Doyoun Kang [Wed, 15 Feb 2023 02:14:30 +0000 (11:14 +0900)]
remove handling enlightenment-tests in post script
Change-Id: I019cd74e344d355d8e878cc8aaa94814b91f2006
Doyoun Kang [Wed, 15 Feb 2023 01:21:29 +0000 (10:21 +0900)]
remove enlightenment-tests in enlightenment package
Change-Id: If6a8c66c1f88d366011d349f9071362dc3d59d54
duna.oh [Fri, 9 Dec 2022 05:42:46 +0000 (14:42 +0900)]
e_devicemgr: handle init_generator_with_sync request
init_generator_with_sync req. is available since tizen_input_devie_manager v.5
e_devicemgr handles this request by sending error_none event when device
is added in E20 and ready to generate input events.
If device is not ready in timer's duration (1 sec), e_devicemgr sends errors.
Change-Id: Ia8412114af58c1f71839c8a111d00fa639f2be9c
Doyoun Kang [Mon, 13 Feb 2023 11:15:46 +0000 (20:15 +0900)]
e_process: check internal window while handling activate/deactivate process
There was a bug that the status of process didn't change to DEACTIVATE when the internal ec got
focused.
For resolve this problem, we check the focus ec is internal ec or not. If focus ec is internal,
then e changes the previous activate process' state to DEACTIVATE.
Change-Id: I62961438c61cfc0685c595e93c9406c41eed963b
Changyeon Lee [Mon, 13 Feb 2023 05:37:36 +0000 (14:37 +0900)]
e_util_transform: fix flickering when client attach null buffer
client attach null buffer when it want to unmap
size of ec is updated by size of buffer and core transform use
size of ec.
so ec is transformed by 0x0 size when window is unmapped and
core transform is enabled.
it cause flickering frame when client attach valid buffer after
it is unmapped.
ex) window is 1920x1080 and core transform size is 3840x2160
1. client attach null buffer and it is umapped
2. transform is updated by 0x0 size because size of ec is 0x0
3. but ec is visible because hide effect
3. client is rendered with transform 0x0 size
4. client attach 1920x1080 size buffer and transform is updated
by 3840x2160
5. client is rendered with transform 3840x2160 size
Change-Id: Icd74973ae3bd18fda47fcadc47d9fad37bc667a5
Doyoun Kang [Mon, 13 Feb 2023 07:02:04 +0000 (16:02 +0900)]
e_comp_wl.h: remove unused function
Change-Id: I020264804229c39426ed21daab7c92ec8a3f7896
Doyoun Kang [Mon, 13 Feb 2023 02:03:56 +0000 (11:03 +0900)]
e_comp_wl: refactor code to handle sync_geometry_set
We refactor code to handle the tizen_move_resize_set_geometry request.
We handled it using the serial data of buffer, but there was a case not to set the serial data.
So, we changed code that e compares the buffer size and requested geometry size instead of the
serial data.
Change-Id: I0994f382797de3108e9c6c3645b0449f750b4832
Changyeon Lee [Mon, 6 Feb 2023 11:36:05 +0000 (20:36 +0900)]
e_comp_object: fix the size of effect object
the size of effect object should be set with the size of comp object
when map of effect object is disabled, not the size of buffer.
because the size of buffer can be different with the size of comp object
when client attach buffer which is not matched comp object.
Change-Id: I609f29db78d666f5efe18a39d1f295fb23a6af02
Doyoun Kang [Mon, 6 Feb 2023 07:23:02 +0000 (16:23 +0900)]
e_desk: check code whether ec is deleted in e_desk_geometry_set
Change-Id: I262d85056fe9a44de5f84626e055daafb906df33
Doyoun Kang [Wed, 1 Feb 2023 09:51:46 +0000 (18:51 +0900)]
e_policy_visibility: check bg_state before calling uniconify_render_pending
Change-Id: I83a38396dcae98aa019cd264c91ad1b8ccec00a2
Changyeon Lee [Thu, 19 Jan 2023 10:00:42 +0000 (19:00 +0900)]
e_comp_wl_tizen_hwc: add wait_commit to log of feedback timeout
Change-Id: I1fcada3c3cd323fde9610324189ffcaacc75c4b7
Doyoun Kang [Wed, 25 Jan 2023 07:46:54 +0000 (16:46 +0900)]
e_comp_wl: modify code sending configure event in ec->frame resize callback
We changed code not to send configure event if the buffer size is same to the ec's size in ec->frame's
resize callback.
Change-Id: Iea5bc6b59ec1acf8a24a03fdf042d8bd0e0327c2
Changyeon Lee [Mon, 16 Jan 2023 05:21:06 +0000 (14:21 +0900)]
e_hwc_windows: fix missing set deactivate of hwc_window
hwc_window should be deactivate in case of hwc window
does not have queue and it is not DEVICE state
Change-Id: I43c1332ff0cf6ab04d7818695f5fdb43e1f42e79
Doyoun Kang [Tue, 10 Jan 2023 04:50:35 +0000 (13:50 +0900)]
e_policy_wl: add a new aux hint for resizing window by ppu
We add a new aux hint value for ppu.
- wm.policy.win.resize.ppu
Change-Id: I563f4621978bcb37d8ce8410b6e0e615350e6c21
Doyoun Kang [Tue, 10 Jan 2023 01:48:23 +0000 (10:48 +0900)]
e_client: support ppu for resizing window
We add a new feature that the window is resized based on the ppu.
If a user sets the ppu value to window and requests to resize window by display server,
e adjusts the window's size according to the ppu value.
Change-Id: Idb7c3d77b9760ca60c4a3d57028af7e53ddb15bd
Junseok Kim [Thu, 5 Jan 2023 03:53:16 +0000 (12:53 +0900)]
e_xdg_shell_v6: Set wait_next_commit flag to True when min/max_size_set
There was a bug that min/max size set didn't work even the client send wl_surface_commit after min/max_size_set
The commit callback didn't work because of unset of the wait_next_commit flag.
And it causes the min/max size didn't adjust to the toplevel surface.
For resolve this issue, set wait_next_commit flag to True when doing min/max_size_set.
Change-Id: If8534017bdae2d19f45f34b06620f0ca03cc9b2d
Changyeon Lee [Wed, 4 Jan 2023 09:03:07 +0000 (18:03 +0900)]
e_hwc_window: handle invalid case of releasing the usable queue buffer
if client attach general buffer after the queue buffer is attached,
the usable queue buffer is released and client can not get the queue buffer
for next frame.
it is invalid sequence of client rendering but we should handle this case.
add checking the usable queue buffer is relased and unset queue of ec.
Change-Id: I31ae261974fb3b4d76bf40cfcfbb02d22dd6b36d
Doyoun Kang [Tue, 3 Jan 2023 01:12:29 +0000 (10:12 +0900)]
e_service_quickpanel: check qp_ignore_indicator_mode config while handling gesture event
If the qp_ignore_indicator_mode is set, then e doesn't send the gesture event to indicator
Change-Id: Ife43fffddd082fc9fcf9b4434b4068d9a1d49b38
Doyoun Kang [Mon, 2 Jan 2023 23:28:45 +0000 (08:28 +0900)]
e_config: add qp_ignore_indicator_mode configuration
If this configuration is set, the quickpanel service window can be shown regardless of the indicator's
state (such as shown, hidden, etc).
FYI,
default value is 0, it means that the quickpanel service can only be shown when the indicator is shown.
Change-Id: I33d13f2e63905d5032cbe60722dd7fc483c72e07
Doyoun Kang [Tue, 27 Dec 2022 10:40:21 +0000 (19:40 +0900)]
e_client: set above_show_pending flag if non alpha window while calculating visibility
Change-Id: If25a7129bd6f4f4edcd9ae869183f80ad75b6026
Changyeon Lee [Mon, 26 Dec 2022 05:10:20 +0000 (14:10 +0900)]
e_info_server: add client name to file of dump buffers
Change-Id: Ia5d68468201949b511af8b9908e24db0d9f1a855
Junseok Kim [Thu, 22 Dec 2022 07:03:11 +0000 (16:03 +0900)]
e_policy_wl: do not send allowed_aux_hint event when delete the aux_hint
The event that "tizen_policy_allowed_aux_hint" means the aux hint allowed, not the request allowed.
But the allowed_aux_hint event sent when the aux_hint_del proceed.
It seems misuse of event and client can understand as the deleted aux hint still allowed.
For prevent misunderstanding and correct the mean of event, don't send allowed_aux_hint when delete the aux hint.
Change-Id: I0d7c5fba29963f0638bc9832b9ca1dd5f34edf45
duna.oh [Tue, 13 Dec 2022 06:25:28 +0000 (15:25 +0900)]
e_comp_wl: add cursor_timer when mouse_wheel and mouse_down/up
add scenarios when setting cursor hide timer
1. when mouse wheel is scrolled.
2. do not set hide timer when mouse down occurs, and set hide timer
when mouse up occurs.
Change-Id: I3ff45dcd1ffcef9dff30acb9487992fe43e8cbe1
Changyeon Lee [Fri, 16 Dec 2022 08:12:59 +0000 (17:12 +0900)]
e_comp_object: fix continuously generate E_EVENT_CLIENT_PROPERTY
if ec has pending geometry but client does not send wl_surface_commit
request, E_EVENT_CLIENT_PROPERTY is continually generated with below
sequence.
1. pending_geometry and changes.pos of client is set true
2. changes.pos is set false and evas_object_move is called in
_e_client_eval(it is called in idle_enterer_before)
3. changes.pos is set true in _e_comp_intercept_move because
ec has pending geometry
4. E_EVENT_CLIENT_PROPERTY is added in _e_client_eval
5. ecore main loop is wake up with E_EVENT_CLIENT_PROPERTY
6. step 2 is repeated
Change-Id: Id817e13569759c0807188e2bbd342274454673ae
Seunghun Lee [Thu, 15 Dec 2022 08:16:10 +0000 (17:16 +0900)]
video: Move video update job to E_MAIN_HOOK_POST_CLIENT_IDLER_BEFORE
This patch is to change the order of jobs like this:
Old:
video update job -> e_client_idler_before() -> hwc commit job
New:
e_client_idler_before() -> video update job -> hwc commit job
The problem with the old way is that if the E_Client for video gets
updated for some reason in the process of e_client_idler_before(), then
video update job couldn't get a chance to be up to date with it and hwc
commit job would end up getting done without the update.
It would be obviously better if we could commit all changes to onscreen
at once as far as possible. And this patch is made to achieve it.
It is necessaty to add a new hook point called
E_MAIN_HOOK_POST_CLIENT_IDLER_BEFORE in e_main, because there is no way
to insert a video update job handler between e_client_idler_before() and
hwc commt job,
The callbacks associated with the hook point will be called right after
the call e_client_idler_before().
This way, video update job will be able to get a chance to be up to date
with changes which might be made in the process of
e_client_idler_before().
Change-Id: I857e0efcb12e7a06c18921bdfc6e6077fd5f521d
Seunghun Lee [Thu, 15 Dec 2022 06:19:16 +0000 (15:19 +0900)]
e_client: Appy base output resolution to subsurfaces
This patch is to apply base output resolution of toplevel window to all
subsurfaces. So, it removes call to e_comp_wl_hook_del() and continue to
listen E_COMP_WL_HOOK_SUBSURFACE_CREATE hook point.
Some wayland clients repeatedly reuse wl_surface to create wl_subsurface
with a different parent from previous one. The problem is that the
subsurface can't get applied with base output resolution of a new
toplevel window in this case.
To apply base output resolution of a new parent to all associated
sub-surfaces, we should call e_client_transform_core_update() for a
toplevel window whenever a subsurface gets created with a new parent.
Since e_client_transform_core_update() recursively updates transform for
all sub-surfaces, calling e_client_transform_core_update() for a
toplevel window should also update transform for all subsurfaces.
Change-Id: I3253f83e0ab1cf1f5381bbfba2693d9d642d88ca
Doyoun Kang [Fri, 9 Dec 2022 02:57:42 +0000 (11:57 +0900)]
e_policy_wl: set desk_iconify_skip to DESKTOP type window
Change-Id: If553ac1b182c2ee7451092460ec977dd25ca6aaf
Doyoun Kang [Fri, 9 Dec 2022 02:54:49 +0000 (11:54 +0900)]
e_client: add e_client_desk_iconify_skip_set/get APIs
We provide APIs for setting/getting a value to skip iconify by desk as below.
- E_API void e_client_desk_iconify_skip_set(E_Client *ec, Eina_Bool skip);
- E_API Eina_Bool e_client_desk_iconify_skip_get(E_Client *ec);
Change-Id: I680a0301d72c71822bf8b83c3fdf0f203559cdc0
Junseok Kim [Wed, 7 Dec 2022 06:46:36 +0000 (15:46 +0900)]
e_client: check ec->frame is visible when finding focusable ec in topmost focus policy
There was a bug that focus given to the iconified window that fully obscured by evas object hide (not unmap, called by E20 module).
When the evas object hide called, the window was uniconic state so that state of visibility client still be E_VIS_ICONIFY_STATE_UNICONIC.
But, the visibility of the window turn to fully obscured, and iconified by visibility.
Finally, the window has property as below:
- stack was topmost window of normal layer
- focused: true
- visibility.obscured : fully obscured (2)
- vc->state : E_VIS_ICONIFY_STATE_UNICONIC (1)
- iconic : true (1)
- exp_iconify.by_client : false (0)
- ec->frame visible : false
- ec->mapped : true
The window will be show again by next visibility calculate, but the window have to loose their focus in this time.
for fix this issue, do not give focus to the window when the evas object is not visible.
NOTE: This is only for topmost focus policy.
In focus history policy case, the window will be show soon so that the window can take focus.
NOTE2: This patch replaces the patch
2b0f37d491a700fbec56e91618825cc2dedd5585,
which patch is prevent issue in topmost focus policy.
Change-Id: I21a0e8fd8610f7aa878e616cfa671ea49c5811e9
Changyeon Lee [Wed, 7 Dec 2022 10:38:36 +0000 (19:38 +0900)]
e_comp_wl: add log of gap idle_exiter between prepare callback of wayland event loop
if the job of main thread has much time or does not get CPU scheduling, the synchronous call
(ex: wl_display roundtrip) is blocked until the job is done. it is the regular operation of
a synchronous call but sometimes it has too much time.
this patch is for debugging it.
Change-Id: Ife0d9f0ad32807fd892c7b2f991d966427c0c72c
Doyoun Kang [Thu, 8 Dec 2022 23:43:45 +0000 (08:43 +0900)]
e_policy_visibility: change vc->state to iconic while canceling grab
If the E_Vis_Client is under running uniconify_render, then we change vc->state value
to E_VIS_ICONIFY_STATE_ICONIC in _e_vis_client_grab_cancel() function.
Change-Id: Ib67e414bb1243f42c23c47c10a7a167b903a51e6
duna.oh [Mon, 5 Dec 2022 03:54:39 +0000 (12:54 +0900)]
e_devicemgr_wl: set the list of 'dev->resources' to NULL after eina_list_free()
Change-Id: If9ee0cf2500c25e4a69dfbe763d2af6d11842955
Seunghun Lee [Tue, 15 Nov 2022 07:19:21 +0000 (16:19 +0900)]
video: Refactor the function for cropping viewport
This is about reconsidering variable and function names to enhance
readability.
No functional changes.
Change-Id: Ia8ea36d7c4af13992e17da82fbc38a7bbbd64351
Seunghun Lee [Thu, 1 Sep 2022 00:09:35 +0000 (09:09 +0900)]
video: Fix transform translation
The function _e_video_hwc_coord_transform() did actually translate the
90 and 270 degree rotations into the inverse of what it's supposed to
do.
Since it's to get the points in buffer-local coordinates from physical
screen-local coordinates, it has to translate 90 and 270 degree
rotations into inverse of buffer_transform().
Change-Id: Ib48fe10dd6213888bf563e5645f8954d34cf36e2
Seunghun Lee [Thu, 11 Aug 2022 05:43:32 +0000 (14:43 +0900)]
video: Handle negative positioned video
If video sub-surface has negative position, e_comp_wl_output_find()
returns NULL unless there is an output which has negative position.
So, it would eventually fail to calculate physical geometry to be
applied to physical device associated with the output.
To resolve this problem, here it uses e_comp_wl_output_find_all() to get
all outputs on which a E_Client is overlapped. And then it uses a first
output among the returned outputs because it's not ready yet to support
for displaying video on multi output.
The same is true of e_comp_zone_xy_get(). As the name suggests, the
e_comp_zone_xy_get() function returns a zone which contains given x, y
position. So it returns nothing with negative x, y position unless there
is a zone which has negative position.
For this reason, this patch uses a zone referenced directly by E_Client
instead of the value returned from e_comp_zone_xy_get().
Change-Id: I5c20927ab418e4194209790a1c5a4b8bc50ebb38
Seunghun Lee [Thu, 11 Aug 2022 05:21:19 +0000 (14:21 +0900)]
e_comp_wl: Add e_comp_wl_output_find_all()
The function returns an array containing all E_Comp_Wl_Output on which
given E_Client is placed.
It's been fine with e_comp_wl_output_find() so far, but if given
E_Client is the window placed on negative poisition, then
e_comp_wl_output_find() returns null unless there is a output that
has negative position which doesn't make sense.
Change-Id: I3ddb891803ff88ff0de48dbc58055ac190a2fa84
Changyeon Lee [Wed, 2 Nov 2022 10:46:32 +0000 (19:46 +0900)]
e_comp_object: defer move,resize,clear when render_update_lock is enabled
there is a visual bug when client move and resize buffer when render_update_lock
is enabled because evas_object is resized and moved but buffer of client is not
updated. (buffer of render_update_lock is fill to evas_image_object)
and buffer of render_update_lock is cleared in comp_object if e_comp_object_clear
is called so defer clear of comp_object until render_update_lock is disabled.
Change-Id: Iabe67a558c7ea62d0f84cd6c041f2f24b32a3c1b
Changyeon Lee [Mon, 31 Oct 2022 06:46:34 +0000 (15:46 +0900)]
e_comp_object: fix missing unrference buffer of render_update_lock
Change-Id: Iaf018573a808b95f7e69666552b981507445078a
Junseok Kim [Thu, 27 Oct 2022 08:49:24 +0000 (17:49 +0900)]
e_policy_wl: Add new aux hint for set visibility.ignore_geometry
Added new aux hint for set visibility.ignore_geometry property of the target client.
hint: "wm.policy.win.visibility.ignore_geometry"
val: "1": set ignore_geometry / "0": unset ignore_geometry
Change-Id: Id912cc1bf2309dabfe77cd1a9c9bb2000c01cb77
Junseok Kim [Thu, 27 Oct 2022 06:09:00 +0000 (15:09 +0900)]
e_config: Add configuration to adjust visibility.ignore_geometry to all windows
For adjust visibility.ignore_geometry to whole windows, added the configuration that calc_vis_ignore_geometry.
When the calc_vis_ignore_geometry is True, all of the new E_Client set the visibility.ignore_geometry to True.
It is not fix each client's visibility.ignore_geometry value to configured value.
each client can change their value at any time.
Change-Id: I7804183a7bb340d7b725eb5b198ba00994123bc1
Junseok Kim [Thu, 27 Oct 2022 06:03:02 +0000 (15:03 +0900)]
e_client: Add ignore_geometry property to E_Client to ignore geometry calc when vis calc
Ignore_geometry is property for ignore geometry calculation during calculate visibility after once the window show and the property set.
The window that set visibility.ignore_geometry always retain unobscured even if the window is stack in lowest after client attach buffer that not NULL.
It makes the client to do something their jobs under other client window.
Change-Id: Idce08255223d451cbd39b09166849a45cc9586c3
duna.oh [Thu, 17 Nov 2022 02:22:57 +0000 (11:22 +0900)]
devicemgr: send max_touch_count events to all tizen_input_device_manager bound
This patch is modification of the following commit.
-------------------------
commit
f57a73ce40323b8c2b1f5847375090165c52dcc6
Author: duna.oh <duna.oh@samsung.com>
Date: Thu Nov 10 14:34:33 2022 +0900
e_devicemgr: create tizen_input_devices only for first bound seat & manager
Change-Id: I92cc3a49f7419fe60352bbaddfdfce59d573a5ac
duna.oh [Fri, 11 Nov 2022 02:35:59 +0000 (11:35 +0900)]
e_devicemgr: add null checking and break from for-loop
Change-Id: I83d7b62195b21997cec9e0c242644506dc37e587
Junseok Kim [Fri, 18 Nov 2022 05:43:22 +0000 (14:43 +0900)]
e_client: Unset current focus when the new focusable ec found
There was a bug that couldn't receive focus changed event when client unobscured by activate using topmost focus policy.
For more detail, in topmost focus policy, issues following below sequence.
1. Focused window(winA) requests Activate below to fully obscured window(winB).
2. winA requests Activate before winB showing.
The Activate below request didn't doing uniconify render, so stack was changed immediately.
and it cause visiblity changes immediately and visibility changed event send to each client.
The winB didn't focused because of the winB is not shown till the uniconify rendering by visibility.
in this case, winA still focused but winA is not focusable window before winB shown or winA Activated.
It can cause the client didn't receive focus changed event after the client unobscured.
(client regards the fully-obscured means loose focus)
for fix this problem, always focus unset before focused window when there're new focusable window.
Change-Id: Iaccfa5ed77bd46f58a4f82acd9109e534da54606
Doyoun Kang [Thu, 17 Nov 2022 04:42:00 +0000 (13:42 +0900)]
e_client: modify return x and y values of e_client_geometry_get()
If the ec->frame's size are 0x0, it means that the geometry of ec->frame is invalid.
In this case, e_client_geometry_get() returned the size value using ec's size itself,
but position value returned by invalid value.
This patch changes the returned geometry of e_client_geometry_get() to ec's geometry
itself in this case.
Change-Id: Ifb465d4d78fb02a0f207b46fd939f03e6433bf4b
Changyeon Lee [Thu, 17 Nov 2022 10:22:21 +0000 (19:22 +0900)]
e_comp_wl_tizen_hwc: reduce hwc_commit_feedback timeout to 5 second
Change-Id: If3e7f03e651ade360d8513611dd52b41f749586c
Junseok Kim [Fri, 11 Nov 2022 01:47:34 +0000 (10:47 +0900)]
e_client: Modify to have focus even the window is in defer_focus_stack
There was a bug that window couldn't get focus even if the window requests activate.
For more detail, there're 2 windows, that one is partial sized, above layer window(winA) like as floating window, and the other is full sized, normal layer window(winB) like as homescreen.
First, winB gets focus by touch as launching new app icon of the user.
By the user operation, a new fullsized, normal layer window (winC) launchs with create, show, and activate request.
Next, the winB fully obscured by new window winC.
In this timing, the winC is not visible but winB prepares hide by wm.
When doing hide, the e_client_revert_focus called for find previously focused window.
The winC is in defer focus stack but it skiped due to it isn't visible yet.
Finally, the winC couldn't get focus even if it requested activate and focus given to winA.
For fix this problem, check the window is in defer_focus_stack when focusable check.
Change-Id: I34e68d41d2cfb183d4eab956abf6472708e295bf
Junseok Kim [Tue, 15 Nov 2022 12:28:35 +0000 (21:28 +0900)]
e_client: do not focus to hidden window
There was a bug that focus given to topmost not unmapped hidden window.
It caused the window hide pended by e20 module, and the window requests activate.
for fix this issue, check the evas_object of the window is visible when focusable check.
Change-Id: I7a3d8c467ea3c39b1d0d77c5692877bcbf9e4f0d
Doyoun Kang [Tue, 15 Nov 2022 02:10:44 +0000 (11:10 +0900)]
e_client: add e_client_pos_set before calling defer_move in e_client_unmaximize
Change-Id: Iefe344d88a1dc3c6c23541caec6d390abed30cf5
Doyoun Kang [Tue, 8 Nov 2022 08:27:23 +0000 (17:27 +0900)]
support a new feature for maximize direction
We add a feature for maximize direction.
- New enumeration : E_Maximize_Direction
- Implement a function for tizen_policy_set_maximize_direction
Change-Id: Ibb7741babcf050fd3f3d73293a0e2a4ea1f2f361
Junseok Kim [Thu, 10 Nov 2022 05:14:00 +0000 (14:14 +0900)]
e_info: Add functionaility for add and delete aux hint of window
Added new functionality for add and delete aux hint of window
enlightenment_info -aux_hint add [win id] [hint id] [hint] [val]
- Add or replace aux hint of the window that matched hint id.
enlightenment_info -aux_hint del [win id] [hint id]
- Delete aux hint of the window that matched hint id.
Change-Id: I7c9ac4f9b2bbd7f36fc8a4e3843bda093fde1228
duna.oh [Thu, 10 Nov 2022 06:36:55 +0000 (15:36 +0900)]
e_devicemgr: use different Eina_List variable for nested for-loop
Change-Id: Iff0453d310d2e84ce01e8583e82046c99c326521
duna.oh [Thu, 10 Nov 2022 05:34:33 +0000 (14:34 +0900)]
e_devicemgr: create tizen_input_devices only for first bound seat & manager
If a wayland_client uses ecore_wl library and efl_util API,
it binds two pairs of wl_seat & tizen_input_device_manager.
ecore_wl listens wayland events (ex. device_add/device_remove) and send
requests properly. (ex. tizen_input_device_release)
However efl_util dispatches event queue only when efl_util API is called.
It doesn't send tizen_input_device_release requests and causes resource leaks.
To fix this resource leak,
this patch is to create tizen_input_device resources only for
the first bound wl_seat & first bound tizen_input_devcie_manager
which ecore_wl binds.
efl_util always binds wl_globals after ecore_wl2 binds ones.
Change-Id: I47c741a36c2205e2bde2a630a4b66d3794aa23ef
duna.oh [Mon, 7 Nov 2022 08:20:01 +0000 (17:20 +0900)]
e_devicemgr: fix the memory leak detected by static analysis tool
Change-Id: I9733fb3941c67d6975f2d1572e0ecebacb86db63
duna.oh [Mon, 7 Nov 2022 08:22:08 +0000 (17:22 +0900)]
e_devicemgr: add error logs when failing to alloc structure
Change-Id: I27ac7b2f3e47be7e963f090130a575d388bcd43e
duna.oh [Fri, 4 Nov 2022 10:15:47 +0000 (19:15 +0900)]
e_devicemgr: add data struct for tizen_input_device_manager
E_Devicemgr_Input_Device_Mgr_Data is a data for tizen_input_device_manager.
This has a Eina_List of E_Device_Input_Device_User_Data and manager's resource.
When a client unbinds tizen_input_device_manager, mgr_data iterates a list
of device_user_data and cleans up device's resource.
Change-Id: I7fa0650a1cf25b177b2a1a88dc6e1b8f80ad4409
Doyoun Kang [Sun, 6 Nov 2022 01:28:55 +0000 (10:28 +0900)]
Revert "e_client: modified code for e_client_unmaximize"
This reverts commit
51cba1f7c4c8cf2fa3508d191c2ba54e9dc3d5ec.
Change-Id: I1960204996f99374ac5b7e725b9462afc3505363
Seunghun Lee [Mon, 7 Nov 2022 08:18:13 +0000 (17:18 +0900)]
viewport: Do not use evas_object_event_callback_del()
For short, the `evas_object_event_callback_del_full()` should be used
instead of `evas_object_event_callback_del()`.
The crash happened as the backtrace shown below, and it seems that the
crash was due to the misuse of `evas_object_event_callback_del()`.
This patch also log more about the viewport just in case the patch may
not fix the crash.
backtrace:
(gdb)bt
#0 0x000d6e98 in e_object_is_del (obj=0x6) at /usr/src/debug/enlightenment-0.20.0/src/bin/e_object.c:106
#1 0x00086088 in _e_comp_wl_viewport_cb_parent_show(data=0x1d50c70, e=<optimized out>, obj=<optimized out>, event_info=<optimized out>) at /usr/src/debug/enlightenment-0.20.0/src/bin/e_comp_wl_viewport.c:572
#2 0xb5d91df6 in _eo_evas_object_cb (data=0x207e4c8, event=<optimized out>) at /usr/src/debug/efl-1.25.1/builddir/../src/lib/evas/canvas/evas_callbacks.c:181
...
The evas_object_event_callback_del() is to remove the most recently
added callback from the object. Having used it in `e_comp_wl_viewport.c`
could have caused unintentional removal of another callback depending on
the timing of creation and destruction of several viewports, and this
eventually could result in a crash.
I couldn't find the clear evidence that this really led the crash, but
let me give you an example what situation can make this crash.
1. There are 2 viewports created with the same parent. So, 2 viewports
would add a callback for its own. I will call both viewport as `A` and
`B`, and `A` is created before `B`.
2. The viewport `A` try to delete its callback from the parent evas
object because of the change of subsurface parent by calling
`evas_object_event_callback_del()`. And this will unintentionally
delete the callback of `B`, not `A` because the callback of `B` is the
most recently added one.
3. Let's say the viewport `A` gets deleted by `tizen_viewport.destroy`,
and it does not call the `evas_object_event_callback_del_full()` because
`viewport->epc` must already be null. And now, the data for the
callback `A` becomes freed memory, but the callback `A` is still in
callback list of parent object.
4. Event if the viewport `B` gets deleted by `tizen_viewport.destroy`,
it won't be able to remove any callbacks with
`evas_object_event_callback_del_full()` because the callback of `B` has
already been removed by the process 2.
As I mentioned, I'm not sure this really led the crash but I believe and
hope that this will fix the crash.
Change-Id: Ie92900df80d2351bbf9054fd3c4e9e6184b67d57
Seunghun Lee [Mon, 7 Nov 2022 08:17:09 +0000 (17:17 +0900)]
Revert "viewport: Cleanup setting parent and more log"
This reverts commit
3cd0f29865d9508cc511228f08fffc64ff46bfbd to re-write
commit message.
Change-Id: I9937d1050fba4cb1052d6a3b478e6bdc183ca70d
Seunghun Lee [Wed, 2 Nov 2022 08:50:12 +0000 (17:50 +0900)]
viewport: Cleanup setting parent and more log
The crash happened as the backtrace shown below but I couldn't find the
cause. So, this patch logs more about viewport to help find the cause of
problem next time if it will happen again.
backtrace:
(gdb)bt
#0 0x000d6e98 in e_object_is_del (obj=0x6) at /usr/src/debug/enlightenment-0.20.0/src/bin/e_object.c:106
#1 0x00086088 in _e_comp_wl_viewport_cb_parent_show (data=0x1d50c70, e=<optimized out>, obj=<optimized out>, event_info=<optimized out>) at /usr/src/debug/enlightenment-0.20.0/src/bin/e_comp_wl_viewport.c:572
#2 0xb5d91df6 in _eo_evas_object_cb (data=0x207e4c8, event=<optimized out>) at /usr/src/debug/efl-1.25.1/builddir/../src/lib/evas/canvas/evas_callbacks.c:181
...
Change-Id: I7be6210c608ca02d55d59a6c65c60a8bb04873ba
Junseok Kim [Thu, 6 Oct 2022 03:10:45 +0000 (12:10 +0900)]
e_policy_wl: calculate aspect ratio including header and footer height
There is a windows that have their own decorator that header and footer.
And the window need to resize using aspect ratio to the content except the fixed decorator size.
for those windows, added aux hint to set their header and footer height and adjusted to calculate aspect ratio resize.
When the client sets the aux hint, then the enlightenment calculates aspect ratio without header and footer height.
aux hint keys:
wm.policy.win.resize.header_height
wm.policy.win.resize.footer_height
aux hint values:
unsigned integer
Change-Id: I033c73197951c975145d2ecaa00bdc96e4e8f798
Doyoun Kang [Mon, 31 Oct 2022 07:26:47 +0000 (16:26 +0900)]
e_policy_wl: tizen_policy version up to 11
We update tizen_policy version to 11 for supporting desktop window type.
Change-Id: Ib630f8f3789137a5061d072459bbc6e4a3613e33
Changyeon Lee [Tue, 25 Oct 2022 04:35:11 +0000 (13:35 +0900)]
e_desk: do not deactive hwc when desk zoom is set
deactive of hwc is not needed because hwc does not use device state
when pp of hwc is set or desk_zoom enable state of client is set.
this patch is for supporting hwc in case of desk_zoom of client is
not appllied.
Change-Id: I7c9b04ff2559d8240373d3a334c16f37f137d905
Doyoun Kang [Thu, 27 Oct 2022 09:32:01 +0000 (18:32 +0900)]
e_client: add APIs for getting desk_zoom enable state
We add APIs to get/set the enable state whether the ec is under running desk_zoom or not.
- void e_client_desk_zoom_enable_set(E_Client *ec, Eina_Bool enable)
- Eina_Bool e_client_desk_zoom_enable_get(E_Client *ec)
Change-Id: I733d84fb0b96066577029aabe4724e9ce7e0ea09
Doyoun Kang [Thu, 27 Oct 2022 08:34:50 +0000 (17:34 +0900)]
e_client: use e_client_frame_geometry_set in e_client_fullscreen
We changed code to use e_client_frame_geometry_set() instead of evas_object_geometry_set()
to move and resize window synchronously in e_client_fullscreen().
Change-Id: I9cf3385cbceeccad2b25f3aaba4b8a71e3140625
Doyoun Kang [Tue, 25 Oct 2022 10:37:44 +0000 (19:37 +0900)]
e_policy: add E_POLICY_HOOK_CLIENT_AUX_HINT_CHANGED hook
Change-Id: I9a824d37f055ccd7161d87aba00ad5579140f1c1
duna.oh [Tue, 25 Oct 2022 04:48:25 +0000 (13:48 +0900)]
Revert "keyrouter: replace eina_safety error log with e_log"
This reverts commit
c82b257197071559463b026880fdd6c912e096d3.
Change-Id: I336deac6e0c9334fe3cce7a4949e7fb1f2f963ba
duna.oh [Tue, 25 Oct 2022 04:48:12 +0000 (13:48 +0900)]
Revert "keyrouter: simplify chained if-statements with AND operator"
This reverts commit
672574a5482de9419e372108762b1d9983df171c.
Change-Id: I9cc0e4a92d9681d01fb44d702e6db2a60f04e3ec
duna.oh [Tue, 25 Oct 2022 04:47:53 +0000 (13:47 +0900)]
Revert "keyrouter: add pname(process name) of client in logs"
This reverts commit
51b8d115c91919115b116bcf5275a9a150d70fde.
Change-Id: I35af6f27219d7fb230b0064315b91d4bb5c4bda0
Doyoun Kang [Fri, 21 Oct 2022 05:57:37 +0000 (14:57 +0900)]
e_client: modify code to call maximize_pre/unmaximize_pre callback
There is a bug that the maximize_done/unmaximize_done evas_object_smart_callback
doesn't call after calling maximize_pre/unmaximize_pre evas_object_smart_callback
in some case.
This resolve this problem
Change-Id: I6d30dc2897b441574c92f46d68e77e3501e412b1
Junseok Kim [Mon, 17 Oct 2022 06:03:42 +0000 (15:03 +0900)]
e_msg: added explicit type conversion to limit length
Change-Id: I5f448d8e88b385de627b6c4c782da5ae7cd41222
duna.oh [Tue, 18 Oct 2022 01:45:44 +0000 (10:45 +0900)]
keyrouter: add pname(process name) of client in logs
Change-Id: I141430ee891e49a953c2ab11a2c52c34b4b8ccfe
duna.oh [Tue, 18 Oct 2022 01:10:24 +0000 (10:10 +0900)]
keyrouter: simplify chained if-statements with AND operator
Change-Id: I4117fcb3fae1284906b30ce6d657e47b7efc3e22
duna.oh [Tue, 18 Oct 2022 01:00:19 +0000 (10:00 +0900)]
keyrouter: replace eina_safety error log with e_log
provide more logs about the client which is exclusively
grabbing the key.
Change-Id: Ia4e14735cb740110204123fc3a0f9766be279c96
Doyoun Kang [Wed, 19 Oct 2022 06:28:31 +0000 (15:28 +0900)]
e_client: modified code for e_client_unmaximize
We changed code to use e_client_util_move_resize_without_frame() instead of
e_policy_visibility_client_defer_move().
An e_client_util_move_resize_without_frame() has a functionality for synchronizing
move and resize.
Change-Id: Idb7d9eee65f0f13fa579b835c8383f1e08e8b99e
Changyeon Lee [Mon, 17 Oct 2022 03:32:59 +0000 (12:32 +0900)]
e_comp_object: add checking null for reducing error log
image object and shadow object can be null because it is created
in intercept show callback.
Change-Id: If83bba9c35abeca0d71d59a5f9416717e2f926e3
Changyeon Lee [Mon, 17 Oct 2022 07:02:06 +0000 (16:02 +0900)]
e_comp_object: change log of mismatch between pixmap and frame size to info level
pixmap and frame size can be misatch for reducing memory in case of video subsurface
Change-Id: I8dffb3cd68b814af26b7d3283eeae257316935fc
Changyeon Lee [Mon, 17 Oct 2022 03:29:51 +0000 (12:29 +0900)]
e_hwc_window: remove error log in case not implemented case
Change-Id: I62666c3542ea0f981b1232f9c9ad8845cd8df766
Junseok Kim [Fri, 14 Oct 2022 07:46:39 +0000 (16:46 +0900)]
e_pixmap: changed eina_safety to if clause to reduce error log
Change-Id: I5f78bda6bc244ec69eef254ede83c2d8db75a603
Junseok Kim [Fri, 14 Oct 2022 02:06:53 +0000 (11:06 +0900)]
e_client: changed eina_safety to if clause to reduce log
Change-Id: I48f5467c8e318d21798bf6d21888f79b8dffb345
Doyoun Kang [Wed, 12 Oct 2022 23:20:18 +0000 (08:20 +0900)]
e_policy_visibility: fix bug for checking a visible state of transient_child
There was a bug that the parent window was uniconified even though its child was not visible.
This was because that the child was uniconic state but its visibility was unknown state.
(this means it was not calculated visibility and it was not shown yet).
To resolve this problem, we add code to check whether child's visibility is unknown
in _e_policy_check_transient_child_visible function.
Change-Id: I80604558afa86fac020e0681a3bcdb8adb75bb84
Junseok Kim [Tue, 4 Oct 2022 10:14:57 +0000 (19:14 +0900)]
e_client: restore the code for resizing mode
There was a bug that the resizing didn't uses the mode that client requested.
It caused by resize_mode did not applied to ec->resize_mode.
For fix it, restore resize_mode code.
And accoding to the policy, the Non-floating window can also resize with interactive resize.
So, removed checking the floating window from legacy code.
Change-Id: I750e4069701de0b144faf1a84cc3b56f5489dcf6
Doyoun Kang [Wed, 12 Oct 2022 01:39:03 +0000 (10:39 +0900)]
winfo: add -desk test option for iconify_toggle
Change-Id: I8d1ff2c87b6d8b1123b84bb8ec73ba126a80dab5
Usage: winfo -desk iconify_toggle
Doyoun Kang [Wed, 12 Oct 2022 01:37:11 +0000 (10:37 +0900)]
e_desk: add new API for getting iconified list
E_API Eina_List * e_desk_visible_client_iconified_list_get(E_Desk *desk)
Change-Id: I86540290f030454be702813a927b4bb652706ad7
Changyeon Lee [Thu, 6 Oct 2022 11:06:46 +0000 (20:06 +0900)]
e_comp_object: reference front buffer while render update is locked
front buffer should be reference for preventing tearing buffer
Change-Id: I449802b815120a85dcf8d05f2d54a07f80099893
Changyeon Lee [Thu, 6 Oct 2022 03:35:25 +0000 (12:35 +0900)]
e_comp_object: change e_comp_object_render_update_lock/unlock/lock_get to E_API
Change-Id: I85fe96e26c1257870a76c2ae40fc380b9def6a7b
duna.oh [Tue, 11 Oct 2022 01:06:54 +0000 (10:06 +0900)]
e_pointer: set e_pointer's object only when object is changed
Change-Id: I050179a6450b831507b8176b5c4362cf1cfd691a
duna.oh [Tue, 11 Oct 2022 01:00:10 +0000 (10:00 +0900)]
add E_POINTER_HOOK_SHOW/HIDE hooks
Change-Id: I919eac923e0e87e6d8daaf1e9c531d35b1d728b4
Changyeon Lee [Tue, 4 Oct 2022 03:03:25 +0000 (12:03 +0900)]
e_comp_object: add log when render update lock enabled/disabled
Change-Id: Ia5e77f3d4be851a2ac57df18f413ad3e5bb139ed
Changyeon Lee [Tue, 4 Oct 2022 02:55:19 +0000 (11:55 +0900)]
e_hwc: add log when norender enabled/disabled
Change-Id: Ib2b39c367614810cfa40326a1c741611b4505519
Doyoun Kang [Sat, 24 Sep 2022 04:06:33 +0000 (13:06 +0900)]
e_policy_wl: support interactive_move/resize_done event of tizen_policy
Change-Id: Iff61f5cee2b2e214188780112587822a3264c7a5
Changyeon Lee [Wed, 28 Sep 2022 03:51:40 +0000 (12:51 +0900)]
e_comp_wl: reference gbm_bo until imported tbm_surface is destroyed
imported tbm_surface can have invalid buffer handle
if buffer handle of gbm_bo is closed when gbm_bo is destroyed.
we do not destroy gbm_bo until imported tbm_surface is destroyed.
Change-Id: Ibbfbeb84aa8e6c17af6471e64571f7dd9f2423e5
Changyeon Lee [Wed, 28 Sep 2022 03:45:50 +0000 (12:45 +0900)]
e_info_server: support tbm_surface which is not wayland-tbm resource in dump buffers
Change-Id: Ib258db8a7b6339b2de8ef8a214fef485cccc5ee5
duna.oh [Wed, 21 Sep 2022 01:38:01 +0000 (10:38 +0900)]
e_comp_wl: move a code checking 'cur_mouse_action' value
If 'cur_mouse_action' is not null, E20 should not send events to clients.
So its value should be checked right before sending wayland events.
This patch allows the cursor to follow mouse movement when user is doing
mouse action such as resizing/moving windows).
Change-Id: Ia53521b1bd4d07400dba8280fad8cde3ea7cf9e4
Doyoun Kang [Tue, 20 Sep 2022 09:06:52 +0000 (18:06 +0900)]
e_desk: add E_Desk_Zoom_Animating_Type information
There was a bug that the disable zoom was not working while running zoom animation
first time.
We add E_Desk_Zoom_Animating_Type information to check an animation type which is
running currently.
For this information, we can control enable/disable zoom animation efficiently.
Change-Id: Ibced55bd644759515679ae4894bc55f326613302
Doyoun Kang [Tue, 20 Sep 2022 04:39:21 +0000 (13:39 +0900)]
e_desk: fix build warning
Change-Id: I1e91e5bc47ba02974a7abebc964340a61b54e2a4
Doyoun Kang [Fri, 16 Sep 2022 09:31:24 +0000 (18:31 +0900)]
support feature - iconify_all and restore_all in desk window
We support new feature to iconify/restore all visible windows in the desk.
For this we add new APIs as below.
- void e_desk_visible_client_iconify_all(E_Desk *desk)
: Iconify all visible windows in the desk except desktop type or user_skip_winlist set window.
- void e_desk_visible_client_restore_all(E_Desk *desk)
: Uniconify all visible windows which were iconified by iconify all windows request.
If a new window is activated after running iconify all windows, then the iconified window list is cleared.
So, there is no window which is uniconified when an user reqeust to restore window.
Change-Id: Iddaa8ee5f4680a580e6eade9060b946d144f476a
Doyoun Kang [Wed, 14 Sep 2022 08:40:48 +0000 (17:40 +0900)]
e_policy_wl: doesn't send iconic state change event to unmapped window
When the window's visibility is changed to fully-obscured, then e sends
an iconic state change event to the window and its children.
But there was a bug that the child window which was not mapped didn't render
because its internal state was iconic.
So, we change code not to send the iconic state change event to unmapped window.
Change-Id: Ib0c38dd419874af7cf1b8b321a724204d5260cc2
Seunghun Lee [Wed, 14 Sep 2022 03:53:18 +0000 (12:53 +0900)]
e_client: Do not move a window to the center of zone
This patch is to fix that sub-surface gets placed wrong position which
is not the position a client wants.
It seems a legacy code of "desktop" window manager to move an
out-of-screen window to the center of zone.
For a desktop window manager, it makes sense because it's window
manager's responsibility to place window appropriately.
Plus, there is no way to move a window except user interacts with window
decoration, e.g, by dragging window title bar. Therefore, the window
manager should at least place movable windows to be visible to a user.
However, in Tizen, it's partly client's responsibility to move its
window, and there exists tizen_position wayland extension for it. So, we
should let clients move its window anywhere even if the position is out
of screen. Some windows like lockscreen already have worked that way
according to the comments which is removed by this patch.
Therefore, this legacy code doesn't make sense when it comes to window
manager for Tizen device environment.
Change-Id: Ia908cf892726a5a9022a1b3061355a0b42c6568a
Doyoun Kang [Wed, 14 Sep 2022 06:31:30 +0000 (15:31 +0900)]
refactoring: use new API instead of exp_iconify.by_client
Add new API for checking whether the window is iconified by client request or not.
- e_client_is_iconified_by_client
We replace code to use e_client_is_iconified_by_client() API instead of
ec->exp_iconify.by_client directly.
Change-Id: I823d3202d55dd8bc7ae220a016fafde544826487