this is a not-great way of hacking around various issues related to
the efl mouse button cancel patches that went in for the 1.15 cycle
which changed the entire mouse input workings of the toolkit.
to avoid further issues, the compositor will explicitly block eventing
on all internal canvases during actions
these are triggered "in passing" when mouse in events occur and do
not necessarily indicate that the mouse has entered this specific window
failing to reject such events can cause mouse-based focus policies to
attempt to set focus onto windows which are not visible, resulting in
an infinite loop where no window is actually focused
in the event that these windows are different, event_window is the parent of window
which may or may not be explicitly tracked by an E_Client, so the wheel events here
should be sent to the parent as is done in mouse button events
fix T2604
canvas grabs changed completely in 1.15, and so it's required that
x11 grab handling also have special runtime cases for this
for such versions the compositor must:
* always grab the internal client window instead of the parent
* always ungrab the window when it has focus
it seems that some clients, eg. libreoffice, don't set the modal window
property on child dialogs. instead of fighting for focus, set up the child
as a modal on the parent and then avoid the whole issue
fix T2594
a client with this flag set here is unreliable to use as a stacking
reference since it has yet to be stacked and can be located anywhere
in the window stack.
fixes internal window stacking on startup
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
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
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)
Summary:
Some implementations of the X server (e.g. Xquartz) define their $DISPLAY
as a path. Since ecore_ipc_server_add() does not create non-existant
directories, and since it may not worth to hide the socket in a complex
path, this patch aims at reducing the $DISPLAY by only keeping its basename.
Reviewers: zmike
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D2465
it's somehow possible to get multiple instances of a module in the config.
since I don't know whether this is a current or past problem, prune the list
on every module load to remove subsequent instances of the module
fix T2181
delfn's on desks being referred to as profile.wait_desk were never
cleared when client was freed. als the desk itself never got an obj
reference - fix that and ref/unref it.
es->cfg->desk_show_mode if false (default) implies a shelf is always
visible - true, as it's sticky (visible on all desktops). but this
didn't check zone. check zone first THEN this.
sometimes while quickly moving the cursor through menu items with submenus
the visibility flag would toggle more quickly than the menu object's visibility.
this is no longer the case
fixes issues where a new child added would appear above other child objects
if the new child was meant to be stacked below other children
mostly noticeable in deskmirror
moderately certain I originally wrote this to work in the other direction
and then failed to remove it when I switched to setting parents instead of
children. regardless, pixmap hash should not be changed here
this speed sup dir listing in efm drastically. first the e fm back end
uses an io thread that just spools through everything fast and sends
it over the mainloop to then send by ipc to e.
and on the e side we no longer use the heavy file internal magic using
api calls that wander all over a file for magic numbers - this is
insanely slow and brings listing to a crawl.
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.
backported to avoid future backport conflicts
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 previous patch(es) had a number of issues which made them unsuitable for general use:
* only checking "/usr" and "/usr/local" paths, despite this only being accurate if e was installed into /usr or /usr/local
* only checking if the paths were at the beginning of the string, when it's possible that they could be anywhere
* failure to also check XDG_CONFIG_DIRS
* improper formatting: this is a bit of a nitpick, but there are no correct instances of 'strcmp(a, b) == 0' in the e codebase.
@fix
this should fix T1988. after some looking here is what i think is
happening. he is asking the system (sysvinit or systemd) to
halt/reboot, but these end up killing the e process before it can save
its config that is lurking around still in ram as opposed to on disk,
so puttting in a config flush before beginning such a system action
will ensure everything is on disk.
init of e_screensaver sets up an event handler for the
powersave_update event (which would be 0 unless powersave is init
first).
Thanks to dt9 for the report ! ;)
Signed-off-by: Chris Michael <cp.michael@samsung.com>
This reverts commit 4aef218d28.
Revert "randr: remember crtcid for fast lookup on reconnect"
This reverts commit 55b4ad41d1.
Revert "randr: further cleanup"
This reverts commit 3e02824663.
these make nvidia randr resolution changing work again. sorry seb.
even though cleanups might be nice... this code is FIDDLY and every
driver seems to work differently, so take it easy and be very careful
- make no assumptions. i know it works on my intel desktop and nvidia
deskop atm. once i get a vga cable i'll fix up intel laptop with
external display that has problems. i can try radeon too later. but
right now - this stuff needs to work. being clean is far less
important than working. :)
When an output is set inactive, it is disconnected from a crtc. Now we
remember the id of the last crtc it was connected to, so we get a faster
lookup on reconnect.
the dp checks were looking for edp stuff. but due to naming they appeard as
eDP or maybe edp or EDP - thus were simplified to dp. put more explicit
checks there.
Primarly two things removed:
1. crtc->outputs should only contain outputs which should be enabled
So looping them twice in e_randr_apply is definitly wrong.
2. _e_randr_crtc_find should find a crtc for an output
So looping our crtcs again in _active_set is definitly wrong. If
_e_randr_crtc_find does not find us a crtc, it should be fixed there.
i've spent a while testing randr on my nvidia box and drviers. it now
works - i can turn screens on and off. i couldn't before. e_randr was
trying to be too smart for its own good. it wasnt matching up outputs
to crtcs, was thinking outputs active when they were not, etc. etc.
etc. - reality is e_randr could do witha lot of simplification, but
at this stage i'm just going to hammer it into shape to work across as
many drivers and setups as i can test against.
Summary:
Otherwise the mouse up event is emitted to the evas but the evas does
not work with it, because it does not know the window it is from. This
leads to a evas_event_down_count_get bigger than 1 which leads to a
missing mouse_out event. With this missing mouse_out event _post_up_handle
in evas_event.c is not called anymore, this means all the click events
are getting emitted to the object which was last clicked.
This fixes T1894.
Test Plan: Drag and Drop a icon somewhere and click somewhere after that, the evas will not get stuck anymore
Reviewers: raster
Subscribers: cedric, abyomi0, zmike, raster
Maniphest Tasks: T1894
Differential Revision: https://phab.enlightenment.org/D1771
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
so e was storing randr config using XIDs to match outputs etc. this is
all kinds of wrong. XIDs are NOT STATIC. they change from xserver to
xserver and from run to run. they MAY be the same. they may not. so
this was just broken.
use output name + edid as a big "string" (name.edid) as a way ofr
identifying config for a specific combination of output plus monitor
and to find/identify the corrent output+monitor to apply it to (of
course missing edid gets replaced with ??? and missing output name is
??? too - i have never seen a missing output name so you get this at
least).
so this FIXES "restore" of screen mode on login for starters. this
does nothing to "fix" the screen setup dialog in any way. there are
separate issues there.
this also breaks e_randr config compat so i bumped epoch so your old
config is rejected. i don't see a sensible way of porting the config
forward.
Summary:
The loop E_CLIENT_REVERSE_FOREACH was used to iconify and uniconify the
icons. The worked well for uninconify.
But if the first client with the focus gets iconifyed the lower one in
the stack will get the focus and will be raised to the top, but the loop
will continue with the next of the now iconifyed e_client, so this one
will be skipped.
Now the forward loop is used to iconify and the reverse to uniconify the
e_clients.
This fixes T1797
Reviewers: zmike
Reviewed By: zmike
Subscribers: cedric, cippp
Maniphest Tasks: T1797
Differential Revision: https://phab.enlightenment.org/D1655
This reverts commit b41dbbe9cf.
Revert this ... it works, but it's not the "proper" fix as it just
causes the crash(s) to happen elsewhere ... time to dig deeper
This was a cause of some memleaks/crashes in the wayland compositor
because the compositor was trying to access properties of the E_Pixmap
after it had already been freed. By setting the user_data to NULL, the
functions in the compositor which were failing will now stop trying to
access the pixmap after it's been freed.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
We are checking for the various mount options here given by configure.
-Wundef showed that we might not have the defines right so better
include config.h here.
all eina hash functions are now useless for any case where the hashed value needs to be stored, so it's necessary for e to provide its own hash function to use for password storage
When running in Wayland only mode we don't have XCB available so a
xcb_connection would fail and we would restart endlessly.
I have been bitten by that often enough now. Better shutdown E cleanly and don't
use the watchdog functionality for now. This needs a wayland specific solution
anyway.
when doing wayland windows (internal dialogs), and in such a case
don't free the window pointer on window hide/destroy.
NB: This fixes the case when running wayland-only and dialogs would
actually show 2 pointers.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
wayland-only and also not working with an X client.
bugfix: Fix case of pointers_size_set not adjusting pointer structure
fields with new size when rendering wayland pointers.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Summary: When we commit a subsurface from cache, we should adjust the
client's shape rectangles so that they get applied when rendering.
Signed-off-by: vivek <vivek.ellur@samsung.com>
Reviewers: devilhorns
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D1412
Summary: mobile policy module provides user interface for small mobile devices like phones.
Test Plan: N/A
Reviewers: raster, devilhorns, zmike
Subscribers: seoz, zmike, cedric
Differential Revision: https://phab.enlightenment.org/D817
Summary: since maximized valuable is used in _e_client_frame_update, so the valuable should be updated.
Test Plan:
(1) run any application.
(2) Maximize the app by key binding.
(3) Change the border type.
(4) Unmaximize the it by key binding.
(5) try to maximize it again, but couldn't
Reviewers: zmike
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D1357
Summary:
Because attached buffer can be NULL for removing surface content,
check of pending.buffer should be done before getting resource value of the buffer.
Reviewers: devilhorns
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D1370
This fixes an issue where internal dialogs were not using the e
cursors but would default back to an x cursor
Signed-off-by: Chris Michael <cp.michael@samsung.com>