not backporting because not sure if correct, but definitely a bug
also there's a similar issue where objects (I'm looking at you, elm_list) don't size until the first render, which causes the initial genlist calc sizes to be 100% wrong every time
time to test/fix accessibility.
Revert "elm access - if we disable apis then disable the tests too - unbreak build"
This reverts commit d8b3cdb28f5334f9429959109af4b960bf414097.
- refer commit 64fd5a825b24591aedb9cac695d03f5b4d867442 for the detail
explanation.
- also move evas_object_resize(win, xx) alongside evas_object_show(win).
- refer commit 64fd5a825b24591aedb9cac695d03f5b4d867442 for the detail
explanation.
- also move evas_object_resize(win, xx) alongside evas_object_show(win).
Elm Flip has some confusing API (and documentation as well) regarding
flip directions and hitsize. Basically, it works just as expected
when everything is set to 1 (all hitsizes are 100% and all directions
are enabled), like in the elm example file :)
BUT, it's not actually possible to restrict the hitsize for one
direction.
Example: You want to flip to left only.
Set direction enabled LEFT.
Then, you must set the hitsize for... RIGHT because that's where the
finger interaction will start from (right to left!). BUT since
RIGHT interaction is not enabled, the histize parameter on RIGHT
will have NO effect on the flip.
Here's what this commit fixes:
- Clear up a bit of the confusion between direction and hit area,
through documentation
- Separate hit area position and direction (bugfix)
- Allow for negative hitsize values to DISABLE a specific area
(define previously undefined behaviour)
So, it will be possible for example:
- Enable left/right interactions
- Only on the top 10% of the widget
NOTE: This breaks behaviour! API/ABI are not quite broken but
undefined behaviour just got defined and fixed.
- Removed unused variable.
- Followed the structure naming convention.
- Fixed formatting.
- Removed unnecessary empty lines and spaces.
- Removed unnecessary casting.
- Removed typos.
- Use ECORE_CALLBACK_RENEW macro for animator callback.
- Just call elm_exit() to exit the main loop.
- Removed unnecessary evas_object_data usage.
- Made a button text more descriptive and correct.
Scenario:
1. Create elm_genlist.
2. Add some items to it.
3. Do not resize the genlist and leave it with width == 0.
PROPOSAL:
1. When the element is queued for recalculation check current width of
widget. If it is equal to 0 do not add idle enterer.
2. In smart callback on resize check if queue of items that need
recalculation is not empty and new width is not 0. In this situation
add idle enterer.
stable release - cherry-pick me!
During animation, if the naviframe wasn't inside another object that would clip
it, you could notice the shadow going over some object around.
Application can delete button itsefl(or its parent) inside the smart callback function.
So the signal_emitting with button object should be done in advance.
Summary: Minor defect in example program fixed.
Test Plan: gcc -g radio_example_01.c -o radio_example_01 `pkg-config --cflags --libs elementary`
Reviewers: seoz, singh.amitesh
Differential Revision: https://phab.enlightenment.org/D363
When the offset is not bigger than 0 && abs(offset) is not smaller than dx(or dy), round up routine could have mis-calculation bug.
It should substract one value more to get a proper round up result.
For example, when x = 1440, offset = -2160, dx = 360, x should become 720 but round up routine made it 1440.
Because it was rounded up from 1080 to 1440. It should be rounded down from 1079 to 720.
Special thanks to Hosang Kim for the help.
Signed-off-by: Daniel Juyung Seo <juyung.seo@samsung.com>
Signed-off-by: Hosang Kim <hosang12.kim@samsung.com>
Summary:
Hello guys,
Below You can find a proof o concept for ATSPI2 support for elementary applications. Currently application can only be listed on bus by clients such as 'accersiser'. I am strongly looking for comments to this design.
Here are my comment regarding following code:
* This pach requires https://phab.enlightenment.org/D327 patch to compile
* I have decided to include whole atspi-constants header, because in fact i will need one-to-one copy of all its enums and defines. Qt has something like 3rdparty folder from which they include this header and don't require atspi dev packages at compilation time. Maybe efl needs the same solution or should just require next package in compilation? I agree that putting atspi-constant.h in main source tree is bad idea, but where? src/utils ??
Reviewers: cedric, raster, kimcinoo
Reviewed By: raster
CC: pkaczmarek-samsung, seoz, jaehwan, kimcinoo
Differential Revision: https://phab.enlightenment.org/D328
Summary: Added support for popup align set
Test Plan: elm_popup_align_set, elm_popup_align_get, elementary_test -to popup
Reviewers: seoz, singh.amitesh
Differential Revision: https://phab.enlightenment.org/D321
Scrollto positions are divided 4 type
NONE : do not scrollto.
TOP : scrollto position is always top viewport.
MIDDLE : scrollto position is always middle of viewport.
IN : scrollto position to fully visible in viewport.
Summary: Changes in the _item_realize to decorate the item T576 test_genlist10
Reviewers: seoz, singh.amitesh
Differential Revision: https://phab.enlightenment.org/D358
If we're walking an item and it's deleted, the memory won't go away
immediately in _item_del_pre_hook() as would in _item_del(), then it's
not enough to remove ourselves from the reverse-relative list of the
item we were relative to, we also need to clean our own relative
pointer so we won't touch it later when the item is not being walked
anymore and _item_del() is called.
I was getting this annoying error with espionage application, opening
an interface of an object and then closing the window or selecting
another bus name (whatever would call elm_genlist_clear()).
During the clear process genlist will flag the next item as "walking"
so it's not gone when the current item dies (would happen if next item
is a subitem). The item would run _item_del_pre_hook() but not
_item_del(), but on the next loop iteration the next item would be the
current, then not walking anymore and during _item_del() it would
access it->item->rel which would point to the now-dead item.
When we scroll the genlist with decorate all mode, newerly realized items looked like a normal mode.
This was caused by a new theme. Genlist item style should handle all the necessary parts when decorate mode is enabled/diabled.
This fixes T576.
Now it looks like a normal efl app.
- Declare variables at the start of the function.
- Resize and show the window right after its creation.
- Set widgets' parents properly.
This fixes " _smart_need_recalculate_set() Object 0x803db8e8 is not stable during recalc loop" issue which was reported by kuuko.
Special thanks to kuuko.
before this, it had the insane logic so the prescale_set() never work.
Now, it works well and the prescale won't be set in default. (before, the default value is 64. why?)
When the content is removed(or unset) the mapbuf didn't clean up the some stuff such as removing event callback for the content.
So the unset content would be tracked still by mapbuf dangling callbacks.
Thank you eo pointer indirection for detecting this.
The issue is a mix of design choices (will discuss on the ML) and
a bug in hoversel which was probably caused because of the confusing design
choices.
The problem is that while widget_item->view is set by whoever created
the item, it's automatically deleted by elm_widget_item_del. This means
that objects that decided to delete their views, but were too sloppy to
update the widget_item->view pointer to NULL, would trigger an attempt
to deleted an already deleted object. This does not happen in 1.7
because widget_item->view was introduced in 1.8.
The automatic view deletion is problematic, as smart subobjects are
already deleted when the parent subobject (the widget) is deleted. So
clearing the views is redundant. However if the widget item is manually
deleted, the view should be deleted. This causes this mixture of cases
that are problematic at best.
Indication error:
Error:
ERR<5394>:eo lib/eo/eo_ptr_indirection.x:275 eo_obj_pointer_get()
obj_id 0x8000002900000149 is not pointing to a valid object. Maybe
it has already been freed.
This fixes T500.
Carsten Haitzler
discomfitor
Daniel Juyung Seo
Daniel Juyung Seo
Ryuan Choi
ChunEon Park
Cedric BAIL
Jean-Philippe Andre
Mike Blumenkrantz
Daniel Zaoui
Cedric Bail
Boris Faure
Stefan Schmidt
Viacheslav Lvov
Yakov Goldberg
Stefan Schmidt
WooHyun Jung
Ryuan Choi
Chinmaya Panigrahi
Lukasz Stanislawski
Anil Kumar Nahak
abhi
Sanghyeon Lee
Jerome Pinot
ami
nirajkr
Gustavo Sverzut Barbieri
Tom Hacohen
Seong-ho Cho
Rafael Antognolli
maxerba
Jihoon Kim
Sanjeev BA