these callbacks must be removed if there is no parent, otherwise they
may trigger once the widget is deleted and trigger invalid object access
@fix
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9429
this is kinda gross, but if the list is already deleted then the list
items are also gone and this is an invalid object access
@fix
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D9402
Summary:
this test is not runnable when it is not triggered directly from pressing
a button in the main window
Reviewers: devilhorns
Reviewed By: devilhorns
Subscribers: devilhorns, bu5hm4n, cedric, #reviewers, #committers
Tags: #efl_widgets
Differential Revision: https://phab.enlightenment.org/D9401
Summary:
this group name was wrongly changed by sed
ref d4526f44b8
Reviewers: segfaultxavi
Reviewed By: segfaultxavi
Subscribers: segfaultxavi, cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9419
Summary: For security concerns we removed the secret keys which could be used improperly by the wrong people.
Reviewers: woohyun, cedric, lauromoura
Reviewed By: lauromoura
Subscribers: #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9377
When user manually free the ecore evas,
it could delete evas internally,
then evas_invalidate would be triggered,
invalidate callback would try free evas again,
this causes double free evas.
TEST SCENARIO:
ee = ecore_evas_new(...);
...
ecore_evas_free(ee);
-> free evas
-> invalidated cb
-> free evas (**double free)
This is a regression bug by 5847886a3f
Summary:
if plug connection fails, a notification can't always be created because
there may be no object passed to this function to create a notify object on
@fix
Reviewers: devilhorns
Reviewed By: devilhorns
Subscribers: cedric, #reviewers, #committers
Tags: #efl_widgets
Differential Revision: https://phab.enlightenment.org/D9400
Summary:
engine internals have changed, so it's necessary to actually check whether
the glview api is available now to determine whether the glview is viable
@fix
Reviewers: devilhorns
Reviewed By: devilhorns
Subscribers: cedric, #reviewers, #committers
Tags: #efl_widgets
Differential Revision: https://phab.enlightenment.org/D9397
Summary:
If you call efl_gfx_filter_program_set in a mouse event callback,
it does not work. Because render_pre removes area uisng evas_render_update_del.
Reviewers: Hermet, jpeg, jsuya, cedric
Reviewed By: Hermet, cedric
Subscribers: cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9301
if scrolling is supposed to continue then just reuse the existing animator
callback and avoid emitting a scroll,start/stop event pair for every new
input event
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9340
Summary:
previously this used a hard linear scroll with a fixed animation time for
mouse wheel events, resulting in an unnatural scroll feel due to abrupt
animation termination
using the decelerate interpolator improves this, and we should be using the
config to determine animation speed
Depends on D9347
Reviewers: bu5hm4n, segfaultxavi
Reviewed By: segfaultxavi
Subscribers: cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9354
somehow it was never checked to see if a parent was being added as a subobject
of a parent's own child object (recursive hierarchy)
@fix
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D9333
We encountered a deadlock case in ecore_evas_image_object in ecore_evas_buffer
that only happens if the ecore_evas_buffer has nothing changed to render,
though it's triggered to rendering.
See this normal scenario that is working fine as our intention.
being ecore_evas_render()
...
-> ecore_evas_buffer_prepare()
-> evas_object_image_data_get()
-> increment lock by backend engine. (egl/tbm ...)
-> render()
-> render_post()
-> _ecore_evas_buffer_update_image()
-> evas_object_image_data_set()
->decrement lock by backend engine (egl/tbm ...)
...
end ecore_evas_render()
The problem is, if the ecore_evas_buffer canvas doesn't changed at all,
render post will be skipped, it could lose the chance to unlock the image data.
Now the host can't render anymore since it's image source lost the lock.
@fix
many of these functions go directly to evas internals with no eo checks,
and the existing "MAGIC_CHECK" macro has somehow become a useless null
check
type checking here is important in order to avoid crazy behavior when the
wrong object types are passed
@fix
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9364
in the case where ecore_main_loop_quit() was called before ecore_main_loop_begin(),
the latter call would exit immediately without ever iterating the main loop
@fix
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9360
this callback isn't called during the main loop, so attempting to
quit the loop has no effect and the test deadlocks
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9367
if a table is moved and no other changes are made to the table or its children,
e.g., if the table is scrolled, then there is no need to loop over the table's
items repeatedly in order to accurately calculate all the item geometries
and positions.
instead, simply apply an offset from the last table calc position to each child
item and handle the position changes more transparently
this yields roughly a 12% perf improvement to the 'efl.ui.scroller simple2' test
and brings rendering up to nearly 60fps
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9346
using alloca like this without any limits is dangerous, so switch to
malloc here in such cases
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D9344
Remove evas_table.
This expect to improve performance by removing internal function call related
evas_table.
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D8615
otherwise we do not clear the internals, and fail to get the count to 0.
Reviewed-by: YeongJong Lee <yj34.lee@samsung.com>
Differential Revision: https://phab.enlightenment.org/D8555
_pack_at is used to add new subobjects to the table. When a object is
already part of this table, then we should not add it again. Additional,
when there is already a gi structure, but the parent is something else,
then we should NOT just reuse this struct, otherwise we might use a
struct reference that we do not own. The struct could be owned by
another table widget.
The test must be adjusted, before we did not error on adding a widget
twice. Now we do (just like in box). Hence we should not do that in
test.
Reviewed-by: YeongJong Lee <yj34.lee@samsung.com>
Differential Revision: https://phab.enlightenment.org/D8554
any time an evas box is moved, it flags itself to do a recalc on all of
its contents in the next render. evas box also inherits from smart clipped
class, however, which means that it will also move all of its contents
immediately on every single move. this results in something like:
move(box) -> for content in box { move(content) } -> render ->
recalc(box) -> for content in box { calc(content); move(content); }
which is massively inefficient and results in box being completely unusable
once it has a large number of contents
by skipping immediate move() calls for all the box contents, we can bring this
performance back to usable levels
@fix
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9336
Summary:
Evas_Coord is a regular int, so this will overflow easily for large
scrollers and return NaN for scroll duration and break the scroll
Reviewers: segfaultxavi
Reviewed By: segfaultxavi
Subscribers: segfaultxavi, cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9355
This is a work around compiler/linker limit on some system as reported
by Romain Naour.
Reviewed-by: Mike Blumenkrantz <michael.blumenkrantz@gmail.com>
Differential Revision: https://phab.enlightenment.org/D9348
if a box is moved and no other changes are made to the box or its children,
e.g., if the box is scrolled, then there is no need to loop over the box's
items repeatedly in order to accurately calculate all the item geometries
and positions.
instead, simply apply an offset from the last box calc position to each child
item and handle the position changes more transparently
this yields roughly a 12% perf improvement to the 'efl.ui.scroller simple' test
and brings rendering up to nearly 60fps
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9342
using alloca like this without any limits is dangerous, so switch to
malloc here in such cases
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9341
Remove legacy stuff from Efl.Ui.Box.
This expect to improve performance by removing internal function call related
evas_box.
Reviewed-by: Xavi Artigas <xavierartigas@yahoo.es>
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D8417
in case we are having a scroller with a lot of elements on it, we are
spending a lot of time in stuff like recalculating clips, even if they
will never be used again.
With this freeze here, we are saving 9s shutdown time in item_container.
Which brings the overall closing time from 10s down to 1s, which is a
win IMO.
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9325
there is a very basic problem in eo events. We are having one central
array of event subscription, if for example a widget is now listening to
changes in its parent, then we are 100% asking for trouble.
As an example:
- A scroller with 100 buttons in it.
- Every button will have a subscription to the FOCUS_MANAGER_CHANGED
event
If you now scroll, the position is updated in the scroller, therefore
the position in scroller is updated. This has the result that the whole
list of 100 event subscriptions is walked, which is obviously bad,
however, this solution here is way easier than fixing eo (i am not even
sure there is a nice solution to it).
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9324
when a logic parent does not have any widgets left, the parent needs to
be reevaluated. However, this only has to happen when there is a change
in state (eg. from 0 -> N or from N -> 0). Every other call can be
safed. This commit introduces this checking, and safes up performance.
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9323
Summary:
text should not be focusable by default. Only focusable if it is
editable. This fixes mysterical focus disappearing in tests using
efl.ui.text.
Depends on D9327
Reviewers: segfaultxavi
Subscribers: cedric, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9328
this value is supposed to always match the current widget state and signal
emission breaks if the states are not consistent
@fix
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D9319
checking part type and part geometry is pretty common, so these base
implementations can fill in gaps in existing functionality
Reviewed-by: Marcel Hollerbach <mail@marcel-hollerbach.de>
Differential Revision: https://phab.enlightenment.org/D9317
This allow evas test to work with an Ecore_Evas directly. It prevent leaking of memory
in the case of half destroying Ecore_Evas.
Reviewed-by: Hermet Park <hermetpark@gmail.com>
Differential Revision: https://phab.enlightenment.org/D9104
Summary:
Decrement grab count of mouse event when mouse up happen during a freeze state.
Currently as grab count is not reduced in this scenario, further touches gets ignored.
Signed-off-by: Godly T.Alias <godlytalias@yahoo.co.in>
Test Plan: (on scenario where proxy exists) mouse down -> freeze events -> mouse up -> unfreeze events
Reviewers: cedric, raster, zmike, Hermet
Reviewed By: zmike
Subscribers: zmike, #reviewers, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D9214
given the following scenario:
* object is added to smartobj
* smartobj is otherwise unchanged
* render occurs
object will never render due to smartobj's render cache not including object
by clearing this cache when reinserting smartobj into the render tree for a
followup render, the object chain will be correctly processed on the next pass
fix T7990
Reviewed-by: Cedric BAIL <cedric.bail@free.fr>
Differential Revision: https://phab.enlightenment.org/D9287