due to recent changes in ecore-input-evas, mouse events are propagated
differently; specifically, there are now "more" events than there previously were.
as a result, grabs on internal wins are no longer necessary, though they probably
never were necessary after the elm conversion
see 5cb6cdbc5e1a13ea0262e155983b494e6519abde in efl
I did an audit of this and it seemed that it no longer served the purpose
for which it was originally intended. specifically, this is for enforcing
click: raise/focus options, and so grabs must be in play on client windows
only when they are not focused to ensure that we get mouse events and can
then focus them. the grabs must then be removed once the window has focus
to avoid spurious mouse eventing
xwayland sets a wrong size on some (eg. menus) clients and wayland
cannot provide geometry or stacking information, so ensure that all
of this is copied over
also remove overrides from focus stack
this should fix the case of mouse-based focus policies trying to reapply
focus after another client has stolen it away without the pointer leaving
the window
in the case of different window <-> event_window, window is a child window
of event_window, and thus checking event_window here is valid (and necessary)
when you have click to focus we have a passive grab set up. somewhere
that window changed to the parent window instead of the client. this
leads to a side effect of a leave and enter event on clients for every
click. generally clients are ok with this, but some seem to have buggy
event handling. these enter/leave events are a side effect of the
passive grab even though we allow/replay the event.
this fixes that by placing passive grabs on the client window itself
instead of the parent.
@fix
these helper functions automatically account for "swapped" xwayland
clients and return the expected value from the wl client comp_data.
in this way, all of the current x11 compositor code can be reused with
minimal changes
in order to maximize the amount of reused code the following details the current
process for xwayland compositing:
* get map request from window
* force reparenting
* show window
* await WL_SURFACE_ID x11 message
* move x11 client data + pixmap onto corresponding wayland client
* business as usual with wayland compositing
this is pretty similar to the method of the reference code in weston,
except that there's no x11 compositor in weston
this follows 56cabf59c6 then
4e5521b4d8 where i have been trying to
fix a crash with e client and comp win references etc. i have gone
over all referencing with a fine tooth comb and found all the nigglies
i can., no leaks now, no crashes, no valgrind complaints etc. so i
call this fixed now. as best i know this is new in e20, so not a
backport fix
the refcoutning for e_comp and e comp clients seemed to be a bit off -
i read over every ref and unref carefully and fix it. this leads to
the com-_data being null (properly now), so now check for that too.
this came in a patch that I take full responsibility for not adequately reviewing at the time.
the names are confusing and counterintuitive, and it does not properly include the client namespace.
this makes highest priority screen the lowest (0) zone. this also
handles missing screesn that you are relative "of". missing clones are
not working atm. also zone reconfigure moves windows now too
this fixes some issues in the new randr2 code that made it not work
right with plug/unplug and lid close/open. now it does work right and
plugging/unplugging displays is seamless (if your driver does not give
plug/unplug events bind a key to update screen config acvtion and e
will figure it out when you hit the key).
the problem is something changes window gravity... what i don't know,
but hey - just forcibly move window to 0,0 which is where we expect it
anyway when resizing.
@fix
the dialog for now is simple and lets you just raw edit the properties
per screen in a dialog. nothing fancy. not user firendly. but it works.
the randr core has been totally rewritten and tested against a range
of drivers and setups before even getting a commit. it works solidly
and configures screens reliably now. drivers tested:
nvidia
intel
radeon
some drivers still are unreliable in terms of delivering plug/unplug
events for outputs (both intel and radeon are flakey - nvidia is solid
and reliable). so to fix this there is now a screen redo action you
can bind to a hotkey or something and have e re-evaluate current
screen setup and apply ny pending config if needed.
also to make reconfiguring prettier the screen is faded to black
first, then configured, then faded back in. some drivers work
flawlessly with this, others still flicker some garbage.
i admit - i haven't tested nouveau, but my general take on this is the
randr code is now in far better shape than where it was (minus pretty
and easy dialog). the dialog can be done next, but i'd like to get the
core in now for more testing.
@fix
there is only one E_Comp which can now be accessed by the e_comp global.
if you're editing a file with some uses of these deprecated functions, replace their usages with appropriate references to this variable
pass -Wno-deprecated-declarations to ignore these warnings during build
according to the shape extension protocol, the number of rectangles returned should be checked to determine a client's shape. if 0 is returned, the client has no shape, meaning that it either should not be drawn or should have no input region. this improves behavior with various client window types such as tooltips
ref T1820
this addresses a vbox gl asccel issue where gl client apps can cause
al of vbox to crash and users find it unexpected that e is set to
software compositing but apps go and use opengl, so it's more
consistent for users
Summary:
1. fix window profile change request with wrong x window id
2. refactoring desktop window profile codes to handle e_client_desk_set correctly
Test Plan:
1. enlightenment: Settings Panel -> Screen -> Virtual Desktops -> Check "Use desktop window profile" -> Apply
2. open an efl app on 1st desktop (0-0)
3. set a window remember on efl app window
4. go to 2nd desktop (1-0)
5. open an efl app again. it should be positioned on the previous desktop (0-0)
Reviewers: zmike, raster
CC: cedric
Differential Revision: https://phab.enlightenment.org/D926
when this feature was added, its behavior was naively set to ignore repeated desk changes. it also was not adequately tested, resulting in a frustratingly large number of bugs.
with these changes, window profiles should no longer be the cause of client visibility being broken as caused by a failure to change desk.
fix TChutney
Summary:
Black area of override redirect window could send damage notify which has unwanted xy position.
(a) Skip x configure notify with 3,55 316x162 of override redirect win
(b) Handle x show notify: create a new client and x damage for override redirect win
(c) Handle x damage notify with 3,55 316x162
To resolve it, E discards unwanted xy position of first damage for override redirect window.
This fixes remained problem of T1180 and T1188.
Test Plan:
1. Run e with sw engine mode
2. Run pcmanfm
3. Select menu in pcmanfm
4. Check whether menu window has black area
Reviewers: raster, zmike, devilhorns
CC: cedric
Differential Revision: https://phab.enlightenment.org/D800
Summary:
Can not turn off screensaver due to condition check bug.
@fixes
Test Plan: enlightenment -> do nothing -> screen saver on -> mouse/key input -> check whether screen saver is turned off or not
Reviewers: raster, cedric, zmike
Reviewed By: raster
CC: seoz, cedric
Differential Revision: https://phab.enlightenment.org/D766
this was added a while ago to fix positioning of windows that wanted to start centered but couldn't accurately calculate xinerama screen sizes, resulting in windows getting centered across the screen split. it ended up being a bit too aggressive, however.
this option causes window activation requests to only activate a window if it is on a currently visible virtual desktop, otherwise it will be set as urgent. I recall that things may have worked this way long ago...
the main idea here is to not DRAW at the time of the first damage to avoid overdraw, but ignoring the fact that the region is ready to be drawn can be problematic when the drawing eventually occurs. best choice here is to keep the region but not the render update
it seems i have an override-redirect window just off the bottom-right
of my screen - i think its the scim input panel status. what happens
is it is "managed" by comp but then deleted (_e_comp_x_hook_client_del
called), BUT _e_comp_x_object_add is called with a deferred event for
that client to add it again (likely this is a race) which finds he
client in a state of not having comp_data as the E_FREE in
_e_comp_x_hook_client_del() frees it and sets it to NULL. move the
comp_data free to the actual client free (which is the last time a
client is valid at all) solves this.
this fixes a race condition when windows open simultaneously and then are stacked under each other: the previous result was that they would end up hanging out at the top of the window stack (above all windows) until another window was raised above them. now they stack as expected
this flag is used to block most cwm actions until after the client is past its first loop iteration. overrides, however, don't need to be blocked and will in fact break sometimes (fuck you gtk) as a result
it's worth noting that showing/hiding overrides extremely fast will still break the compositor a little in that I'm not going to render a single frame while constantly playing edje hide/show animations which are longer than the actual render. with animations disabled, however, rendering occurs as expected.
T831
so comp would call ecore_x_composite_window_events_enable in
_e_comp_x_client_shape_input_rectangle_set()
_e_comp_x_hook_client_post_new_client() all the time.. because
ec->need_shape_merge was always set.. why was it always set.. because
ecore_x_composite_window_events_enable woudl set shape rectangles to
make the window visible to input events... so e would feed itsefl with
events all day long via x. this shortcuts that to check if event rects
are the SAME then don't set them as the ones that are stored.
we can assume that these are always going to be ready for drawing immediately, and sometimes X fucks up the damages so it's best to go with the full frame from the beginning
there's an event from ecore-x which tells us when this property has changed, which means we no longer have to be constantly checking to see if other apps have changed it
when not dealing with a re_manage client (persisting from previous E instance), we're almost definitely going to get multiple full-window damages as a new client starts up. the first damage USUALLY results in a garbage render, so ignoring it makes the first frame become actual window content (good!) or a transparent rect until the client finishes drawing (not good!), which is at least an improvement over showing random pixels
xorg 1.15 introduces this extension which has a magical event to notify when a pixmap's size changes, which means that the size never needs to be manually fetched
if we're running in a non-X environment, we can't very well expect that using X calls to hide/show windows for desklock is going to work as expected. now a compositor backend can add a pre or post desklock hook to hide/show its clients as necessary
if a client's X attributes fail to be fetched on the first attempt, it's possible that we might succeed on a second attempt, which would prevent us from losing a client
pointers in E now come in two flavors, one for each protocol that we currently support. each is created from the owner compositor backend:
X pointers still work as usual
Wayland pointers are actually surfaces that we dump image data into periodically to render them as evas images
some small amounts of hack is necessary to make this work, namely blocking X pointers when a Wayland one is in use if we're running a multi-protocol compositor
* try to clear up build system for separating out ecore-x
* add #ifdefs for lots of ecore-x stuff
* break out some internal e wl functions for reuse in api
* store wl surface buffers as an inlist
* add protocol-specific client compositor data
** move lots of X client attributes here
* add pixmap type checks to a number of X-specific things, such as grabinput, to block them for non-X clients
* rearrange startup order to work with wayland
* move X screensaver code to e_comp_x
* flag modules still requiring X with -DNEED_X
huge fustercluck commit because there wasn't really a way to separate out the changes. better to just rip it all out at once.
* compositor and window management completely rewritten. this was the goal for E19, but it pretty much required everything existing to be scrapped since it wasn't optimized, streamlined, or sensible. now instead of having the compositor strapped to the window manager like an outboard motor, it's housed more like an automobile engine.
** various comp structs have been merged into other places (eg. E_Comp_Zone is now just part of E_Zone where applicable), leading to a large deduplication of attributes
** awful E_Comp_Win is totally dead, having been replaced with e_comp_object smart objects which work just like normal canvas objects
** protocol-specific window management and compositor functionality is now kept exclusively in backend files
** e_pixmap api provides generic client finding and rendering api
** screen/xinerama screens are now provided directly by compositor on startup and re-set on change
** e_comp_render_update finally replaced with eina_tiler
** wayland compositor no longer creates X windows
** compositor e_layout removed entirely
* e_container is gone. this was made unnecessary in E18, but I kept it to avoid having too much code churn in one release. its sole purpose was to catch some events and handle window stacking, both of which are now just done by the compositor infra
* e_manager is just for screensaver and keybind stuff now, possibly remove later?
* e_border is gone along with a lot of its api. e_client has replaced it, and e_client has been rewritten completely; some parts may be similar, but the design now relies upon having a functional compositor
** window configuration/focus functions are all removed. all windows are now managed solely with evas_object_X functions on the "frame" member of a client, just as any other canvas object can be managed.
*** do NOT set interceptors on a client's comp_object. seriously.
* startup order rewritten: compositor now starts much earlier, other things just use attrs and members of the compositor
* ecore_x_pointer_xy_get usage replaced with ecore_evas_pointer_xy_get
* e_popup is totally gone, existing usage replaced by e_comp_object_util_add where applicable, otherwise just placed normally on the canvas
* deskmirror is (more) broken for now
* illume is totally fucked
* Ecore_X_Window replaced with Ecore_Window in most cases
* edge binding XWindows replaced with regular canvas objects
* some E_Win functionality has changed such that delete callbacks are now correctly called in ALL cases. various dialogs have been updated to not crash as a result
comp files and descriptions:
e_comp.c - overall compositor functions, rendering/update loop, shape cutting
e_comp_x.c - X window management and compositor functionality
e_comp_wl.c - Wayland surface management and compositor functionality
e_comp_canvas.c - general compositor canvas functions and utilities
e_comp_object.c - E_Client->frame member for managing clients as Evas_Objects, utility functions for adding objects to the compositor rendering systems
additional authors: ivan.briano@intel.com
feature: new compositor
removal: e_border, e_container, e_popup
Mike Blumenkrantz
Carsten Haitzler
Yomi
Chris Michael
Chris Michael
Mike Blumenkrantz
Gwanglim Lee
Seunghun Lee
Wonguk Jeong
Daniel Kolesa
ChunEon Park
discomfitor