path: root/src
AgeCommit message (Collapse)AuthorFilesLines
2011-02-24Merge branch 'master' of ↵HEADmasterSam Spilsbury7-39/+215
2011-02-24Update for new gtk-w-d changesSam Spilsbury2-1/+2
2011-02-13Be a little bit smarter about that - also make the parent appearSam Spilsbury2-1/+42
unfocused when the dialog goes out of focus
2011-02-13Display transient parent as "active" if the current window has aSam Spilsbury4-1/+59
transient dialog New UXD specification
2011-02-10Add support for the switcher window being override-redirectSam Spilsbury6-43/+120
2011-02-03Explicitly include gdk/gdk.h to avoid offset problems on amd64Malcolm Lewis1-0/+1
2011-02-01Support semantic difference between frame extents and frame geometrySam Spilsbury3-11/+15
2011-01-31Revert "Don't dep on system libdecoration but on source built-one"Sam Spilsbury1-1/+2
This reverts commit 8e515dd1dd7eaf8cbf4ee5e1ef182378757aba20.
2011-01-30Fix right side grab area offsetSam Spilsbury1-0/+1
2011-01-30Redraw decoration even if the size update failed. Fixes LP #704413Sam Spilsbury1-30/+42
2011-01-29Fix typo which caused window shadows not to appear.Sam Spilsbury2-2/+2
2011-01-29Revert "Revert "Don't dep on system libdecoration but on source built-one""Sam Spilsbury1-2/+1
This reverts commit 86b00f1e1a9baff35992a58ebc962cc30d929457.
2011-01-28Revert "Don't dep on system libdecoration but on source built-one"Sam Spilsbury1-1/+2
This reverts commit 8e515dd1dd7eaf8cbf4ee5e1ef182378757aba20.
2011-01-24Fix non boolean return if there is no frame styleSam Spilsbury1-1/+1
2011-01-19Don't dep on system libdecoration but on source built-oneAurélien Gâteau1-2/+1
(as the decorator is a distro-patch for now in the compiz source)
2010-12-10Bye-Bye tooltips. It was good to have you on board but design says you are ↵Sam Spilsbury5-233/+16
2010-12-10Fix button offsetSam Spilsbury1-0/+14
2010-12-10Fix typo and orderingSam Spilsbury1-29/+29
2010-12-10Allow decoration to specify an invisible grab region for the frame (to ↵Sam Spilsbury1-0/+92
handle small frame sizes)
2010-12-10Allow the theme to specify a shadow opacity. Also nuke debug messagesSam Spilsbury2-5/+4
2010-12-10Allow metacity themes to specify shadows. WARNING: requires patched ↵Sam Spilsbury4-24/+112
libmetacity private
2010-12-10Add the ability for themes to specify active/inactive shadows, right now ↵Sam Spilsbury5-94/+269
shadow values are hardcoded
2010-12-09Testing branch for active/inactive shadowsSam Spilsbury6-32/+175
2010-12-09Fix not reading gconf properties and generate inactive shadowsSam Spilsbury4-13/+63
2010-12-09Add theme_get_shadow method. Currently doesn't do muchSam Spilsbury3-0/+23
2010-12-09Expand shadows to allow for active and inactive shadows.Sam Spilsbury5-22/+26
Currenty only does active shadows
2010-12-07Rename to unity-window-decoratorSam Spilsbury1-25/+20
2010-12-07Make cmake workSam Spilsbury1-64/+151
2010-12-07Added sourcesSam Spilsbury21-146/+8670
2010-12-07Add standalone building gtk-window-decoratorSam Spilsbury30-19604/+162
2010-12-06Fix startup order.Sam Spilsbury4-93/+183
Previously we were doing something like this: init -> screen init -> init core plugin screen -> create windows -> eventLoop -> updatePlugins -> initPlugin (initScreen (initWindow))). This order has a number of problems with it - namely that config plugins will be loaded on one full pass of the event loop and then core options are initialized, before which we were doing certain startup procedures on the default options, which resulted in strange race condition bugs. Also, if we are compositing we have a brief (visible) switch from 2D mode to 3D mode - in reality we want to be in 3D mode ASAP. The new order is init -> updatePlugins (load plugins specified on command line) -> screen init -> load command line plugins -> update plugins (load new plugins specified by config plugins on the command line) -> init screens -> create window (init plugins for window) -> eventLoop. This means that it takes longer for us to get to the event loop, but at least we can zip through the initialization phase quite quickly
2010-12-05Fix windows being forever treated as "not managed" in focus () and not ↵Sam Spilsbury1-0/+1
entering into scale mode, etc We need to re-set unmanaging too once managed is set again
2010-12-03Fix some focus issues:Sam Spilsbury4-2/+16
- Don't give soon-to-be-unmanaged windows focus when closing a window - Transfer focus to another window if a window that is about to get focus is closed Forward port 5a78a90a3ec9de02b8e52790fbc376a2502cead7 to master. Fixes gnome-do issue
2010-12-03Revert "Ensure that we can change the input focus, even if we requested a ↵Sam Spilsbury4-52/+15
client to take focus and it failed to do it before it was unmapped" This reverts commit 1084f0be5c200d46ad3b4358a477963d31040736. Danny suggested a better way to do this from a commit to 0.8
2010-12-03Ensure that we can change the input focus, even if we requested a client to ↵Sam Spilsbury4-15/+52
take focus and it failed to do it before it was unmapped
2010-12-02Make w->managed () wrapableSam Spilsbury1-0/+5
2010-11-11Listen for FocusIn/FocusOut on NotifyGrab NotifyUngrab and set an internal ↵Sam Spilsbury3-1/+18
bit that plugins can query
2010-11-11Make sure that we properly unregister actions when removing the option ↵Sam Spilsbury1-24/+26
private with operator=
2010-11-09[PATCH] Be a little bit more clever when handing option construction in ↵Danny Baumann4-20/+25
handleActionEvent. Previously we would have a static CompAction::Vector which we would resize to lower sizes (so that triggerFooBindings doesn't die) but this causes us to be reconstruction CompOptions all the time which is expensive. Instead just reset the option to make it "dead" (eg keep it in memory for longer) but don't create and recreate them all the time Also copy the active state of a CompAction on copy-construct
2010-11-08Don't increase core action register reference counts if the same action was ↵Sam Spilsbury3-4/+23
already registered
2010-11-06Don't increase leave pendingMaps at a non-zero number once the window is mapped.Sam Spilsbury1-19/+19
We used to assume that for every MapRequest event that we got for which we also sent an XMapWindow that we would get a MapNotify. According to the man XMapWindow this doesn't make any sense since trying calling XMapWindow on an already mapped window for which we have SubstructureRedirectMask set results in a NoOp anyways (so we'll only ever get one MapNotify and multiple MapRequests). This fixes a case where no input focus would be set on closing windows with multiple remaining pendingMaps since we don't move the input focus if this is the case
2010-10-24Use XFreeStringList to recursively free the entire arraySam Spilsbury1-2/+2
2010-10-24Cleanup (static analysis)Sam Spilsbury1-0/+3
2010-10-24Fix some memory leaks (it looks like there are some more in ↵Sam Spilsbury1-18/+29
XTextPropertyToStringList too)
2010-10-24core: general code cleanup (static analysis)Sam Spilsbury5-10/+21
2010-10-07Remove unused variables to silence warnings.Scott Moreau1-5/+1
2010-09-30Clean up isViewableSam Spilsbury1-4/+1
2010-09-09Force input focus on button 1/2/3 press.Sam Spilsbury1-3/+2
Sometimes there is a case where the window is active on one screen already, and because compiz is not multi-screen aware, when another window is focussed on another screen, the window will continue to be "active" but not focussed. Previously, when clicking on this window again, the window would not be focussed again since it was already active. This commit forces input focus regardless of the "active" state
2010-08-28Better checking if the window is destroyed and use the default visualSam Spilsbury1-13/+28
and colormap, unless the window has request an ARGB one, in which case use that (like metacity). This fixes problems with windows which set incorrect colormap/visual combinations in their XSetWindowAttribs, like some SDL windows such as frozen-bubble and SDL/testwm
2010-08-20Fix warning.Scott Moreau1-1/+1