFocusTo ();
301
302=== modified file 'src/privatewindow.h'
303--- src/privatewindow.h 2012-03-02 18:02:07 +0000
304+++ src/privatewindow.h 2012-04-04 05:26:30 +0000
305@@ -34,7 +34,6 @@
306 #include <core/timer.h>
307 #include "privatescreen.h"
308
309-
310 typedef CompWindowExtents CompFullscreenMonitorSet;
311
312 class PrivateWindow {
313@@ -167,6 +166,11 @@
314
315 bool handleSyncAlarm ();
316
317+ void move (int dx, int dy, bool sync);
318+ bool resize (int dx, int dy, int dwidth, int dheight, int dborder);
319+ bool resize (const CompWindow::Geometry &g);
320+ bool resize (const XWindowAttributes &attrib);
321+
322 void configure (XConfigureEvent *ce);
323
324 void configureFrame (XConfigureEvent *ce);
325@@ -241,13 +245,8 @@
326 XSizeHints sizeHints;
327 XWMHints *hints;
328
329- struct timeval lastGeometryUpdate;
330- struct timeval lastConfigureRequest;
331-
332 bool inputHint;
333 bool alpha;
334- int width;
335- int height;
336 CompRegion region;
337 CompRegion inputRegion;
338 CompRegion frameRegion;
339@@ -290,8 +289,6 @@
340 typedef std::pair <XWindowChanges, unsigned int> XWCValueMask;
341
342 compiz::X11::PendingEventQueue pendingConfigures;
343- CompTimer mClearCheckTimeout;
344- bool pendingPositionUpdates;
345
346 char *startupId;
347 char *resName;
348@@ -327,6 +324,8 @@
349
350 bool closeRequests;
351 Time lastCloseRequestTime;
352+
353+ bool nextMoveImmediate;
354 };
355
356 #endif
357
358=== modified file 'src/screen.cpp'
359--- src/screen.cpp 2012-03-30 06:26:33 +0000
360+++ src/screen.cpp 2012-04-04 05:26:30 +0000
361@@ -4657,6 +4657,8 @@
362 return false;
363 }
364
365+ /* Use synchronous behaviour when running with --sync, useful
366+ * for getting stacktraces when X Errors occurr */
367 XSynchronize (dpy, synchronousX ? True : False);
368
369 snprintf (displayString, 255, "DISPLAY=%s",
370
371=== modified file 'src/window.cpp'
372--- src/window.cpp 2012-03-30 11:43:20 +0000
373+++ src/window.cpp 2012-04-04 05:26:30 +0000
374@@ -77,8 +77,8 @@
375 PrivateWindow::isInvisible() const
376 {
377 return attrib.map_state != IsViewable ||
378- attrib.x + width + output.right <= 0 ||
379- attrib.y + height + output.bottom <= 0 ||
380+ attrib.x + geometry.width () + output.right <= 0 ||
381+ attrib.y + geometry.height () + output.bottom <= 0 ||
382 attrib.x - output.left >= (int) screen->width () ||
383 attrib.y - output.top >= (int) screen->height ();
384 }
385@@ -810,292 +810,14 @@
386 if (!serverFrame)
387 return;
388
389-
390- gettimeofday (&lastConfigureRequest, NULL);
391- /* Flush any changes made to serverFrameGeometry or serverGeometry to the server
392- * since there is a race condition where geometries will go out-of-sync with
393- * window movement */
394-
395- window->syncPosition ();
396-
397- if (serverInput.left || serverInput.right || serverInput.top || serverInput.bottom)
398- {
399- int bw = serverGeometry.border () * 2;
400-
401- xwc.x = serverGeometry.x () - serverInput.left;
402- xwc.y = serverGeometry.y () - serverInput.top;
403- xwc.width = serverGeometry.width () + serverInput.left + serverInput.right + bw;
404- if (shaded)
405- xwc.height = serverInput.top + serverInput.bottom + bw;
406- else
407- xwc.height = serverGeometry.height () + serverInput.top + serverInput.bottom + bw;
408-
409- if (shaded)
410- height = serverInput.top + serverInput.bottom;
411-
412- if (serverFrameGeometry.x () == xwc.x)
413- valueMask &= ~(CWX);
414- else
415- serverFrameGeometry.setX (xwc.x);
416-
417- if (serverFrameGeometry.y () == xwc.y)
418- valueMask &= ~(CWY);
419- else
420- serverFrameGeometry.setY (xwc.y);
421-
422- if (serverFrameGeometry.width () == xwc.width)
423- valueMask &= ~(CWWidth);
424- else
425- serverFrameGeometry.setWidth (xwc.width);
426-
427- if (serverFrameGeometry.height () == xwc.height)
428- valueMask &= ~(CWHeight);
429- else
430- serverFrameGeometry.setHeight (xwc.height);
431-
432- /* Geometry is the same, so we're not going to get a ConfigureNotify
433- * event when the window is configured, which means that other plugins
434- * won't know that the client, frame and wrapper windows got shifted
435- * around (and might result in display corruption, eg in OpenGL */
436- if (valueMask == 0)
437- {
438- XConfigureEvent xev;
439- XWindowAttributes attrib;
440- unsigned int nchildren = 0;
441- Window rootRet = 0, parentRet = 0;
442- Window *children = NULL;
443-
444- xev.type = ConfigureNotify;
445- xev.event = screen->root ();
446- xev.window = priv->serverFrame;
447-
448- XGrabServer (screen->dpy ());
449-
450- if (XGetWindowAttributes (screen->dpy (), priv->serverFrame, &attrib))
451- {
452- xev.x = attrib.x;
453- xev.y = attrib.y;
454- xev.width = attrib.width;
455- xev.height = attrib.height;
456- xev.border_width = attrib.border_width;
457- xev.above = None;
458-
459- /* We need to ensure that the stacking order is
460- * based on the current server stacking order so
461- * find the sibling to this window's frame in the
462- * server side stack and stack above that */
463- XQueryTree (screen->dpy (), screen->root (), &rootRet, &parentRet, &children, &nchildren);
464-
465- if (nchildren)
466- {
467- for (unsigned int i = 0; i < nchildren; i++)
468- {
469- if (i + 1 == nchildren ||
470- children[i + 1] == ROOTPARENT (window))
471- {
472- xev.above = children[i];
473- break;
474- }
475- }
476- }
477-
478- if (children)
479- XFree (children);
480-
481- if (!xev.above)
482- xev.above = (window->serverPrev) ? ROOTPARENT (window->serverPrev) : None;
483-
484- xev.override_redirect = priv->attrib.override_redirect;
485-
486- }
487-
488- compiz::X11::PendingEvent::Ptr pc =
489- boost::shared_static_cast<compiz::X11::PendingEvent> (compiz::X11::PendingConfigureEvent::Ptr (
490- new compiz::X11::PendingConfigureEvent (
491- screen->dpy (), serverFrame, valueMask, &xwc)));
492-
493- pendingConfigures.add (pc);
494- if (priv->mClearCheckTimeout.active ())
495- priv->mClearCheckTimeout.stop ();
496- priv->mClearCheckTimeout.start (boost::bind (&PrivateWindow::checkClear, priv),
497- 2000, 2500);
498-
499- XSendEvent (screen->dpy (), screen->root (), false,
500- SubstructureNotifyMask, (XEvent *) &xev);
501-
502- XUngrabServer (screen->dpy ());
503- XSync (screen->dpy (), false);
504- }
505- else
506- {
507- compiz::X11::PendingEvent::Ptr pc =
508- boost::shared_static_cast<compiz::X11::PendingEvent> (compiz::X11::PendingConfigureEvent::Ptr (
509- new compiz::X11::PendingConfigureEvent (
510- screen->dpy (), serverFrame, valueMask, &xwc)));
511-
512- pendingConfigures.add (pc);
513- if (priv->mClearCheckTimeout.active ())
514- priv->mClearCheckTimeout.stop ();
515- priv->mClearCheckTimeout.start (boost::bind (&PrivateWindow::checkClear, priv),
516- 2000, 2500);
517- XConfigureWindow (screen->dpy (), serverFrame, valueMask, &xwc);
518- }
519-
520- if (shaded)
521- {
522- XUnmapWindow (screen->dpy (), wrapper);
523- }
524- else
525- {
526- XMapWindow (screen->dpy (), wrapper);
527- XMoveResizeWindow (screen->dpy (), wrapper, serverInput.left, serverInput.top,
528- serverGeometry.width (), serverGeometry.height ());
529- }
530- XMoveResizeWindow (screen->dpy (), id, 0, 0,
531- serverGeometry.width (), serverGeometry.height ());
532- window->sendConfigureNotify ();
533- window->windowNotify (CompWindowNotifyFrameUpdate);
534- }
535- else
536- {
537- int bw = serverGeometry.border () * 2;
538-
539- xwc.x = serverGeometry.x ();
540- xwc.y = serverGeometry.y ();
541- xwc.width = serverGeometry.width () + bw;
542-
543- /* FIXME: It doesn't make much sense to allow undecorated windows to be
544- * shaded */
545- if (shaded)
546- xwc.height = bw;
547- else
548- xwc.height = serverGeometry.height () + bw;
549-
550- if (serverFrameGeometry.x () == xwc.x)
551- valueMask &= ~(CWX);
552- else
553- serverFrameGeometry.setX (xwc.x);
554-
555- if (serverFrameGeometry.y () == xwc.y)
556- valueMask &= ~(CWY);
557- else
558- serverFrameGeometry.setY (xwc.y);
559-
560- if (serverFrameGeometry.width () == xwc.width)
561- valueMask &= ~(CWWidth);
562- else
563- serverFrameGeometry.setWidth (xwc.width);
564-
565- if (serverFrameGeometry.height () == xwc.height)
566- valueMask &= ~(CWHeight);
567- else
568- serverFrameGeometry.setHeight (xwc.height);
569-
570- /* Geometry is the same, so we're not going to get a ConfigureNotify
571- * event when the window is configured, which means that other plugins
572- * won't know that the client, frame and wrapper windows got shifted
573- * around (and might result in display corruption, eg in OpenGL */
574- if (valueMask == 0)
575- {
576- XConfigureEvent xev;
577- XWindowAttributes attrib;
578- unsigned int nchildren = 0;
579- Window rootRet = 0, parentRet = 0;
580- Window *children = NULL;
581-
582- xev.type = ConfigureNotify;
583- xev.event = screen->root ();
584- xev.window = priv->serverFrame;
585-
586- XGrabServer (screen->dpy ());
587-
588- if (XGetWindowAttributes (screen->dpy (), priv->serverFrame, &attrib))
589- {
590- xev.x = attrib.x;
591- xev.y = attrib.y;
592- xev.width = attrib.width;
593- xev.height = attrib.height;
594- xev.border_width = attrib.border_width;
595- xev.above = None;
596-
597- /* We need to ensure that the stacking order is
598- * based on the current server stacking order so
599- * find the sibling to this window's frame in the
600- * server side stack and stack above that */
601- XQueryTree (screen->dpy (), screen->root (), &rootRet, &parentRet, &children, &nchildren);
602-
603- if (nchildren)
604- {
605- for (unsigned int i = 0; i < nchildren; i++)
606- {
607- if (i + 1 == nchildren ||
608- children[i + 1] == ROOTPARENT (window))
609- {
610- xev.above = children[i];
611- break;
612- }
613- }
614- }
615-
616- if (children)
617- XFree (children);
618-
619- if (!xev.above)
620- xev.above = (window->serverPrev) ? ROOTPARENT (window->serverPrev) : None;
621-
622- xev.override_redirect = priv->attrib.override_redirect;
623-
624- }
625-
626- compiz::X11::PendingEvent::Ptr pc =
627- boost::shared_static_cast<compiz::X11::PendingEvent> (compiz::X11::PendingConfigureEvent::Ptr (
628- new compiz::X11::PendingConfigureEvent (
629- screen->dpy (), serverFrame, valueMask, &xwc)));
630-
631- pendingConfigures.add (pc);
632- if (priv->mClearCheckTimeout.active ())
633- priv->mClearCheckTimeout.stop ();
634- priv->mClearCheckTimeout.start (boost::bind (&PrivateWindow::checkClear, priv),
635- 2000, 2500);
636-
637- XSendEvent (screen->dpy (), screen->root (), false,
638- SubstructureNotifyMask, (XEvent *) &xev);
639-
640- XUngrabServer (screen->dpy ());
641- XSync (screen->dpy (), false);
642- }
643- else
644- {
645- compiz::X11::PendingEvent::Ptr pc =
646- boost::shared_static_cast<compiz::X11::PendingEvent> (compiz::X11::PendingConfigureEvent::Ptr (
647- new compiz::X11::PendingConfigureEvent (
648- screen->dpy (), serverFrame, valueMask, &xwc)));
649-
650- pendingConfigures.add (pc);
651- if (priv->mClearCheckTimeout.active ())
652- priv->mClearCheckTimeout.stop ();
653- priv->mClearCheckTimeout.start (boost::bind (&PrivateWindow::checkClear, priv),
654- 2000, 2500);
655-
656- XConfigureWindow (screen->dpy (), serverFrame, valueMask, &xwc);
657- }
658-
659- if (shaded)
660- {
661- XUnmapWindow (screen->dpy (), wrapper);
662- }
663- else
664- {
665- XMapWindow (screen->dpy (), wrapper);
666- XMoveResizeWindow (screen->dpy (), wrapper, 0, 0,
667- serverGeometry.width (), serverGeometry.height ());
668- }
669-
670- XMoveResizeWindow (screen->dpy (), id, 0, 0,
671- serverGeometry.width (), serverGeometry.height ());
672- window->sendConfigureNotify ();
673- window->windowNotify (CompWindowNotifyFrameUpdate);
674- }
675+ xwc.x = serverGeometry.x ();
676+ xwc.y = serverGeometry.y ();
677+ xwc.width = serverGeometry.width ();
678+ xwc.height = serverGeometry.height ();
679+ xwc.border_width = serverGeometry.border ();
680+
681+ window->configureXWindow (valueMask, &xwc);
682+ window->windowNotify (CompWindowNotifyFrameUpdate);
683 window->recalcActions ();
684 }
685
686@@ -1138,8 +860,8 @@
687
688 for (unsigned int i = 0; i < n; i++)
689 {
690- x1 = rects[i].x + priv->geometry.border ();
691- y1 = rects[i].y + priv->geometry.border ();
692+ x1 = rects[i].x + priv->serverGeometry.border ();
693+ y1 = rects[i].y + priv->serverGeometry.border ();
694 x2 = x1 + rects[i].width;
695 y2 = y1 + rects[i].height;
696
697@@ -1147,17 +869,17 @@
698 x1 = 0;
699 if (y1 < 0)
700 y1 = 0;
701- if (x2 > priv->width)
702- x2 = priv->width;
703- if (y2 > priv->height)
704- y2 = priv->height;
705+ if (x2 > priv->serverGeometry.width ())
706+ x2 = priv->serverGeometry.width ();
707+ if (y2 > priv->serverGeometry.height ())
708+ y2 = priv->serverGeometry.height ();
709
710 if (y1 < y2 && x1 < x2)
711 {
712- x1 += priv->geometry.x ();
713- y1 += priv->geometry.y ();
714- x2 += priv->geometry.x ();
715- y2 += priv->geometry.y ();
716+ x1 += priv->serverGeometry.x ();
717+ y1 += priv->serverGeometry.y ();
718+ x2 += priv->serverGeometry.x ();
719+ y2 += priv->serverGeometry.y ();
720
721 ret += CompRect (x1, y1, x2 - x1, y2 - y1);
722 }
723@@ -1185,17 +907,19 @@
724 {
725 int order;
726
727+ /* We should update the server here */
728+ XSync (screen->dpy (), false);
729+
730 boundingShapeRects = XShapeGetRectangles (screen->dpy (), priv->id,
731 ShapeBounding, &nBounding, &order);
732 inputShapeRects = XShapeGetRectangles (screen->dpy (), priv->id,
733 ShapeInput, &nInput, &order);
734-
735 }
736
737- r.x = -priv->geometry.border ();
738- r.y = -priv->geometry.border ();
739- r.width = priv->width + priv->geometry.border ();
740- r.height = priv->height + priv->geometry.border ();
741+ r.x = -priv->serverGeometry.border ();
742+ r.y = -priv->serverGeometry.border ();
743+ r.width = priv->serverGeometry.width () + priv->serverGeometry.border ();
744+ r.height = priv->serverGeometry.height () + priv->serverGeometry.border ();
745
746 if (nBounding < 1)
747 {
748@@ -1726,7 +1450,7 @@
749 priv->attrib.map_state = IsUnmapped;
750 priv->invisible = true;
751
752- if (priv->shaded && priv->height)
753+ if (priv->shaded)
754 {
755 priv->updateFrameWindow ();
756 }
757@@ -1794,7 +1518,7 @@
758 }
759
760 bool
761-CompWindow::resize (XWindowAttributes attr)
762+CompWindow::resize (const XWindowAttributes &attr)
763 {
764 return resize (Geometry (attr.x, attr.y, attr.width, attr.height,
765 attr.border_width));
766@@ -1811,7 +1535,7 @@
767 }
768
769 bool
770-CompWindow::resize (CompWindow::Geometry gm)
771+PrivateWindow::resize (const CompWindow::Geometry &gm)
772 {
773 /* Input extents are now the last thing sent
774 * from the server. This might not work in some
775@@ -1828,12 +1552,8 @@
776 priv->geometry.height () != gm.height () ||
777 priv->geometry.border () != gm.border ())
778 {
779- int pw, ph;
780 int dx, dy, dwidth, dheight;
781
782- pw = gm.width () + gm.border () * 2;
783- ph = gm.height () + gm.border () * 2;
784-
785 dx = gm.x () - priv->geometry.x ();
786 dy = gm.y () - priv->geometry.y ();
787 dwidth = gm.width () - priv->geometry.width ();
788@@ -1843,37 +1563,60 @@
789 gm.width (), gm.height (),
790 gm.border ());
791
792- priv->width = pw;
793- priv->height = ph;
794+ if (priv->attrib.override_redirect)
795+ {
796+ priv->serverGeometry = priv->geometry;
797+ priv->serverFrameGeometry = priv->frameGeometry;
798+ }
799
800 if (priv->mapNum)
801 priv->updateRegion ();
802
803- resizeNotify (dx, dy, dwidth, dheight);
804+ window->resizeNotify (dx, dy, dwidth, dheight);
805
806 priv->invisible = priv->isInvisible ();
807 }
808 else if (priv->geometry.x () != gm.x () || priv->geometry.y () != gm.y ())
809 {
810- int dx, dy;
811-
812- dx = gm.x () - priv->geometry.x ();
813- dy = gm.y () - priv->geometry.y ();
814-
815- priv->geometry.setX (gm.x ());
816- priv->geometry.setY (gm.y ());
817-
818- priv->region.translate (dx, dy);
819- priv->inputRegion.translate (dx, dy);
820- if (!priv->frameRegion.isEmpty ())
821- priv->frameRegion.translate (dx, dy);
822-
823- priv->invisible = priv->isInvisible ();
824-
825- moveNotify (dx, dy, true);
826+ move (gm.x () - priv->geometry.x (),
827+ gm.y () - priv->geometry.y (), true);
828 }
829
830- updateFrameRegion ();
831+ window->updateFrameRegion ();
832+
833+ return true;
834+}
835+
836+bool
837+PrivateWindow::resize (const XWindowAttributes &attr)
838+{
839+ return resize (CompWindow::Geometry (attr.x, attr.y, attr.width, attr.height,
840+ attr.border_width));
841+}
842+
843+bool
844+PrivateWindow::resize (int x,
845+ int y,
846+ int width,
847+ int height,
848+ int border)
849+{
850+ return resize (CompWindow::Geometry (x, y, width, height, border));
851+}
852+
853+bool
854+CompWindow::resize (const CompWindow::Geometry &gm)
855+{
856+ XWindowChanges xwc = XWINDOWCHANGES_INIT;
857+ unsigned int valueMask = CWX | CWY | CWWidth | CWHeight | CWBorderWidth;
858+
859+ xwc.x = gm.x ();
860+ xwc.y = gm.y ();
861+ xwc.width = gm.width ();
862+ xwc.height = gm.height ();
863+ xwc.border_width = gm.border ();
864+
865+ configureXWindow (valueMask, &xwc);
866
867 return true;
868 }
869@@ -2041,13 +1784,7 @@
870 ce->border_width);
871 else
872 {
873- if (ce->override_redirect)
874- {
875- priv->serverGeometry.set (ce->x, ce->y, ce->width, ce->height,
876- ce->border_width);
877- }
878-
879- window->resize (ce->x, ce->y, ce->width, ce->height, ce->border_width);
880+ resize (ce->x, ce->y, ce->width, ce->height, ce->border_width);
881 }
882
883 if (ce->event == screen->root ())
884@@ -2128,7 +1865,7 @@
885 if (priv->syncWait)
886 priv->syncGeometry.set (x, y, width, height, ce->border_width);
887 else
888- window->resize (x, y, width, height, ce->border_width);
889+ resize (x, y, width, height, ce->border_width);
890
891 if (priv->restack (ce->above))
892 priv->updatePassiveButtonGrabs ();
893@@ -2137,27 +1874,6 @@
894
895 if (above)
896 above->priv->updatePassiveButtonGrabs ();
897-
898- if (!pendingConfigures.pending ())
899- {
900- /* Tell plugins its ok to start doing stupid things again but
901- * obviously FIXME */
902- CompOption::Vector options;
903- CompOption::Value v;
904-
905- options.push_back (CompOption ("window", CompOption::TypeInt));
906- v.set ((int) id);
907- options.back ().set (v);
908- options.push_back (CompOption ("active", CompOption::TypeInt));
909- v.set ((int) 0);
910- options.back ().set (v);
911-
912- /* Notify other plugins that it is unsafe to change geometry or serverGeometry
913- * FIXME: That API should not be accessible to plugins, this is a hack to avoid
914- * breaking ABI */
915-
916- screen->handleCompizEvent ("core", "lock_position", options);
917- }
918 }
919
920 void
921@@ -2180,23 +1896,34 @@
922 {
923 if (dx || dy)
924 {
925- gettimeofday (&priv->lastGeometryUpdate, NULL);
926-
927- /* Don't allow window movement to overwrite working geometries
928- * last received from the server if we know there are pending
929- * ConfigureNotify events on this window. That's a clunky workaround
930- * and a FIXME in any case, however, until we can break the API
931- * and remove CompWindow::move, this will need to be the case */
932-
933- if (!priv->pendingConfigures.pending ())
934+ XWindowChanges xwc = XWINDOWCHANGES_INIT;
935+ unsigned int valueMask = CWX | CWY;
936+
937+ xwc.x = priv->serverGeometry.x () + dx;
938+ xwc.y = priv->serverGeometry.y () + dy;
939+
940+ priv->nextMoveImmediate = immediate;
941+
942+ configureXWindow (valueMask, &xwc);
943+ }
944+}
945+
946+void
947+PrivateWindow::move (int dx,
948+ int dy,
949+ bool immediate)
950+{
951+ if (dx || dy)
952+ {
953+ priv->geometry.setX (priv->geometry.x () + dx);
954+ priv->geometry.setY (priv->geometry.y () + dy);
955+ priv->frameGeometry.setX (priv->frameGeometry.x () + dx);
956+ priv->frameGeometry.setY (priv->frameGeometry.y () + dy);
957+
958+ if (priv->attrib.override_redirect)
959 {
960- priv->geometry.setX (priv->geometry.x () + dx);
961- priv->geometry.setY (priv->geometry.y () + dy);
962- priv->frameGeometry.setX (priv->frameGeometry.x () + dx);
963- priv->frameGeometry.setY (priv->frameGeometry.y () + dy);
964-
965- priv->pendingPositionUpdates = true;
966-
967+ priv->serverGeometry = priv->geometry;
968+ priv->serverFrameGeometry = priv->frameGeometry;
969 priv->region.translate (dx, dy);
970 priv->inputRegion.translate (dx, dy);
971 if (!priv->frameRegion.isEmpty ())
972@@ -2204,19 +1931,7 @@
973
974 priv->invisible = priv->isInvisible ();
975
976- moveNotify (dx, dy, immediate);
977- }
978- else
979- {
980- XWindowChanges xwc = XWINDOWCHANGES_INIT;
981- unsigned int valueMask = CWX | CWY;
982- compLogMessage ("core", CompLogLevelDebug, "pending configure notifies on 0x%x, "\
983- "moving window asyncrhonously!", (unsigned int) priv->serverId);
984-
985- xwc.x = priv->serverGeometry.x () + dx;
986- xwc.y = priv->serverGeometry.y () + dy;
987-
988- configureXWindow (valueMask, &xwc);
989+ window->moveNotify (dx, dy, true);
990 }
991 }
992 }
993@@ -2227,22 +1942,6 @@
994 return !mEvents.empty ();
995 }
996
997-bool
998-PrivateWindow::checkClear ()
999-{
1000- if (pendingConfigures.pending ())
1001- {
1002- /* FIXME: This is a hack to avoid performance regressions
1003- * and must be removed in 0.9.6 */
1004- compLogMessage ("core", CompLogLevelWarn, "failed to receive ConfigureNotify event on 0x%x\n",
1005- id);
1006- pendingConfigures.dump ();
1007- pendingConfigures.clear ();
1008- }
1009-
1010- return false;
1011-}
1012-
1013 void
1014 compiz::X11::PendingEventQueue::add (PendingEvent::Ptr p)
1015 {
1016@@ -2464,21 +2163,6 @@
1017 mValueMask (valueMask),
1018 mXwc (*xwc)
1019 {
1020- CompOption::Vector options;
1021- CompOption::Value v;
1022-
1023- options.push_back (CompOption ("window", CompOption::TypeInt));
1024- v.set ((int) w);
1025- options.back ().set (v);
1026- options.push_back (CompOption ("active", CompOption::TypeInt));
1027- v.set ((int) 1);
1028- options.back ().set (v);
1029-
1030- /* Notify other plugins that it is unsafe to change geometry or serverGeometry
1031- * FIXME: That API should not be accessible to plugins, this is a hack to avoid
1032- * breaking ABI */
1033-
1034- screen->handleCompizEvent ("core", "lock_position", options);
1035 }
1036
1037 compiz::X11::PendingConfigureEvent::~PendingConfigureEvent ()
1038@@ -2488,57 +2172,6 @@
1039 void
1040 CompWindow::syncPosition ()
1041 {
1042- gettimeofday (&priv->lastConfigureRequest, NULL);
1043-
1044- unsigned int valueMask = CWX | CWY;
1045- XWindowChanges xwc = XWINDOWCHANGES_INIT;
1046-
1047- if (priv->pendingPositionUpdates && !priv->pendingConfigures.pending ())
1048- {
1049- if (priv->serverFrameGeometry.x () == priv->frameGeometry.x ())
1050- valueMask &= ~(CWX);
1051- if (priv->serverFrameGeometry.y () == priv->frameGeometry.y ())
1052- valueMask &= ~(CWY);
1053-
1054- /* Because CompWindow::move can update the geometry last
1055- * received from the server, we must indicate that no values
1056- * changed, because when the ConfigureNotify comes around
1057- * the values are going to be the same. That's obviously
1058- * broken behaviour and worthy of a FIXME, but requires
1059- * larger changes to the window movement system. */
1060- if (valueMask)
1061- {
1062- priv->serverGeometry.setX (priv->geometry.x ());
1063- priv->serverGeometry.setY (priv->geometry.y ());
1064- priv->serverFrameGeometry.setX (priv->frameGeometry.x ());
1065- priv->serverFrameGeometry.setY (priv->frameGeometry.y ());
1066-
1067- xwc.x = priv->serverFrameGeometry.x ();
1068- xwc.y = priv->serverFrameGeometry.y ();
1069-
1070- compiz::X11::PendingEvent::Ptr pc =
1071- boost::shared_static_cast<compiz::X11::PendingEvent> (compiz::X11::PendingConfigureEvent::Ptr (
1072- new compiz::X11::PendingConfigureEvent (
1073- screen->dpy (), priv->serverFrame, 0, &xwc)));
1074-
1075- priv->pendingConfigures.add (pc);
1076-
1077- /* Got 3 seconds to get its stuff together */
1078- if (priv->mClearCheckTimeout.active ())
1079- priv->mClearCheckTimeout.stop ();
1080- priv->mClearCheckTimeout.start (boost::bind (&PrivateWindow::checkClear, priv),
1081- 2000, 2500);
1082- XConfigureWindow (screen->dpy (), ROOTPARENT (this), valueMask, &xwc);
1083-
1084- if (priv->serverFrame)
1085- {
1086- XMoveWindow (screen->dpy (), priv->wrapper,
1087- priv->serverInput.left, priv->serverInput.top);
1088- sendConfigureNotify ();
1089- }
1090- }
1091- priv->pendingPositionUpdates = false;
1092- }
1093 }
1094
1095 bool
1096@@ -3339,14 +2972,23 @@
1097 {
1098 unsigned int frameValueMask = 0;
1099
1100- /* Immediately sync window position
1101- * if plugins were updating w->geometry () directly
1102- * in order to avoid a race condition */
1103-
1104- window->syncPosition ();
1105+ if (id == screen->root ())
1106+ {
1107+ compLogMessage ("core", CompLogLevelWarn, "attempted to reconfigure root window");
1108+ return;
1109+ }
1110
1111 /* Remove redundant bits */
1112
1113+ xwc->x = valueMask & CWX ? xwc->x : serverGeometry.x ();
1114+ xwc->y = valueMask & CWY ? xwc->y : serverGeometry.y ();
1115+ xwc->width = valueMask & CWWidth ? xwc->width : serverGeometry.width ();
1116+ xwc->height = valueMask & CWHeight ? xwc->height : serverGeometry.height ();
1117+ xwc->border_width = valueMask & CWBorderWidth ? xwc->border_width : serverGeometry.border ();
1118+
1119+ /* FIXME: This is a total fallacy for the reparenting case
1120+ * at least since the client doesn't actually move here, it only
1121+ * moves within the frame */
1122 if (valueMask & CWX && serverGeometry.x () == xwc->x)
1123 valueMask &= ~(CWX);
1124
1125@@ -3413,18 +3055,15 @@
1126 compLogMessage ("core", CompLogLevelWarn, "restack_mode not Above");
1127 }
1128
1129- frameValueMask = valueMask;
1130+ frameValueMask = CWX | CWY | CWWidth | CWHeight | (valueMask & (CWStackMode | CWSibling));
1131
1132- if (frameValueMask & CWX &&
1133- serverFrameGeometry.x () == xwc->x - serverGeometry.border () - serverInput.left)
1134+ if (serverFrameGeometry.x () == xwc->x - serverGeometry.border () - serverInput.left)
1135 frameValueMask &= ~(CWX);
1136
1137- if (frameValueMask & CWY &&
1138- serverFrameGeometry.y () == xwc->y - serverGeometry.border () - serverInput.top)
1139+ if (serverFrameGeometry.y () == xwc->y - serverGeometry.border () - serverInput.top)
1140 frameValueMask &= ~(CWY);
1141
1142- if (frameValueMask & CWWidth &&
1143- serverFrameGeometry.width () == xwc->width + serverGeometry.border () * 2
1144+ if (serverFrameGeometry.width () == xwc->width + serverGeometry.border () * 2
1145 + serverInput.left + serverInput.right)
1146 frameValueMask &= ~(CWWidth);
1147
1148@@ -3434,19 +3073,76 @@
1149
1150 if (shaded)
1151 {
1152- if (frameValueMask & CWHeight &&
1153- serverFrameGeometry.height () == serverGeometry.border () * 2
1154+ if (serverFrameGeometry.height () == serverGeometry.border () * 2
1155 + serverInput.top + serverInput.bottom)
1156 frameValueMask &= ~(CWHeight);
1157 }
1158 else
1159 {
1160- if (frameValueMask & CWHeight &&
1161- serverFrameGeometry.height () == xwc->height + serverGeometry.border () * 2
1162+ if (serverFrameGeometry.height () == xwc->height + serverGeometry.border () * 2
1163 + serverInput.top + serverInput.bottom)
1164 frameValueMask &= ~(CWHeight);
1165 }
1166
1167+ /* Don't allow anything that might generate a BadValue */
1168+ if (valueMask & CWWidth && !xwc->width)
1169+ {
1170+ compLogMessage ("core", CompLogLevelWarn, "Attempted to set < 1 width on a window");
1171+ xwc->width = 1;
1172+ }
1173+
1174+ if (valueMask & CWHeight && !xwc->height)
1175+ {
1176+ compLogMessage ("core", CompLogLevelWarn, "Attempted to set < 1 height on a window");
1177+ xwc->height = 1;
1178+ }
1179+
1180+ if (valueMask & CWStackMode &&
1181+ ((xwc->stack_mode != TopIf) && (xwc->stack_mode != BottomIf) && (xwc->stack_mode != Opposite) &&
1182+ (xwc->stack_mode != Above) && (xwc->stack_mode != Below)))
1183+ {
1184+ compLogMessage ("core", CompLogLevelWarn, "Invalid stack mode %i", xwc->stack_mode);
1185+ valueMask &= ~(CWStackMode | CWSibling);
1186+ }
1187+
1188+ /* Don't allow anything that might cause a BadMatch error */
1189+
1190+ if (valueMask & CWSibling && !(valueMask & CWStackMode))
1191+ {
1192+ compLogMessage ("core", CompLogLevelWarn, "Didn't specify a CWStackMode for CWSibling");
1193+ valueMask &= ~CWSibling;
1194+ }
1195+
1196+ if (valueMask & CWSibling && xwc->sibling == (serverFrame ? serverFrame : id))
1197+ {
1198+ compLogMessage ("core", CompLogLevelWarn, "Can't restack a window relative to itself");
1199+ valueMask &= ~CWSibling;
1200+ }
1201+
1202+ if (valueMask & CWBorderWidth && attrib.c_class == InputOnly)
1203+ {
1204+ compLogMessage ("core", CompLogLevelWarn, "Cannot set border_width of an input_only window");
1205+ valueMask &= ~CWBorderWidth;
1206+ }
1207+
1208+ if (valueMask & CWSibling)
1209+ {
1210+ CompWindow *sibling = screen->findTopLevelWindow (xwc->sibling);
1211+
1212+ if (!sibling)
1213+ {
1214+ compLogMessage ("core", CompLogLevelWarn, "Attempted to restack relative to 0x%x which is "\
1215+ "not a child of the root window or a window compiz owns", static_cast <unsigned int> (xwc->sibling));
1216+ valueMask &= ~(CWSibling | CWStackMode);
1217+ }
1218+ else if (sibling->frame () && xwc->sibling != sibling->frame ())
1219+ {
1220+ compLogMessage ("core", CompLogLevelWarn, "Attempted to restack relative to 0x%x which is "\
1221+ "not a child of the root window", static_cast <unsigned int> (xwc->sibling));
1222+ valueMask &= ~(CWSibling | CWStackMode);
1223+ }
1224+ }
1225+
1226 /* Can't set the border width of frame windows */
1227 frameValueMask &= ~(CWBorderWidth);
1228
1229@@ -3473,11 +3169,8 @@
1230 + serverInput.top + serverInput.bottom);
1231 }
1232
1233-
1234 if (serverFrame)
1235 {
1236- gettimeofday (&lastConfigureRequest, NULL);
1237-
1238 if (frameValueMask)
1239 {
1240 XWindowChanges wc = *xwc;
1241@@ -3493,13 +3186,10 @@
1242 screen->dpy (), priv->serverFrame, frameValueMask, &wc)));
1243
1244 pendingConfigures.add (pc);
1245- if (priv->mClearCheckTimeout.active ())
1246- priv->mClearCheckTimeout.stop ();
1247- priv->mClearCheckTimeout.start (boost::bind (&PrivateWindow::checkClear, priv),
1248- 2000, 2500);
1249
1250 XConfigureWindow (screen->dpy (), serverFrame, frameValueMask, &wc);
1251 }
1252+
1253 valueMask &= ~(CWSibling | CWStackMode);
1254
1255 /* If the frame has changed position (eg, serverInput.top
1256@@ -3697,6 +3387,9 @@
1257 CompWindow::configureXWindow (unsigned int valueMask,
1258 XWindowChanges *xwc)
1259 {
1260+ int dx = valueMask & CWX ? xwc->x - priv->serverGeometry.x () : 0;
1261+ int dy = valueMask & CWY ? xwc->y - priv->serverGeometry.y () : 0;
1262+
1263 if (priv->managed && (valueMask & (CWSibling | CWStackMode)))
1264 {
1265 CompWindowList transients;
1266@@ -3751,6 +3444,16 @@
1267 {
1268 priv->reconfigureXWindow (valueMask, xwc);
1269 }
1270+
1271+ if (!overrideRedirect () && (dx || dy))
1272+ {
1273+ priv->region.translate (dx, dy);
1274+ priv->inputRegion.translate (dx, dy);
1275+ if (!priv->frameRegion.isEmpty ())
1276+ priv->frameRegion.translate (dx, dy);
1277+ moveNotify (dx, dy, priv->nextMoveImmediate);
1278+ priv->nextMoveImmediate = true;
1279+ }
1280 }
1281
1282 int
1283@@ -4302,10 +4005,6 @@
1284 screen->dpy (), serverFrame, valueMask, &lxwc)));
1285
1286 pendingConfigures.add (pc);
1287- if (priv->mClearCheckTimeout.active ())
1288- priv->mClearCheckTimeout.stop ();
1289- priv->mClearCheckTimeout.start (boost::bind (&PrivateWindow::checkClear, priv),
1290- 2000, 2500);
1291 }
1292
1293 /* Below with no sibling puts the window at the bottom
1294@@ -5473,7 +5172,7 @@
1295 m = priv->geometry.x () + offX;
1296 if (m - priv->input.left < (int) s->width () - vWidth)
1297 rv.setX (offX + vWidth);
1298- else if (m + priv->width + priv->input.right > vWidth)
1299+ else if (m + priv->geometry.width () + priv->input.right > vWidth)
1300 rv.setX (offX - vWidth);
1301 else
1302 rv.setX (offX);
1303@@ -5488,7 +5187,7 @@
1304 m = priv->geometry.y () + offY;
1305 if (m - priv->input.top < (int) s->height () - vHeight)
1306 rv.setY (offY + vHeight);
1307- else if (m + priv->height + priv->input.bottom > vHeight)
1308+ else if (m + priv->geometry.height () + priv->input.bottom > vHeight)
1309 rv.setY (offY - vHeight);
1310 else
1311 rv.setY (offY);
1312@@ -6038,7 +5737,7 @@
1313
1314 if (m - priv->output.left < (int) screen->width () - vWidth)
1315 wx = tx + vWidth;
1316- else if (m + priv->width + priv->output.right > vWidth)
1317+ else if (m + priv->geometry.width () + priv->output.right > vWidth)
1318 wx = tx - vWidth;
1319 }
1320
1321@@ -6048,7 +5747,7 @@
1322
1323 if (m - priv->output.top < (int) screen->height () - vHeight)
1324 wy = ty + vHeight;
1325- else if (m + priv->height + priv->output.bottom > vHeight)
1326+ else if (m + priv->geometry.height () + priv->output.bottom > vHeight)
1327 wy = ty - vHeight;
1328 }
1329
1330@@ -6273,9 +5972,6 @@
1331 priv->serverFrameGeometry = priv->frameGeometry = priv->syncGeometry
1332 = priv->geometry = priv->serverGeometry;
1333
1334- priv->width = priv->attrib.width + priv->attrib.border_width * 2;
1335- priv->height = priv->attrib.height + priv->attrib.border_width * 2;
1336-
1337 priv->sizeHints.flags = 0;
1338
1339 priv->recalcNormalHints ();
1340@@ -6297,8 +5993,7 @@
1341
1342 if (priv->attrib.c_class != InputOnly)
1343 {
1344- priv->region = CompRegion (priv->attrib.x, priv->attrib.y,
1345- priv->width, priv->height);
1346+ priv->region = CompRegion (priv->serverGeometry);
1347 priv->inputRegion = priv->region;
1348
1349 /* need to check for DisplayModal state on all windows */
1350@@ -6543,8 +6238,6 @@
1351 hints (NULL),
1352 inputHint (true),
1353 alpha (false),
1354- width (0),
1355- height (0),
1356 region (),
1357 wmType (0),
1358 type (CompWindowTypeUnknownMask),
1359@@ -6579,7 +6272,6 @@
1360 pendingUnmaps (0),
1361 pendingMaps (0),
1362 pendingConfigures (screen->dpy ()),
1363- pendingPositionUpdates (false),
1364
1365 startupId (0),
1366 resName (0),
1367@@ -6781,8 +6473,8 @@
1368 priv->frameRegion &= CompRegion (r);
1369 }
1370
1371- x = priv->geometry.x () - priv->input.left;
1372- y = priv->geometry.y () - priv->input.top;
1373+ x = priv->serverGeometry.x () - priv->input.left;
1374+ y = priv->serverGeometry.y () - priv->input.top;
1375
1376 XShapeCombineRegion (screen->dpy (), priv->serverFrame,
1377 ShapeBounding, -x, -y,
1378@@ -6823,6 +6515,11 @@
1379
1380 priv->updateSize ();
1381 priv->updateFrameWindow ();
1382+
1383+ /* Always send a moveNotify
1384+ * whenever the frame extents update
1385+ * so that plugins can re-position appropriately */
1386+ moveNotify (0, 0, true);
1387 }
1388
1389 /* Use b for _NET_WM_FRAME_EXTENTS here because
1390
1391=== modified file 'src/windowgeometry.cpp'
1392--- src/windowgeometry.cpp 2012-01-23 05:44:19 +0000
1393+++ src/windowgeometry.cpp 2012-04-04 05:26:30 +0000
1394@@ -61,22 +61,22 @@
1395 int
1396 CompWindow::width () const
1397 {
1398- return priv->width +
1399+ return priv->geometry.width () +
1400 priv->geometry.border () * 2;
1401 }
1402
1403 int
1404 CompWindow::height () const
1405 {
1406- return priv->height +
1407+ return priv->geometry.height () +
1408 priv->geometry.border () * 2;;
1409 }
1410
1411 CompSize
1412 CompWindow::size () const
1413 {
1414- return CompSize (priv->width + priv->geometry.border () * 2,
1415- priv->height + priv->geometry.border () * 2);
1416+ return CompSize (priv->geometry.width () + priv->geometry.border () * 2,
1417+ priv->geometry.height () + priv->geometry.border () * 2);
1418 }
1419
1420 int

Subscribers

People subscribed via source and target branches