Compiz

[regression r3729] Plugin classes not destructed properly on unload

Bug #1193596 reported by Sami Jaktholm on 2013-06-22

This bug affects 6 people

Affects		Status	Importance	Assigned to	Milestone
	Compiz	Fix Released	Critical	Sam Spilsbury	Compiz 0.9.10.0
	compiz (Ubuntu)	Fix Released	Undecided	Unassigned

Bug Description

When a plugin is unloaded, the screen and window classes of that plugin aren't properly destructed. This is a regression caused by r3729.

Here's what I've figured out so far:
1) Plugin list changes, CompPlugin::pop is called for plugins
2) CompPlugin::pop calls CompManager::finiPlugin
3) CompManager::finiPlugin calls p->vTable->markNoFurtherInstantiation
4) markNoFurtherInstantiation calls disallowInstantiations for screen (and window)
5) disallowInstantiations sets mPluginLoaded to false
6) CompManager::finiPlugin calls p->vTable->finiScreen
7) CompPlugin::VTable::finiScreen calls PluginClassHandler::get
8) PluginClassHandler::get returns NULL because mPluginLoaded is false
9) CompPlugin::VTable::finiScreen destructs the returned NULL instead of the PluginScreen instance

I'm not sure what happens after (9), but the destructors of PluginClasses are never called.

A crash due to this bug occurs with following steps:
1) Enable Opacify plugin
2) Activate it if not automatically activated
3) Disable Opacify plugin
4) Crash.

Related branches

lp:~compiz-team/compiz/compiz.fix_1193596

Merged into lp:compiz/0.9.10 at revision 3743

PS Jenkins bot (community): Approve (continuous-integration) on 2013-06-23

Sami Jaktholm (community): Approve on 2013-06-23

MC Return: Approve on 2013-06-23

lp:ubuntu/saucy-proposed/compiz

lp:ubuntu/trusty-proposed/compiz

Revision history for this message

Sami Jaktholm (sjakthol) wrote on 2013-06-22:

A possible fix is to call markNoFurtherInstantiation after finiScreen has completed. However, I'm not sure if is's a good idea to let the destructors (possibly) create new instances of PluginClasses we're trying to destroy.

MC Return (mc-return) on 2013-06-22

Changed in compiz:
milestone:	none → 0.9.10.0

MC Return (mc-return) on 2013-06-22

Changed in compiz:
importance:	Undecided → High

Revision history for this message

Sam Spilsbury (smspillaz) wrote on 2013-06-22:

Actually, you're right - it needs to call markNoFurtherInstantiation after finiScreen is completed. Thanks for finding that.

Sam Spilsbury (smspillaz) on 2013-06-22

Changed in compiz:
assignee:	nobody → Sam Spilsbury (smspillaz)
importance:	High → Critical

Revision history for this message

Sam Spilsbury (smspillaz) wrote on 2013-06-22:

Additional note: allowing the destructors to create new instances isn't a problem - a new instance will only be created if a plugin class instance didn't already exist for that particular core structure.

Calling PluginWindow::get (w) inside of a PluginScreen destructor is more or less undesired behavior anyways - previously it would just result in ::get returning NULL (which by convention, is never checked), now the behavior is undefined. The proper solution is probably to have a "mark no further instantiations" phase for each attachable class (eg, PluginWindow::markNoFurtherInstantiations, PluginScreen::markNoFurtherInstantiations), but just delaying the marking until the end of the unload process is good enough for now.

Revision history for this message

PS Jenkins bot (ps-jenkins) wrote on 2013-06-23:

Fix committed into lp:compiz at revision None, scheduled for release in compiz, milestone 0.9.10.0

Changed in compiz:
status:	New → Fix Committed

Stephen M. Webb (bregma) on 2013-07-23

Changed in compiz:
status:	Fix Committed → Fix Released

Revision history for this message

Launchpad Janitor (janitor) wrote on 2013-08-22:

Download full text (70.8 KiB)

This bug was fixed in the package compiz - 1:0.9.10+13.10.20130822-0ubuntu1

---------------
compiz (1:0.9.10+13.10.20130822-0ubuntu1) saucy; urgency=low

[ Sam Spilsbury ]
* Bump version to 0.9.10

  [ Łukasz 'sil2100' Zemczak ]
  * Remove debian/patches/unity_support_test.patch:
    - Running the support test from compiz has bad side effects, from now
      on we run it from Xsession.d
  * Automatic snapshot from revision 3644

  [ Iven Hsu ]
  * Opacify: Only dim the windows above the active window.(LP:
    #1189374). (LP: #1189374)
  * KWD: Fix compile errors with KDE 4.11. The KWin developers made
    kdecorationbridge.h private. See:
    http://lists.freedesktop.org/archives/compiz/2013-March/003479.html
    (LP: #1193792). (LP: #1193792)

  [ Nikolay Martynov ]
  * When static switcher is enabled and has an option to show
    application icon turned on the icons are expected to be ~1/3 of a
    thumbnail (48px). Instead they are displayed in 512px size and
    completely cover everything. This change addresses this issue. See
    LP #1173914. (LP: #1173914, #1186426)

  [ BryanFRitt ]
  * Fixed the non-working Annotate 'Clear' Button. Moved this option's
    CCSM position upwards to keep the button shortcuts together. (LP:
    #1202907). (LP: #1202907)

  [ Mehrdad Afshari ]
  * Added "move window to previous monitor" feature to compiz Put
    plugin. (LP: #1178581)

  [ Hu Kang ]
  * gtk-window-decorator: destroy action menu when any of the (close,
    min, max) buttons on the title bar is pressed. (LP: #1101648)
  * Remove redundant src/logmessage/include/core/logmessage.h (LP:
    #1067246). (LP: #1067246)

  [ Steve Langasek ]
  * Fix for bug #763148 (with added test cases): when the desktop is
    resized, windows should stay on their original workspace. (LP:
    #763148)

This bug was fixed in the package compiz - 1:0.9.10+13.10.20130822-0ubuntu1

---------------
compiz (1:0.9.10+13.10.20130822-0ubuntu1) saucy; urgency=low

[ Sam Spilsbury ]
  * Bump version to 0.9.10

[ BryanFRitt ]
  * Fixed the non-working Annotate 'Clear' Button. Moved this option's
    CCSM position upwards to keep the button shortcuts together. (LP:
    #1202907). (LP: #1202907)

[ Mehrdad Afshari ]
  * Added "move window to previous monitor" feature to compiz Put
    plugin. (LP: #1178581)

[ Steve Langasek ]
  * Fix for bug #763148 (with added test cases): when the desktop is
    resized, windows should stay on their original workspace. (LP:
    #763148)

[ Brandon Schaefer ]
  * Unrevert 3728, fix failing tests. Change the behaviour of
    undecorating windows. Previously when a window was undecorated, we
    would shift it back to an appropriate position according to its
    gravity member. That behaviour was problematic because in the
    StaticGravity case the window has to just stay in the same place.
    But then if you had a window with StaticGravity which then did get a
    decoration and later removed it, it would be placed as though it was
    decorated and appear to be in the wrong place. The correct behaviour
    is to place all windows as though they have decorations, and then
    when decorations are removed, to move the window back to the corner
    as indicated in its gravity and then expand its size to cover the
    obscured regions no longer hidden because the decorations went away.
    (LP: #1165343).   1. Completely remove decorOffsetMove and other
    related code from      decor.cpp. Put the logic to handle the
    window->input () - window->border ()      placement offset inside of
    setWindowFrameExtents instead. Now the window      will always be
    offset from its original non-decorated position to the new
         decorated position, rather than having to guess between
    decoration sizes.   2. Make saveGeometry and restoreGeometry work
    relative to window->border ()      as opposed to including it in the
    saved geometry. It is possible that the      border size might
    change during maximization, as such, we don't want to      save the
    position with the border before maximizing. Instead save the
    position      as if it were never decorated so that when the window
    is restored it can be      restored to its original position and
    then adjusted for its new border size.   3. Fix a few typoes in the
    tests.   4. Moved some commonly used matchers into compiz::testing
      5. Make COMPIZ_PLUGIN_DIR accept multiple directories and look in
    each one      of them for the plugin   6. Set COMPIZ_PLUGIN_DIR
    appropriately for each plugin that we wish to load      on startup
    so that we load locally built plugins as opposed to installed
         ones.   7. Uncomment compiz_discover_tests for the acceptance
    tests. Now they are      run by default. (LP: #1195522). (LP:
    #1195522, #1165343)
  * If we receive a stateNotifyChange, we check if we are in a max
    state. If so, overwrite the xwc with either X | W, or Y | H
    depending on the semi max state with the orig position before we
    went into a semi max state. This way when we leave any sort of max
    state the orig position is restored. If the window is being grabbed
    we do not overwrite the position, which it gets its own value at
    that point (instead of the orig pos). (LP: #892012)

[ Marco Trevisan (Treviño) ]
  * OpenGL Screen: Ignore the MSAA configs during initialization This
    fixes a crash with new MESA drivers and high monitor resolutions
    caused by the fact that we're accidentally using some MSAA configs
    with some drivers in Mesa 9.0. Taking in account the values of
    GLX_SAMPLES and GLX_SAMPLE_BUFFERS fixes the issue. See mesa bug
    https://bugs.freedesktop.org/show_bug.cgi?id=61182. (LP: #1174495)
  * OpenGLES Screen: Ignore the MSAA configs during initialization.
  * GLScreen: break the iteration as soon as we've found a valid egl
    config.

[ Michail Bitzes ]
  * Plugin wizard ported to OpenGL|ES. Use GLVertexBuffer. Enable
    building for GLES. Enable architectures armel and armhf. (LP:
    #1196003). (LP: #1196003)

[ Chris Townsend ]
  * Fixed issue where minimizing an unredirect full screen window would
    repaint the full screen window after minimizing it even though it
    isn't really there. . (LP: #1053895)

[ Sami Jaktholm ]
  * *Grid code: Prevent center and corner gridded windows from jumping
    viewports. (Thanks and credits for this go to Sami Jaktholm) Prevent
    top and bottom gridded windows from jumping viewports by making
    those semi-maximize horizontally. As those are actually semi-
    maximized horizontally, we will treat them as such and let core
    handle the restoring, just like we already do for vertically semi-
    maximized grid windows (left/right). Now "Strg+Super+Down" will
    restore top and bottom gridded windows correctly as well. Also
    multiple gridding to top, bottom, left or right will not overwrite
    the stored original size anymore. Restore windows also when
    workspace switcher (expo) is active. (Thanks and credits for this go
    to Sami Jaktholm) Allow cycling for all gridded windows if
    explicitly specified by the user in CCSM. Forbid cycling through
    different sizes for corner and center-gridded windows also per
    default, now fully fixing bug #878820 and following the design
    specification by Ayatana Design there, making behaviour consistent.
    Introduced 3 new bools: horzMaximizedGridPosition,
    vertMaximizedGridPosition, anyMaximizedGridPosition Used these bools
    inside the if condition checks. Simplified complicated if condition
    by removing redundant additional size check. Cleanup all around.
    *Grid xml: Added cycle_sizes bool option, which allows the user to
    choose the prefered behaviour (fixed versus flexible sizes on
    multiple presses on the same grid keyboard shortcut). Default of
    this option is off, cycling disabled as specified by design. Added
    punctuation where missing and removed it where usually is none.
    Fixed typos, improved description and titles. *Expo code: Tell grid
    when viewport change is in progress. (Thanks and credits for this go
    to Sami Jaktholm) (fixes: LP: #878820, LP: #879218, LP: #882754 and
    LP: #1082001, partially fixes: #1116538, #1164332). (LP: #882754,
    #878820, #1164332, #1082001, #879218, #1116538)
  * Grid: Check that composite and opengl ABIs are COMPIZ_COMPOSITE_ABI
    and COMPIZ_OPENGL_ABI instead of CORE_ABIVERSION. (LP: #1181461)
  * Fade: Fix a typo that causes fade to constantly damage windows with
    a modified saturation. (LP: #1159054)
  * Grid: Fix window misplacement if a window is restored but the size
    doesn't change. - Set isGridResized to false before configuring the
    window in GridScreen::restoreWindow. If the restored geometry is the
    same as current geometry, configureXWindow will trigger moveNotify
    instead of resizeNotify. As isGridResized is still true,
    GridWindow::moveNotify blocks the restoration movement and the
    pointer detaches from the decoration. - Don't use pointerBufs when
    calculating position of the restored window with snapback windows
    disabled. The window will be positioned a bit too high (or low) and
    the pointer detaches from decoration if the movement was actually
    tracked. This was a workaround for the first point and can be safely
    removed. (LP: #1115344)
  * Fix memory leaks in titleinfo plugin. (LP: #1101569, #1101454)
  * Expo plugin: 1) Always fetch the dndButton action in dndFini. The
    action given to dndFini might be a key or edge binding action if
    expo was terminated during window DnD. We shouldn't set their state
    to CompAction::StateInitButton or the bindings break. 2) Pass NULL
    action to termExpo if invoked from handleEvent Previously we passed
    the action of expoKey to termExpo which in turn passed it to
    dndFini. DndFini changed the action state to
    CompAction::StateInitButton that broke the keybinding. Now that
    dndFini fetches the dndButton action itself, there's no need to
    fetch it in handleEvent too as dndFini is the only one using the
    action. (LP: #1131106)
  * Expo: Cast denominators of integer divisions to floats before
    performing a matrix translation to viewports with the results.
    Otherwise rounding errors in cases the results are not an integers
    cause the viewports to be painted offscreen. (LP: #438580)
  * Port dbus introspection to compiz 0.9. - move xml creation to a
    separate class (IntrospectionResponse) for easier memory management
    (allocate buffer and writer in ctor, free in dtor). - move
    duplicated response sending code to a separate method that takes
    IntrospectionResponse and sends the resulting xml. - Refactor
    handle*IntrospectMessage to work with compiz 0.9 interfaces. This
    also fixes the broken list method which was a result of logic error.
    The code to invoke list handler was never reached. This fixes most
    of the issues noted in bug 749084. (LP: #749084)
  * Bias the current viewport in addWindowSizeChanges. If window is
    visible in the current viewport, use that viewport when calculating
    geometries in addWindowSizeChanges. Otherwise the current method is
    used. This way visible windows are maximized in the current
    viewport. However, offscreen windows won't jump back to the active
    viewport if window maximizes by itself or addWindowSizeChanges is
    called for some other reason. (LP: #776435)
  * Expo: Recompute glow quads if a desktop window is resized. (LP:
    #1090713). (LP: #1090713)
  * Decor: Use maximized border extents only if window is fully
    maximized. The decorator draws a normal border around semi-maximized
    windows. When maximized border extents were used for semi-maximized
    windows, compiz didn't reserve any space for the border in its
    geometry calculations. At least following problems are a result of
    this behavior: - Semi-maximized windows have 1px borders drawn on
    adjacent workspaces (LP: #986051). - Grid placed window overlaps the
    adjacent viewport (LP: #898870). (LP: #986051, #898870)

[ gebner@2b7e.org ]
  * Adding a uniform to a GLVertexBuffer that uses AutoProgram causes
    compiz to segfault. Example: gWindow->addShaders("cms", "",
    fragment_shader); gWindow->vertexBuffer()->addUniform("cms_lut",
    unit); // segfault happens later in PrivateVertexBuffer::render The
    patch modifies PrivateVertexBuffer::render to set the uniform on the
    generated AutoProgram instead of the provided program, which in this
    case is NULL, causing a segfault. (LP: #1162598). (LP: #1162598)

[ MC Return ]
  * Cube-addon (Cube Reflection and Deformation): If the user disables
    "Draw top face"/"Draw bottom face" we do not want to draw anything
    (LP: #1162484). The original, non-deformed caps will only work for
    the non-deformed cube, so we can just use the original function in
    that case. We need to clear the texture if no texture files are
    specified in "Image files", otherwise the old texture would still be
    drawn, even if all image files are removed (LP: #1162711). Now we
    will default back and use the cube cap colors and opacities defined
    in the "Desktop Cube" plugin (if "Draw top/bottom face" are enabled
    only, see above). This way the user can choose between (for
    top/bottom): 1. Do not draw a cube cap face at all ("Draw top/bottom
    face" option disabled) 2. Use color and opacity specified in
    "Desktop Cube" (empty images list) 3. Use a texture for the cap
    (image is in the list, which is default) Removed redundant mCurrent
    = mCurrent % mFiles.size (); calculation, this has already been
    done: cap->mCurrent = (cap->mCurrent + change + count) % count;
    count and change both need to be != 0 for mCurrent to change. Fixed
    indentation, removed redundant brackets and newlines, declaration
    and assignment of local variables in one line, if possible, minor
    cleanup. (LP: #1162484, LP: #1162711). (LP: #1162484, #1162711)
  * Stack (Window) Switcher (currently excluded from compilation): Fixed
    window title drawn although "Show Window Title" is disabled. (LP:
    #1070782). (LP: #1070782)
  * *Grid code: Prevent center and corner gridded windows from jumping
    viewports. (Thanks and credits for this go to Sami Jaktholm) Prevent
    top and bottom gridded windows from jumping viewports by making
    those semi-maximize horizontally. As those are actually semi-
    maximized horizontally, we will treat them as such and let core
    handle the restoring, just like we already do for vertically semi-
    maximized grid windows (left/right). Now "Strg+Super+Down" will
    restore top and bottom gridded windows correctly as well. Also
    multiple gridding to top, bottom, left or right will not overwrite
    the stored original size anymore. Restore windows also when
    workspace switcher (expo) is active. (Thanks and credits for this go
    to Sami Jaktholm) Allow cycling for all gridded windows if
    explicitly specified by the user in CCSM. Forbid cycling through
    different sizes for corner and center-gridded windows also per
    default, now fully fixing bug #878820 and following the design
    specification by Ayatana Design there, making behaviour consistent.
    Introduced 3 new bools: horzMaximizedGridPosition,
    vertMaximizedGridPosition, anyMaximizedGridPosition Used these bools
    inside the if condition checks. Simplified complicated if condition
    by removing redundant additional size check. Cleanup all around.
    *Grid xml: Added cycle_sizes bool option, which allows the user to
    choose the prefered behaviour (fixed versus flexible sizes on
    multiple presses on the same grid keyboard shortcut). Default of
    this option is off, cycling disabled as specified by design. Added
    punctuation where missing and removed it where usually is none.
    Fixed typos, improved description and titles. *Expo code: Tell grid
    when viewport change is in progress. (Thanks and credits for this go
    to Sami Jaktholm) (fixes: LP: #878820, LP: #879218, LP: #882754 and
    LP: #1082001, partially fixes: #1116538, #1164332). (LP: #882754,
    #878820, #1164332, #1082001, #879218, #1116538)
  * *Resizeinfo, xml changes: Added option for bold/normal font, default
    is still bold. Added option to change the font size (10-14), default
    is still 12 pixel. Enhanced and corrected a few tooltips.
    *Resizeinfo, code changes: Choose between PANGO_WEIGHT_BOLD and
    PANGO_WEIGHT_NORMAL. Use individual font size specified by the user
    in CCSM. Fixed computation of wrong damageRegion, it was
    additionally adding the window size, making it way too large.
    Removed useless declaration of int width and height. Declaration and
    assignment of local variables in one line, if possible. Minor
    indentation fixes. (LP: #1166195, LP: #1166196 and LP: #1166245).
    (LP: #1166195, #1166196, #1166245)
  * *Move xml: Implemented options to configure: "Snapoff Distance"
    "Snapback Semimaximized Windows" and "Snapback Distance" Improved a
    few tooltips. *Move code: Replaced SNAP_BACK and SNAP_OFF hardcoded
    constants and made those configurable. Implemented a strategy to
    snap off horizontally maximized windows by dragging them along the x
    axis. Implemented snapping back of horizontally maximized windows
    and fixed the snapping for vertically maximized windows (wrong
    cursor calculation). Fixed a few wrong calculations in the if
    condition checks responsible for snapping off and back. Merged if
    condition checks. Just compute various local variables if we do not
    return false. Removed redundant brackets, fixed indentation and
    improved readability. (LP: #1165198, LP: #1167933). (LP: #1167208,
    #1167933, #1165198)
  * Freewins xml: Fixed False->false and True->true typos. Credits and
    thanks for finding cause and fix go to raveit65 ! (LP: #1163606).
    (LP: #1163606)
  * Screenshot code cleanup: Used standard bool for mGrab. Merged a few
    if condition checks. Declaration and assignment of local variables
    in one line. Removed redundant brackets. Added and removed newlines
    to improve readability. Added a missing break (just a style issue).
    Fixed indentation. C++ style for comments.
  * *Move xml: Slightly improved long description of the move plugin.
    Implemented option to change the key move increment size. Removed
    <precision> settings as they just work for floats. *Move code:
    Removed hardcoded KEY_MOVE_INC variable. Introduced int
    moveIncrement to not have to execute optionGetKeyMoveInc () to get
    the value twice. (LP: #1167933, part 2). (LP: #1167933)
  * Shift Switcher code cleanup: Merged a few if-if condition checks.
    Declaration and assignment of local variables in one line. Removed
    redundant brackets. Added and removed newlines to improve
    readability. Fixed indentation.
  * Composite code cleanup: Merged a few if-if condition checks.
    Declaration and assignment of local variables in one line. Removed
    redundant brackets. Added and removed newlines to improve
    readability. Fixed indentation.
  * *Grid code: Simplified bool GridScreen::restoreWindow (...): If Grid
    has not touched the window we return false and let core do the dirty
    work. (Thanks go to Sam Spilsbury for suggesting this solution.)
    *Grid xml: Improved the CCSM Grid put_restore_key tooltip. (LP:
    #1116538). (LP: #1116538)
  * *Cube Gears: Added glDepthMask (GL_TRUE); and glDepthFunc (GL_LESS);
    to void GearsScreen::cubePaintInside (...) to fix the culling on
    multimonitor systems (green gear was always rendered in front). Also
    fixed indentation and removed some redundant newlines. (LP:
    #1168919). (LP: #1168919)
  * *Showmouse code cleanup: Assignment and declaration of local
    variables in one line. Removed redundant brackets. Fixed
    indentation. Removed and added newlines for better readability.
    *Showmouse code changes: Removed these redundant gl calls: glEnable
    (GL_TEXTURE_2D); glDisable (GL_TEXTURE_2D); (partially fixes LP:
    #1051290 and LP: #1051295). (LP: #1051290, #1051295)
  * *Workspacenames: New Feature: Implemented option to adjust vertical
    offset for top and bottom of screen workspacename placement. Set the
    default offset to 50 pixel. Cleanup: Added brackets to for loop. Use
    prefix increments. Indentation fixes. Added and removed newlines for
    better readability. Added comment. Removed TEXT_BORDER constant.
    Improved tooltips. (LP: #1172063). (LP: #1172063)
  * *Screenshot, xml: Better plugin summary. Added option to disable the
    drawing of the screenshot selection indicator. Added options to
    change color and opacity of the outline and the inside of the
    rectangle indicator, default colors have not been changed. Other
    minor improvements. *Screenshot, code: Introduced the new variable
    bool selectionSizeChanged, which is true if the size of the
    selection has changed. We just want to draw the screenshot selection
    box, if we are grabbed, the size has changed and the CCSM option to
    draw it is enabled. Do not render anything if indicator is disabled.
    Indicator colors are now configurable. Increased indicator box
    outline line width to 2.0. Only damage the full screen once, if the
    grab gets terminated, during grab just damage the region painted,
    which is the screenshot selection box (LP: #1172234). *SCreenshot,
    cleanup: Minor cleanup in the rendering code. Removed redundant
    #ifndef USE_GLES compiler option, GLES can cope with enabling and
    disabling GL_BLEND. Use prefix- instead of postfix increments.
    Declaration and assignment of local variables in one line. Reduced
    the scope of the variables x1, y1, x2 and y2. Replaced magic number
    65535.0f with const float MaxUShortFloat = std::numeric_limits
    <unsigned short>::max ();. Indentation fixes. C++ style for
    comments, added comments. (LP: #1169353, LP: #1172234). (LP:
    #1169353, #1172234)
  * Shift Switcher code cleanup: Use prefix instead of postfix increment
    and decrement. Declaration and assignment of local variables in one
    line. Merged if statements. Added brackets to for loops. Fixed
    indentation. Added and removed newlines for better readability.
    Added comment.
  * Cube Addon .xml cleanup: Better plugin description. Uppercase
    titles. Improved tooltips and names.
  * Animation code cleanup: Use prefix- instead of postfix- in- and
    decrements. Removed redundant brackets. Added and removed newlines
    where appropriate. Fixed indentation.
  * Grid, code cleanup: Merged if statements. Replaced magic number
    65535.0f with const float MaxUShortFloat = std::numeric_limits
    <unsigned short>::max ();. Use prefix- instead of postfix-increment.
    Removed redundant (float) casts. Removed redundant brackets. Added
    and removed newlines to improve readability. Minor indentation
    fixes.
  * Workspacenames .xml cleanup: Better description of the plugin's
    functionality. Improved tooltips.
  * Cube .xml cleanup: Better plugin description. Uppercase titles.
    Improved tooltips.
  * Shift Switcher .xml cleanup: Better description of the plugin's
    function. Fixed load before "bs" plugin typo -> it is the obs
    plugin. Created a "Switcher Mode" tab and moved Flip/Cover mode
    settings there. Created a subgroup for mouse-specific settings.
    Better tooltips with enhanced descriptions. Improved titles. Fixes
    all around (typos, etc.).
  * Refactor screenshot code to make it clearer and also to allow us to
    take screenshots in glPaintOutput as opposed to paint (). Taking
    screenshots in paint () was probably correct pre-GLES, but isn't
    really correct now - we want to be able to read off of the currently
    bound scratch framebuffer where we last painted the frame. Reading
    off the frontbuffer is an imprecise operation because the contents
    of both buffers are really undefined after a swap. In the case where
    we are not painting to a scratch framebuffer object, we'll just end
    up reading from the backbuffer anyways, which would be correct to do
    mid-frame. Also added the new static const GLenums
    DRAW_FRAMEBUFFER_BINDING and READ_FRAMEBUFFER_BINDING to opengl.h.
    (LP: #771875). (LP: #771875)
  * shift.h: Minor indentation fixes (mostly replaced tabs with spaces).
  * Show Desktop .xml cleanup: Improved plugin description. Renamed
    "Misc. Options" tab to "General" Improved tooltips.
  * Stack Window Switcher (stackswitch): Declaration and assignment of
    local variables in one line. Merged if condition checks. Use prefix
    de- and increments. Removed redundant brackets. Fixed indentation.
    Note: stackswitch is still excluded from compilation, because it has
    not been ported to the new GLES-compatible modern post-r3320 Compiz
    codebase yet, so this cannot really be tested at the moment (except
    in Compiz pre-r3320). Nevertheless this MP is needed to reduce the
    diff size for the GLES port of this plugin, which will follow soon.
  * Ring Switcher .xml cleanup: Better description of the "Ring
    Switcher". Uppercase for titles. More precise names for the tabs.
    Improved tooltip wording and descriptions. Changed the order of
    options in "General Options" tab: "Ring Windows" and "Overlay Icon"
    options are now on top. No settings have been changed.
  * *Shift Window Switcher, new features: Implemented new option to
    adjust the vertical offset of the text placement and fixed vertical
    placement calculation for top and bottom positions. The default
    offset is 50 pixels now. Removed hardcoded and redundant float
    border. Use 256x256 pixels resolution for icon textures, instead of
    96x96. Minor indentation fixes. (LP: #1173684). (LP: #1173684)
  * Opacity, Brightness and Saturation (obs) .xml cleanup: Better plugin
    description. Improved setting names and tooltips.
  * Opacity, Brightness and Saturation (obs) code cleanup: Replaced
    #define MODIFIER_COUNT 3 with const unsigned short MODIFIER_COUNT =
    3; (the variable MODIFIER_COUNT is needed in obs.h and obs.cpp). Use
    prefix instead of postfix increment. Merged if condition checks.
    Declaration and assignment of local variables in one line. Removed
    redundant brackets. Added and removed newlines if appropriate.
    Indentation fixes.
  * Improved and optimized ABI logic checks consistently for all
    plugins. (LP: #1178228). (LP: #1178228)
  * *Grid, code changes and fixes: Fixed segfault in Compiz when
    unchecking "Snap windows back to original size". This option now
    fully works and will snap back windows to current size, if they are
    dragged off via mouse. Fixed segfault in Compiz when unchecking
    "Snapoff Maximized/Semi-maximized Windows" - this option is now
    removed as this setting is configurable via "Move" plugin already.
    The snapoff threshold global setting is now configurable instead of
    hardcoded. Fixed broken cursor y-coordinate calculation. Fixed
    window placement cursor calculation for non-original-size-restored
    windows. Make sure grid-maximizing will never overwrite the windows'
    original size. Removed useless code until someone can convince me
    that we need it (LP: #1020857). This fixes the weird positioning
    (50, 50) when restoring mouse-grid-maximized windows via keyboard.
    Added missing ABI checks of OpenGL and Composite. *Grid, xml changes
    and fixes: Put Corners/Edges setting into own tab. New tab named
    "Resize Actions". Removed hardcoded SNAPOFF_THRESHOLD and made it
    configurable. Improved tooltips. Re-applied the quilt patch to the
    changed grid.xml.in. Changed the put_restore_key to
    "Ctrl+Super+Down" to fix restoring of grid-maximized windows for
    Ubuntu as well. (LP: #1172641) *Grid, minor code cleanup: Removed
    redundant (int) casts. Improved readability for the if condition
    checks. Added and removed newlines, if appropriate. Added comments.
    Indentation fixes. (LP: #745159, LP: #1020857, LP: #1172641). (LP:
    #1020857, #745159, #1172641)
  * Resize, minor code cleanup: Declaration and assignment of local
    variables in one line. Added and removed newlines, if appropriate.
    Removed redundant brackets. Indentation fixes.
  * Wobbly Windows, code cleanup: Declaration and assignment of local
    variables in one line. Merged if statements. Removed redundant
    brackets. Added missing break. Removed and added newlines for better
    readability. Fixed indentation.
  * Ring, Shift and Stack Window Switcher: Increased the non-power-of-2
    icon texture resolution from 96x96 to 512x512 pixels, in the case of
    the ring switcher from 64 to 512. Static Switcher: Replaced #define
    of MAX_ICON_SIZE with type-safe extern const unsigned short and set
    the icon resolution to 512 as well. Compiztoolbox: Increased the
    non-power-of-2 icon texture resolution const unsigned short
    ICON_SIZE from 48 to 512 and const unsigned int MAX_ICON_SIZE from
    256 to 512 as well. Changed the type of MAX_ICON_SIZE from const
    unsigned int to const unsigned short. Scale: Increased the non-
    power-of-2 icon texture resolution from 96x96 to 512x512 pixels. KDE
    Compatibility: Increased the icon texture resolution from 256x256 to
    512x512 pixels. (LP: #1173914). (LP: #1173914)
  * Removed redundant Expo configuration changes from ubuntu-
    config.patch: The reflection scale change, because reflection is
    disabled by default anyway. The inactive viewport saturation change
    from 100.0% to 40.0%, because this option does not work (and never
    did AFAICR), but if it finally gets fixed Expo's look would
    dramatically change. (LP: #1178965). (LP: #1178965)
  * Expo plugin: Fixed inactive viewports saturation setting being
    ignored. Let's "fade to grey"... :) (LP: #1179183). (LP: #1179183)
  * compizconfig/*, libdecoration/*: Use prefix instead of postfix de-
    and increments.
  * *Ring Switcher, new feature: Implemented "Vertical Offset" option
    with a default of 50 pixels. This allows the CCSM user to easily and
    exactly configure where the window title text should be displayed.
    Ring Switcher, cleanup: Simplified
    RingWindow::compareRingWindowDepth (...). Bail out of functions
    ASAP, do not initialize anything you do not need, if you exit
    anyway. Merged if condition checks. Declaration and assignment of
    local variables in one line. Removed redundant brackets. Added and
    removed newlines, if appropriate. Fixed indentation. (LP: #1173684).
    (LP: #1173684)
  * gtk/*, include/*, kde/*: Use prefix instead of postfix de- and
    increments.
  * Scale Addons code cleanup: Always bail out of functions ASAP, do not
    calculate unnecessary stuff before. Merged if-condition checks.
    Declaration and assignment of local variables in one line. Reduced
    the scope of some variables. Use pre- instead of postfix in- and
    decrements. Added and removed newlines where appropriate. Added
    brackets to foreach macro loops and removed them otherwise (if
    redundant). Fixed indentation. Note: This plugin has a huge chunk of
    code, which has been commented out. (apparantly/according to the
    comment, because it does not work) I was very cautious in
    refactoring this part of the code, as it currently is excluded from
    compilation and thus cannot be tested. TODO: Test, repair and
    integrate this code (or remove it if it is redundant). TODO: Fix
    half-broken "Exit Scale On Pull". TODO: Fix the scale-will-only-
    terminate-if-initiate_key-is-defined bug.
  * *EZoom (Enhanced Zoom Desktop) code cleanup: Optimized ABI check.
    Use prefix- instead of postfix-increment. Assignment and declaration
    of local variables in one line. Merged if statements. Removed
    redundant brackets. Fixed indentation. Added and removed newlines to
    improve readability. *EZoom (Enhanced Zoom Desktop) code changes:
    Replaced #define WIDTH (x2 - x1) with int width = x2 - x1; and
    #define HEIGHT (y2 - y1) with int height = y2 - y1; Replaced WIDTHOK
    and HEIGHTOK macros with local variables: WIDTHOK and HEIGHTOK just
    need to be calculated once and can then be used for all cases in
    void EZoomScreen::ensureVisibilityArea (...). static_cast <float>
    for int denominators in divisions. Removed these redundant gl calls:
    glEnable (GL_TEXTURE_2D); glDisable (GL_TEXTURE_2D); Zoom Box fixes:
    Enabled GL_BLEND for the zoom box to fix transparency issues. Made
    outline and fill color and transparency of the zoom box
    configurable. Outline width is 2.0 for the zoom box as well now.
    (same like for grid preview and screenshot selector outlines) Fixed
    zoom box damage handling. Minor restructuring in zoom box rendering
    code. Minor ezoom.xml.in cleanup and enhancements. Load after
    mousepoll. (fixes LP: #1171665, partially fixes LP: #1051290 and LP:
    #1051295). (LP: #1171665, #1051290, #1051295)
  * Wall: Fixed broken wall edge flipping functions: "Edge Flip Pointer"
    "Edge Flip Move" Removed void WallScreen::updateScreenEdgeRegions ()
    completely. This function not needed hereat all, no other plugin
    needs to re-define the screen edge regions by itself - just core
    should do that. The screen edge region updating also confused the
    edge flipping functionality. case ConfigureNotify: just breaks out
    now Added missing default: case and a break (just a style issue).
    Minor cleanup in bool WallScreen::initiateFlip (...). One
    declaration per line. Note: "Edge Flip DnD" still seems to be
    broken, but will be fixed in a follow-up proposal. (LP: #771448, LP:
    #858845). (LP: #771448, #858845)
  * Wallpaper .xml.in cleanup: Better plugin description. Improved
    option names. Better wording of tooltips. Removed redundant
    newlines. Wallpaper code cleanup: Use prefix instead of postfix
    increments. Declaration and assignment of variables in one line.
    Added some newlines to improve readability. Removed redundant
    brackets. Fixed indentation.
  * EZoom plugin, code improvements: Bail out of functions ASAP, do not
    calculate stuff you might not need. Replaced: if (zoom_level == 1.0f
    && zooms.at (out).newZoom == 1.0f) with: if (zoom_level == zooms.at
    (out).newZoom) because we always want to get out of here if the
    target level is the same as current zoom level, not only in the case
    both are 1.0f. EZoom plugin, code cleanup: Assignment and
    declaration of variables in one line, if possible. Added and removed
    some newlines. Merged if condition checks. Indentation fixes.
  * Mousepoll code cleanup: Removed the unused variables
    MP_OPTION_MOUSE_POLL_INTERVAL and MP_OPTION_NUM. Declaration and
    assignment of variables in one line. Added and removed newlines.
    Removed redundant brackets. Fixed indentation.
  * Annotate plugin, code cleanup: Use prefix instead of postfix
    increment. Declaration and assignment of local variables in one
    line. Removed redundant brackets. Added and removed newlines if
    appropriate. Fixed indentation. Annotate plugin, fixes: Use
    GL_LINE_LOOP to draw the rectangle outline as glRecti is not
    supported on GLES. Simplified the coordinate calculation as the
    offset is not really needed here.
  * Titleinfo, minor code cleanup: Declaration and assignment of local
    variables in one line. Removed redundant brackets. Added and removed
    newlines. Fixed indentation.
  * Text plugin code cleanup: Do not declare variables you do not need
    if you return anyway. Use prefix instead of postfix increments.
    Merged if condition checks. Declaration and assignment of variables
    in one line. Added and removed newlines. Removed redundant brackets.
    Fixed indentation.
  * ccp code cleanup: Declare variables if you need them, do not declare
    them if you return. Declare CCSSettingValue *val; outside the
    foreach loop. Use prefix instead of postfix increments. Merged if
    condition checks. Declaration and assignment of variables in one
    line. Removed redundant brackets. Added and removed newlines. Fixed
    indentation.
  * Commands, minor code cleanup: Declaration and assignment of
    variables in one line. Return ASAP. Added and removed newlines.
    Fixed indentation.
  * Cube code cleanup: Replaced 0xffff with GLfloat MaxUShortFloat =
    static_cast <GLfloat> (std::numeric_limits <unsigned short>::max
    ());. Use prefix instead of postfix de- and increments. Declaration
    and assignment of variables in one line. Declare variables shortly
    before you need them. Merged if condition checks. Added missing
    breaks (just a style issue). Added and removed brackets. Added and
    removed newlines. Fixed indentation.
  * action.cpp and actions.cpp code cleanup: Use pre- instead of postfix
    increment. Declaration and assignment of variables in one line.
    Merged if condition checks. Removed redundant brackets. Added and
    removed newlines. Fixed indentation.
  * Rotate Cube code cleanup: Simplified a few calculations of this
    type: (360.0 / (x * 2.0)) == 180 / x Declaration and assignment of
    variables in one line. Use pre- instead of postfix de- and
    increments. Merged if condition checks. Added and removed newlines
    and brackets. Fixed indentation. Added TODO comments.
  * Fade plugin, minor code cleanup: Declaration and assignment of local
    variables in one line. Merged if condition checks. Use pre- instead
    of postfix de- and increments. Removed redundant brackets. Added and
    removed newlines. Fixed indentation.
  * Opacify code cleanup: Declaration of local variables outside of
    loops. Calculate activeOpacity and targetOpacity just once, not for
    each window. Use prefix instead of postfix increments. Merged if
    condition checks. Declaration and assignment of local variables in
    one line. Added and removed newlines. Removed redundant brackets.
    Fixed indentation. Opacify, functionality fix: Improved the toggling
    logic by setting isToggle in the constructor according to the
    startup setting, then calling setFunctions (isToggle);,which then
    calls screen->handleEventSetEnabled (os, os->isToggle); to fix
    Opacify ignoring the "Toggle Opacify on by default" setting. (LP:
    #787814). (LP: #787814)
  * Widget, minor code cleanup: Declaration and assignment of local
    variables in one line. Merged if condition checks. Added and removed
    newlines. Added default case to switch (just a style issue). Removed
    redundant brackets. Fixed indentation.
  * window.cpp code cleanup: Declare variables outside of loops. Reduced
    the scope of variables. Do not assign values to variables if those
    values are not used. Always bail out of functions ASAP, do not
    calculate stuff you might not need. Use pre- instead of postfix de-
    and increments. Declaration and assignment of variables in one line.
    Merged if condition checks. Added and removed brackets. Added and
    removed newlines. Fixed indentation. Added TODOs (removal of magic
    numbers).
  * Fix for cube.cpp, regression r3720 - Inner sides of cube invisible
    now. Wrong variable type was accidentially used, cullInv is of type
    int, not bool. (LP: #1183852). (LP: #1184852, #1183852)
  * src/output*.cpp cleanup: Return ASAP, do not calculate stuff you do
    not need. Declaration and assignment of variables in one line.
    Declare local variables outside of loops. Use prefix instead of
    postfix increment. Added and removed newlines. Fixed indentation.
  * Expo, code cleanup: Declaration of local variables outside of loops.
    Use prefix instead of postfix increments. Declaration and assignment
    of local variables in one line, if possible. Reduced the scope of
    some variables. Removed redundant brackets. Used static_cast <>
    (type) instead of (type) cast. Added and removed newlines if
    appropriate. Fixed indentation. Added TODOs. Expo, speed
    improvements: Do not calculate screen->vpSize ().width () and
    screen->vpSize ().height () multiple times, instead save them in the
    unsigned ints vpCountHorz and vpCountVert and use those variables
    instead. Speed up the curve calculations by using additional
    variables to save results to not have to re-calculate those all the
    time. The new variables introduced are degToRad, screenWidth,
    screenWidthSquared, curveDistSquaredPlusQuarter,
    pOne2MinusCurveDist, v0Squared and v2Squared. Also introduced const
    float halfGapX = gapX / 2.0 and used this variable in the
    calculations of curveDistance and curveRadius. Simplified
    calculations of this type: (M_PI / 180.0f) * curveAngle / 2.0, which
    is equal to: (M_PI / 360.0f) * curveAngle Do not call
    optionGetGroundSize (); twice, instead save the value in the float
    groundSize and use that variable in the following calculations.
    Removed the creation of the redundant bool hide, which is just used
    once in an if-condition check. It does not help to have this bool.
    Try to avoid redundant GL_BLEND state changes, they are expensive -
    only enable GL_BLEND if it is disabled and just disable it, if it
    was disabled before, otherwise do nothing. Try to avoid redundant GL
    filter changes - just query the filter state if the mipmap option in
    CCSM is enabled. Also just set back the filter to the previous state
    if we actually changed it, otherwise do nothing. -5.5f * 2 = -11.0f.
    No need to calculate M_PI / 180.0f 720 times: Calculate this value
    once instead and save it in the const float mpi, use mpi in the
    following looped calculations. Use const int scw in the same loop
    instead of calling screen->width () 360 times. glow.cpp: Massively
    increased calculation speed of the glow texture: No need for any
    macros here -> removed them and replaced them with local variables.
    Precalculate values, store them in local variables and use those in
    the following calculations (new variables are the ints winRealX,
    winRealY, winRealWidth, winRealHeight, halfWinRealWidth,
    halfWinRealHeight, xPlusHalfWidth, yPlusHalfHeight, xPlusGlowOff,
    yPlusGlowOff, xMinusGlowOff, yMinusGlowOff and the float glowPart).
    Used w->geometry ().widthIncBorders () and w->geometry
    ().heightIncBorders () to determine winRealWidth and winRealHeight.
    Expo, fixes: Do not force "One wall per output" on the user, if his
    displays use different resolutions. "One big wall" makes a lot of
    sense for many multi-screen configs, where not all of the screens
    have exactly the same resolution, so if the user explicitely chooses
    this mode, Compiz should respect the user's choice. (LP: #1009592).
    (LP: #1009592)
  * Compiz, redundant and outdated files cleanup: Removed outdated NEWS
    files. Removed outdated AUTHORS and ChangeLog files. Removed
    plugins/freewins/COPYING - we do not need another copy of the GPL v2
    here. Removed another copy of the GPL v2 from plugins/workarounds.
    This will reduce package sizes, download time and bring some order
    to chaos.
  * Expo, .xml.in fixes: Better description of the plugin. Improved and
    fixed tooltips. Moved the "Curve Strength" option closer to
    "Deformation". Refreshed the quilt patch for Ubuntu: There is no
    need to set the default distance to 0.005 as this setting has no
    effect without deformation. Set default multi-screen mode to "One
    wall per output" for Ubuntu, because "One big wall" works now for
    all display configurations and could potentially change the
    appearance for Ubuntu users. (LP: #1074487). (LP: #1074487)
  * Added contribute directory. Added QtCreatorConfig.xml, a
    configuration file for Qt Creator, which adjusts Qt Creator's C++
    indentation style to be the same that is used in the Compiz project
    (X11, mixed spaces and tabs).
  * Mousepoll: Fixed mousepoll version mismatch. (COMPIZ_MOUSEPOLL_ABI
    was missing) (LP: #1195659). (LP: #1195659)
  * Expo: Simplified the GL filter handling.
  * Fixed broken text in all plugins (missing COMPIZ_TEXT_ABI). (LP:
    #1196493). (LP: #1196493)
  * CCSM: Fixed vertical alignment of the text on the stylized keys.
    (Shift, Super, Ctrl, Alt) (LP: #1196345). (LP: #1196345)
  * Thumbnail, cleanup: Merged if condition checks. Declaration of
    variables when they are needed (C++ style). Removed redundant
    brackets. Removed useless "/* Could someone please explain how this
    works */" comment. Removed extern const unsigned short
    TEXT_DISTANCE, because this variable does not exist anymore (value
    is configurable already). Changed if (something > 0.0) to if
    (something), 10 times. Changed pointedWin = 0; to pointedWin =
    NULL;. Used centerX () and centerY () abstractions from
    compiz::window::Geometry to determine the icon center coordinates.
    Added and removed newlines, if appropriate. Fixed indentation.
    Thumbnail, code speedup: Return ASAP, do not calculate stuff you
    might not need. No need for WIN_W (w) and WIN_H (w) macros, removed
    those and replaced them with the local variables int winWidth, int
    winHeight, int dockWidth and int dockHeight. No need for WIN_X (w)
    and WIN_Y (w) macros, replaced those with int dockX and int dockY.
    Introduced int tHeight = thumb.height;, int tWidth = thumb.width;,
    int halfTWidth = tWidth / 2; and int halfTHeight = tHeight / 2; and
    used those variables in the following calculations. Use
    igMidPoint[0] and igMidPoint[1] as arguments in the screen-
    >outputDeviceForPoint (arg1, arg2) function, instead of re-
    calculating the arguments again. Thumbnail, GL speedup: Introduced
    GLfloat wxPlusWidth = wx + width;, GLfloat wyPlusHeight = wy +
    height;, GLfloat wxPlusWPlusOff = wxPlusWidth + off;, GLfloat
    wyPlusHPlusOff = wyPlusHeight + off;, GLfloat wxMinusOff = wx - off;
    and GLfloat wyMinusOff = wy - off; and used those coordinates in the
    vertexData arrays. No need to re- calculate those values multiple
    times. Just query optionGetShowDelay () once and use int showDelay
    later. We just enable blending if it is currently disabled and
    disable blending only, if it was disabled before.
  * src/session.cpp cleanup: Declaration and assignment of local
    variables in one line. (this *should* fix bug #1101405) Use pre-
    instead of postfix increments. Added and removed newlines. Added
    missing break (just a style issue). Fixed indentation. (LP:
    #1101405). (LP: #1101405)
  * Firepaint, code cleanup: Declare variables outside of loops.
    Declaration and assignment of local variables in one line. Use pre-
    instead of postfix increment. Removed redundant newlines. Removed
    redundant casts. (from (float) optionGetFireColor) Removed redundant
    brackets. Added missing break (style issue only). Added comment and
    TODO. Fixed indentation. Firepaint, speedup: We just enable GL_BLEND
    if it is disabled and we just disable GL_BLEND if it was disabled
    before. Introduced the GLfloats xMinusW, xPlusW, yMinusH and yPlusH
    to store calculated coordinates instead of recalculating them
    multiple times. Introduced float fireLife = optionGetFireLife ();,
    float fireWidth = optionGetFireSize ();, float fireHeight =
    fireWidth * 1.5f; and bool mystFire = optionGetFireMystical (); and
    used those variables inside the loop. Firepaint, .xml.in cleanup:
    Firepaint now has 2 tabs, 'General' and 'Particle Settings'. The
    name of this plugin is Firepaint, <short> should not contain the
    description. Better description of the plugin. Uppercase option
    titles. Punctuation for tooltips. Improved tooltips. Firepaint,
    fixes: Initialize all class member variables in the ParticleSystem::
    ParticleSystem () ctor (LP: #1101512, LP: #1101580). (LP: #1101512,
    #1101580)
  * img* plugins code cleanup: Always bail out of function ASAP, do not
    calculate stuff you might not need. Declare variables outside of
    loops so they won't be re-declared in each loop. Use prefix instead
    of postfix increments. Declaration and assignment of variables in
    one line. Merged if condition checks. Added and removed brackets.
    Added and removed newlines, if appropriate. Minor code structure
    improvements, declare variables when you need them, not much
    earlier. Fixed indentation.
  * Text, speed improvements: Introduced const float halfPi = PI / 2.0f;
    and const float triHalfPi = halfPi * 3; and used those to draw the
    rounded background. Introduced GLfloat xPlusWidth = x + width; and
    GLfloat yMinusHeight = y - height; and used those coordinates for
    the vertexData array. Text, cleanup: Declaration of local variables
    outside of loops. Fixed indentation.
  * cube.cpp, code cleanup: #include "privates.h", not <privates.h>.
    Merged if condition checks. Declaration of variables, where you need
    them to improve the code structure and readability. Use C++ standard
    bool for topDir, bottomDir, allCaps and mCapsPainted[output] (bool
    true/false instead Bool TRUE/FALSE). Initialize all of the class
    member variables in the constructor. Replaced if (priv->mSky.size ()
    > 0) with if (!priv->mSky.empty ()). Removed redundant brackets.
    Fixed indentation. Added TODO. cube.cpp, code speedup: Do not call
    optionGetInactiveOpacity () twice, instead save the value as float
    inactiveOpacity and use this variable in the following calculation.
    Introduced the variables float halfHsize = hsize / 2.0; and float
    tsSize = 360.0f / size; and used those in following calculations to
    not have to recalculate those values multiple times. Do not
    calculate M_PI / sides and 2 * M_PI / sides multiple times in loops,
    instead save those values in the GLfloats mps and tmps and use those
    values inside the loops. Use GLdouble mDist2 = 0.5 / mDistance; in
    the GLdouble clipPlane* arrays, instead of recalculating this value
    multiple times. Use GLfloat oneMinusFStepX = 1.0f - fStepX; and
    GLfloat oneMinusFStepY = 1.0f - fStepY; instead of calculating those
    values multiple times. Use float outputWidth = outputPtr->width ();
    and float outputHeight = outputPtr->height (); instead of calling
    the functions multiple times. Introduced the GLfloat normInvert =
    0.5f * invert; and used this variable instead of recalculating this
    value multiple times. Introduced the GLfloats mSkyWidth and
    mSkyHeight and used those in the following loops instead of calling
    mSkySize.width () and mSkySize.height () multiple times. Try to
    avoid redundant GL_BLEND state changes, because OpenGL will blindly
    change the global state, no matter what it currently is set to and
    that state change is expensive. So we query and save the actual
    blending state in the variable GLboolean glBlendEnabled =
    glIsEnabled (GL_BLEND); and just enable it if it isn't already and
    just disable it, if it was disabled before. (LP: #1101422, LP:
    #1195977) cube.cpp, other fixes: Return false in function
    PrivateCubeScreen::updateGeometry (int, int) if sides should ever be
    0, which should never happen, but should make Coverity happy. (LP:
    #1101541) Initialize all class member variables in the
    PrivateCubeScreen:: PrivateCubeScreen (CompScreen *) constructor.
    (mTc, mNOutput, mOutput, mOutputMask, mCleared, mCapsPainted) (LP:
    #1101575). (LP: #1195977, #1101541, #1101422, #1101575)
  * plugins/ezoom/src/ezoom.cpp: Also initialize xtrans and ytrans in
    the ctors. Just just enable blending if it is disabled and just
    disable it if it was disabled before. Introduced the variables
    oWidth, oHeight, halfOWidth and halfOHeight to speed up following
    calculations. Return ASAP, do not calculate possibly redundant
    stuff. Declaration and assignment of local variables in one line.
    Improved readability. Added and removed newlines. Minor indentation
    fixes. plugins/ezoom/ezoom.xml.in: Removed redundant whitespaces
    from a tooltip. (LP: #1101572). (LP: #1101572)

[ Hu Kang stevenhooke11@gmail.com ]
  * workarounds.cpp: Fix typo in call to XShapeSelectInput. (LP:
    #1167284). (LP: #1167284)
  * Use XIDefineCursor rather than XDefineCursor. compiz should show a
    busy cursor during the application startup stage. the corresponding
    code is in scr/screen.cpp:
    cps::StartupSequenceImpl::updateStateFeedback () the patch changes
    all XDefineCursor to XIDefineCursor (see XInput2 extension) (LP:
    #1179155). (LP: #1179155)

[ Łukasz 'sil2100' Zemczak ]
  * Qt is stupid since it uses the stupid X11 protocol - let's not set
    all the attributes (especially override_redirect) every time, only
    in cases when it's actually needed (LP: #1141079). (LP: #1141079)

[ Andrea Azzarone ]
  * Satisfy matchers on getProperty as soon as they are set. The order
    of evaluation for matchers in Google Mock appears to be undefined -
    this means that we can't rely on the first argument being matched
    first and the second argument being matched afterwards. In turn,
    this means that any GValue may be passed to a GValueMatch which, by
    a design flaw, is unable to handle any values of a type it does not
    expect (at least not without an API change). It will silently pass
    the incorrect type to g_type_get_* which causes internal assertion
    failures. At the moment we're just interleaving the calls to
    getProperty and get_property - that way the expectations are
    satisfied and go away as soon as they're set. This in turn means
    that Google Mock only has to traverse one matcher rather than
    multiple matchers. (LP: #1187468). (LP: #1187468)
  * Extends invisible borders to the title bar too. (LP: #717444). (LP:
    #717444)
  * Disable grid/put_maximize_key on ubuntu.
  * Port to new gmock. Disable some faling tests because of weird leaks.
    (LP: #1185265)
  * Update window geo only on new decoration size only if mapped. (LP:
    #1198000). (LP: #1198000)
  * Disable show_desktop_key by default on ubuntu. Will provide the same
    option under the unityshell panel. (LP: #1204664)
  * Use Glib::RefPtr<Glib::Source> in glib_integration_test and compiz
    core too. After the fix of bug
    https://bugzilla.gnome.org/show_bug.cgi?id=561885
    Glib::Source::~Source is called when both Source::unreference() and
    SourceCallbackData::destroy_notify_callback() are called. (LP:
    #1214459)

[ Sam Spilsbury ]
  * Also take into account the gravity requirements when validating
    reposition requests. Get most of that code under test too. (LP:
    #1159324). (LP: #1159324)
  * Install compizconfig backends to the correct libdir (LP: #1163611).
    (LP: #1163611)
  * Don't timeout after 3000ms when asking compiz for a startup message.
    It can sometimes take longer in the valgrind case. Also timeouts are
    bad design for tests. (LP: #1169170). (LP: #1169170)
  * Check for USE_GCONF if USE_GSETTINGS is unavailable when
    instantiating different types of settings objects. (LP: #1169172).
    (LP: #1169172)
  * Link in CMAKE_THREAD_LIBS_INIT through GTEST_BOTH_LIBRARIES so it
    gets included in everything. Remove redundant links. Require
    FindThreads.cmake. (LP: #1171364). (LP: #1171364)
  * Explicitly check if the smart pointers are null rather than
    converting implicitly to bool as boost 1.53 now marks the bool
    operator as explicit (LP: #1172601). (LP: #1172601)
  * Rewrite screen size change tests. 1. Rename the overly-terse
    variables names to variables names that have some significance 2.
    Split the giant three-test structure into multiple tests each with
    only one assert 3. Remove a lot of redundant calculation 4. Remove
    magic numbers peppered throughout the code. Use constants and
    express asserts as relationships between those constants. 5.
    Refactor some of the more common test-advancement code (such as
    changing the screen size) into the test fixture as common methods.
    (LP: #1167983). (LP: #1167983)
  * Refactor screenshot code to make it clearer and also to allow us to
    take screenshots in glPaintOutput as opposed to paint (). Taking
    screenshots in paint () was probably correct pre-GLES, but isn't
    really correct now - we want to be able to read off of the currently
    bound scratch framebuffer where we last painted the frame. Reading
    off the frontbuffer is an imprecise operation because the contents
    of both buffers are really undefined after a swap. In the case where
    we are not painting to a scratch framebuffer object, we'll just end
    up reading from the backbuffer anyways, which would be correct to do
    mid-frame. Also added the new static const GLenums
    DRAW_FRAMEBUFFER_BINDING and READ_FRAMEBUFFER_BINDING to opengl.h.
    (LP: #771875). (LP: #771875)
  * Set libcompizconfig_libdir before the modules get a chance to see
    it. (LP: #1173799). (LP: #1173799)
  * Try to launch tests on another server if there's tests running in
    parallel. Provide our own compiz_xorg_gtest_main and subclass
    xorg::testing::Environment to try and launch tests on another
    display if there's tests running on one already. This isn't by any
    means perfect - there are still race conditions surrounding
    XOpenDisplay and parallel test runs but it makes a more smaller time
    gap for conditions such as:  1. Client has a server grab on the
    display we're checking     and won't let go.  2. Two servers get
    launched on one port and one set of tests     interfere with the
    other. It also means that we're now unable to configure the display
    port, log file and config file on the command line. But we weren't
    using that anyways. Finally, the logs now point to
    /tmp/Compiz.Xorg.GTest.displaynum.log (LP: #1178514). (LP: #1178514)
  * Fix some files not installing: 1. gsettings schemas never get added
    to the install_manifest list because we use install (CODE) so we
    need to add a custom uninstall target for those, and also need to
    recompile on uninstall too 2. Run the gtk-update-icon-cache stage
    during install for ccsm and also add icon-theme.cache to the
    uninstall target (LP: #1168475). (LP: #1168475)
  * Wall: Fixed broken wall edge flipping functions: "Edge Flip Pointer"
    "Edge Flip Move" Removed void WallScreen::updateScreenEdgeRegions ()
    completely. This function not needed hereat all, no other plugin
    needs to re-define the screen edge regions by itself - just core
    should do that. The screen edge region updating also confused the
    edge flipping functionality. case ConfigureNotify: just breaks out
    now Added missing default: case and a break (just a style issue).
    Minor cleanup in bool WallScreen::initiateFlip (...). One
    declaration per line. Note: "Edge Flip DnD" still seems to be
    broken, but will be fixed in a follow-up proposal. (LP: #771448, LP:
    #858845). (LP: #771448, #858845)
  * Add a small compiz_autopilot_acceptance_tests binary based on gtest.
    This will find autopilot and run a predefined list of tests, wrapped
    up in the familiar google test format. The predefined list is
    basically a bunch of tests which effectively interact with the
    window manager. (LP: #1170013). (LP: #1170013)
  * Change the behaviour of undecorating windows. Previously when a
    window was undecorated, we would shift it back to an appropriate
    position according to its gravity member. That behaviour was
    problematic because in the StaticGravity case the window has to just
    stay in the same place. But then if you had a window with
    StaticGravity which then did get a decoration and later removed it,
    it would be placed as though it was decorated and appear to be in
    the wrong place. The correct behaviour is to place all windows as
    though they have decorations, and then when decorations are removed,
    to move the window back to the corner as indicated in its gravity
    and then expand its size to cover the obscured regions no longer
    hidden because the decorations went away. (LP: #1165343). (LP:
    #1165343)
  * Added some new hooks to PluginClassHandler to allow a VTable to
    specify if loaded. PluginClassHandler::get () was designed to simply
    instantiate an instance of that class for the core structure, but it
    did this without checking if the plugin was loaded. Added some new
    methods to PluginClassHandler exposed by LoadedPluginClassBridge and
    only accessible by those who implement PluginKey to specify globally
    whether or not a plugin is actually loaded, so that
    PluginClassHandler can return accordingly. Integration and unit
    tests added as appropriate (LP: #1169620) (LP: #1101026). (LP:
    #1101026, #1169620)
  * Satisfy matchers on getProperty as soon as they are set. The order
    of evaluation for matchers in Google Mock appears to be undefined -
    this means that we can't rely on the first argument being matched
    first and the second argument being matched afterwards. In turn,
    this means that any GValue may be passed to a GValueMatch which, by
    a design flaw, is unable to handle any values of a type it does not
    expect (at least not without an API change). It will silently pass
    the incorrect type to g_type_get_* which causes internal assertion
    failures. At the moment we're just interleaving the calls to
    getProperty and get_property - that way the expectations are
    satisfied and go away as soon as they're set. This in turn means
    that Google Mock only has to traverse one matcher rather than
    multiple matchers. (LP: #1187468). (LP: #1187468)
  * Revert revision 3278.
  * Immediately update the window matrices and regions if the number of
    textures changed. (LP: #1189369). (LP: #1189369)
  * Provide a basic decor plugin acceptance test suite. This change
    provides a simple acceptance test suite for the decor plugin. It
    creates a fake window decorator (cdt::FakeDecorator) and allows
    users to create fake decorations (eg cdt::FakePixmapDecoration)
    which can be serialized as decoration properties and set on windows
    under xorg-gtest. It also launches compiz with the opengl, composite
    and decor plugins loaded and runs some basic tests. Among them:
    BaseDecorAcceptance. Startup: Basic canary "can we start compiz with
    these plugins" test
    FakeDecoratorSessionOwnerNameSetOnSelectionOwner: Create a fake
    decorator, ensure that libdecoration sets our session name "fake" on
    the selection owner FakeDecoratorReceiveClientMessage: Create fake
    decorator, ensure that libdecoration posts a client message to the
    root window announcing that the new decorator exists.
    DecorationSupportsWindowType: Create a fake decorator, announce
    support for the WINDOW type decorations and ensure that the correct
    atom is set on the session owner window.
    DecorationSupportsPixmapType: Create a fake decorator, announce
    support for the PIXMAP type decorations and ensure that the correct
    atom is set on the session owner window.
    DecorFakeDecoratorAcceptance. WindowDefaultFallbackNoExtents: By
    default, newly created windows should recieve a fallback decoration
    but they should not have any frame extents.
    DecorWithPixmapDefaultsAcceptance. These tests create a default
    pixmap decoration to use in the hypothetical situation that the
    window decorator hasn't yet generated a decoration for this window.
    FallbackRecieveInputFrameNotify: Verify that we get a
    _COMPIZ_WINDOW_DECOR_INPUT_FRAME property set on the client when it
    is created and mapped (eg, an input frame window was annonuced to be
    created) FallbackHasInputFrameInParent: Verify that a second window
    exists in the frame window after this message was recieved.
    FallbackNormalWindowExtentOnDecoration: Verify that
    _NET_FRAME_EXTENTS is set to the default window extents for the
    default decoration when the window was mapped.
    FallbackNormalWindowInputOnFrame: Verify that the input window
    matches the extents set. PixmapDecoratedWindowAcceptance. These
    tests create a new window and an associated unique decoration for
    that window. MaximizeBorderExtentsOnMaximize: Maximize the window
    and ensure that the correct border extents are used.
    MaximizeBorderExtentsOnVertMaximize: Vertically maximize the window
    and ensure that the correct border extents are used.
    MaximizeBorderExtentsOnHorzMaximize: Horizontally maximize the
    window and ensure that the correct border extents are used.
    MaximizeFrameWindowSizeEqOutputSize: Maximie the window and ensure
    that the frame window exactly equals the output size.
    VertMaximizeFrameWindowSizeEqOutputYHeight: Maximize the window and
    ensure that the frame window's Y and Height values are consistent
    with the output size. HorzMaximizeFrameWindowSizeEqOutputXWidth:
    Ditto horizontal maximization.
    DISABLED_VertMaximizeFrameWindowSizeSameXWidth: Maximize the window
    vertically and ensure that the frame window's border-relative X
    position and width does not change. This test is disabled, as the
    behavior in compiz is currently broken, (but it means that we can
    fix it later and enable the test).
    DISABLED_HorzMaximizeFrameWindowSizeSameYHeight Maximize the window
    horizontally and ensure that the frame window's border-relative Y
    position and height does not change. This test is disabled, as the
    behavior in compiz is currently broken, (but it means that we can
    fix it later and enable the test). (LP: #1188900). (LP: #1188900)
  * Only mark for no further instantiations once we've finished
    destructing the plugin (LP: #1193596). (LP: #1193596)
  * Add more acceptance tests for the decor plugin.
    PixmapDecoratedWindowAcceptance. UndecoratedWindowExpandToOrigSize
    Test that upon undecoration, the window has exactly the same
    geometry as it did before decoration.
    DISABLED_UndecorateStaticGravityWindow Tests that for windows with a
    static gravity, the window has exactly the same geometry as it did
    before decoration. Disabled, as core is currently not moving the
    window with the static gravity back to where it started.
    AdjustmentExtents/PixmapDecorationAdjustment.
    AdjustRestoredWindowBorderMovesClient/P Test that changing the
    border extents causes the client window's absolute geometry to
    change. DISABLED_AdjustRestoredWindowBorderShrinkClient/P Tests that
    changing the border extents causes the client window's absolute
    geometry to shrink by the amount of border. Disabled, as the current
    behavior is to expand the frame window as opposed to shrinking the
    client (but this will be changed).
    DISABLED_ClientExpandsAsBorderShrinks/P Tests that as the border
    shrinks away, the client expands back to its original size.
    Disabled, as the current behavior is to expand the frame window as
    opposed to shrinking the client (but this will be changed).
    DISABLED_ClientExpandsAsBorderShrinksWhilstMaximized/P Tests that
    even if the client is maximized, if the restored border shrinks away
    it demaximizes back to the same position while also accounting for
    any change in the restored window border size. Disabled, as the
    current behavior is to expand the frame window as opposed to
    shrinking the client (but this will be changed).
    DISABLED_ClientExpandsAsBorderShrinksWhilstUndecorated/P Tests that
    even when the client is not permitted to be decorated because the
    decoration hint was removed, when it is redecorated it will be moved
    to the correct position taking into account any changes in its
    border size. Disabled, as the current behavior is to expand the
    frame window as opposed to shrinking the client (but this will be
    changed). DISABLED_AdjustRestoredWindowInputNoMoveClient/P Tests
    that adjusting the input extents (as independent from the border
    extents) of the client does not cause the client to move. Disabled,
    as this behavior appears to be broken in core. Move the
    "GetImmediateParent" function into the compiz::testing namespace so
    that it can be used by other tests. (LP: #1188900)
  * Unrevert 3728, fix failing tests. Change the behaviour of
    undecorating windows. Previously when a window was undecorated, we
    would shift it back to an appropriate position according to its
    gravity member. That behaviour was problematic because in the
    StaticGravity case the window has to just stay in the same place.
    But then if you had a window with StaticGravity which then did get a
    decoration and later removed it, it would be placed as though it was
    decorated and appear to be in the wrong place. The correct behaviour
    is to place all windows as though they have decorations, and then
    when decorations are removed, to move the window back to the corner
    as indicated in its gravity and then expand its size to cover the
    obscured regions no longer hidden because the decorations went away.
    (LP: #1165343).   1. Completely remove decorOffsetMove and other
    related code from      decor.cpp. Put the logic to handle the
    window->input () - window->border ()      placement offset inside of
    setWindowFrameExtents instead. Now the window      will always be
    offset from its original non-decorated position to the new
         decorated position, rather than having to guess between
    decoration sizes.   2. Make saveGeometry and restoreGeometry work
    relative to window->border ()      as opposed to including it in the
    saved geometry. It is possible that the      border size might
    change during maximization, as such, we don't want to      save the
    position with the border before maximizing. Instead save the
    position      as if it were never decorated so that when the window
    is restored it can be      restored to its original position and
    then adjusted for its new border size.   3. Fix a few typoes in the
    tests.   4. Moved some commonly used matchers into compiz::testing
      5. Make COMPIZ_PLUGIN_DIR accept multiple directories and look in
    each one      of them for the plugin   6. Set COMPIZ_PLUGIN_DIR
    appropriately for each plugin that we wish to load      on startup
    so that we load locally built plugins as opposed to installed
         ones.   7. Uncomment compiz_discover_tests for the acceptance
    tests. Now they are      run by default. (LP: #1195522). (LP:
    #1195522, #1165343)
  * Add a simple script for making releases.
  * Port to new gmock. Disable some faling tests because of weird leaks.
    (LP: #1185265)
  * Don't add the frame to the toplevel stack if it hasn't been created
    yet. In the event that a window is unreparented or destroyed, we
    usually need to add its frame window to the toplevel window stack
    until the time at which we recieve a DestroyNotify for it, as there
    may be incoming ConfigureNotify events puporting to stack other
    windows relative to that frame. However, this does not apply in the
    case where we have not yet received a CreateNotify for the frame
    window. In that case, it is not possible for any stacking requests
    to be made relative to this window, so it does not need to be added
    immediately. Instead, we can add it at the time that we recieve a
    CreateNotify for it as a regular override redirect window until the
    time that it is later destroyed. (LP: #1171314). (LP: #1171314)
  * Remove redundant src/logmessage/include/core/logmessage.h (LP:
    #1067246). (LP: #1067246)
  * Adjust acceptance tests for the fix to LP: #1198000 1. Always set
    frame extents before updating the frame 2. Added
    COMPIZ_NO_CONFIGURE_BUFFER_LOCKS so that we don't have to deal with
    the delayed configure requests logic in the tests where it does not
    help 3. Split restoredDecorationSize in to
    restoredDecorationBorderSize and restoredDecorationInputSize so that
    we can measure the size of the decoration both in terms of its
    visible and invisible border 4. Wait for the default decoration
    properly 5. Capture the initial geometry of the window with and
    without decorations when the window gets its first decoration. Use
    these values to calculate the size changes in the window from then
    on instead of its creation values 6. Just flag when the window was
    first decorated in setWindowFrameExtents rather than checking if it
    is mapped. We only want to not resize the window once. (LP:
    #1198000)
  * Release version 0.9.10.0
  * Bump VERSION to 0.9.10.2

[ Ubuntu daily release ]
  * Automatic snapshot from revision 3781
 -- Ubuntu daily release <ps-jenkins@lists.canonical.com>   Thu, 22 Aug 2013 06:58:07 +0000

Changed in compiz (Ubuntu):
status:	New → Fix Released

Report a bug

This report contains Public information

Everyone can see this information.

Duplicates of this bug

You are

Subscribing...

Edit bug mail

Other bug subscribers

Remote bug watches

gnome-bugs #561885
[RESOLVED FIXED] Edit
freedesktop-bugs #61182 Edit

Bug watches keep track of this bug in other bug trackers.