platform/upstream/efl.git
7 years agojs: Change test eos to ptr()
Lauro Moura [Fri, 10 Feb 2017 14:55:43 +0000 (11:55 -0300)]
js: Change test eos to ptr()

7 years agodocgen: add a way to get numerical id of a function
Daniel Kolesa [Wed, 22 Feb 2017 17:29:07 +0000 (18:29 +0100)]
docgen: add a way to get numerical id of a function

7 years agodocgen: more useful function listings in class pages
Daniel Kolesa [Wed, 22 Feb 2017 14:54:36 +0000 (15:54 +0100)]
docgen: more useful function listings in class pages

Now there are C signatures visible.

7 years agogenlist: Simplify code (remove one argument)
Jean-Philippe Andre [Wed, 22 Feb 2017 06:34:34 +0000 (15:34 +0900)]
genlist: Simplify code (remove one argument)

Parameter "qadd" is always true when calling _item_process_post(),
so simplify the code by removing it.

7 years agogenlist: Fix more tree issues (expanded state)
Jean-Philippe Andre [Wed, 22 Feb 2017 04:53:03 +0000 (13:53 +0900)]
genlist: Fix more tree issues (expanded state)

This fixes a lot of cases where a genlist node (of any type)
with children may have appeared in the invalid expanded or
contracted state.

Before this patch, the test case "Genlist tree, Relative insert"
looked like below (all items are programmatically added):

> A
  > 1
  > 2
> B
  > 3
  > 4

The problem above is that A and B have visible children but
still believe they are in contracted state. This patch ensures
that A and B will be marked as expanded, but will do so without
firing an "expanded" signal and definitely without the
"expand,request" signal.

After this patch, the test case will look like this:

v A
  > 1
  > 2
v B
  > 3
  > 4

Which is more correct. Note that this test case does not handle
any expand/contract signal.

NOTE: This is a behaviour break!

7 years agogenlist: Fix insert after with a tree
Jean-Philippe Andre [Tue, 21 Feb 2017 13:04:10 +0000 (22:04 +0900)]
genlist: Fix insert after with a tree

This fixes a case where inserting item "C" after item "B" in this
tree would go wrong:

A
B
- 1
- 2

Before this patch, 1 and 2 lose their parent:

A
B
C
- 1
- 2

After this patch, 1 and 2 retain their parent:

A
B
- 1
- 2
C

Insert before worked by luck, no need to fix it.
Note that this patch may require the next one to actually
work (ensuring expanded state flag).

NOTE: This is a behaviour break!

7 years agogenlist: Fix sorted_insert with tree
Jean-Philippe Andre [Tue, 21 Feb 2017 12:48:11 +0000 (21:48 +0900)]
genlist: Fix sorted_insert with tree

This fixes the test case "Genlist Tree, Insert Sorted".
This is a pretty ugly patch... but the genlist code is already
pretty ugly, as it keeps a flat inlist of items (sd->items)
as well as a tree structure in parallel.

Before this patch, the following configuration led to issues:

 1
 3
 - A
 - B

Adding item "2" led to a crash. Adding item 4 led to this:

 1
 3
 4
 - A
 - B

Items A and B lost their parent "3". Subsequent sorted inserts
would lead to insane bahaviour, where for instance "8" would
appear before "3".

This patch fixes all sorted inserts, at the cost of performance
(an optimized code path is avoided). Subsequent patches will
increase the robustness of the tree structure.

NOTE: This is a behaviour break!

Fixes T4850

7 years agoedje_edit: fix scripts compilation
Andrii Kroitor [Tue, 21 Feb 2017 17:33:51 +0000 (19:33 +0200)]
edje_edit: fix scripts compilation

@fix

7 years agodocgen: inherit overridden method/property brief descriptions on class pages
Daniel Kolesa [Tue, 21 Feb 2017 15:48:16 +0000 (16:48 +0100)]
docgen: inherit overridden method/property brief descriptions on class pages

7 years agoslider: theme - fix warnings
Amitesh Singh [Tue, 21 Feb 2017 15:23:51 +0000 (20:53 +0530)]
slider: theme - fix warnings

Fixes warnings related to  "disabled_visible" states
not present in "limits" part.

fixes T5123

7 years agoelementary gengrid: fix for working item reorder mode correctly
Minkyu Kang [Tue, 21 Feb 2017 09:32:30 +0000 (15:02 +0530)]
elementary gengrid: fix for working item reorder mode correctly

Summary:
Change the item indexing to start 1 after reordering animation.
Change the logic of edge checking to get the row or col correctly.

Signed-off-by: Minkyu Kang <mk7.kang@samsung.com>
Test Plan:
elementary_test -to gengrid2
append 6 items
enable the reorder mode
check reordering is working properly (4 to 1)

@fix

Reviewers: singh.amitesh, cedric

Subscribers: jehun.lim, jpeg

Differential Revision: https://phab.enlightenment.org/D4676

7 years agodayselector: Fix DaySelector widget
Jean-Philippe Andre [Tue, 21 Feb 2017 09:14:36 +0000 (18:14 +0900)]
dayselector: Fix DaySelector widget

An invalid strcmp() led to always reject the swallowed
widgets, which were Elm.Check instead of Elm_Check.
Use efl_isa() instead.

Fixes T5034

7 years agoedje: Fix a major issue with recalc loop
Jean-Philippe Andre [Tue, 21 Feb 2017 08:49:44 +0000 (17:49 +0900)]
edje: Fix a major issue with recalc loop

With the new Efl.Ui.Clock widget, which implements Datetime,
most parts are relative to each other, and trigger a near-inifinite
recalc of all the parts. Indeed edje was recalculating a part even
if it had already calculated it (flag is set).

This seemingly simple patch fixes Datetime and probably a lot
of other edje widgets.

Fixes T4909

@fix

7 years agoevas: Optimize out some nop calls to evas text
Jean-Philippe Andre [Tue, 21 Feb 2017 08:48:08 +0000 (17:48 +0900)]
evas: Optimize out some nop calls to evas text

Some property set calls to Evas Text are NOPs if the property was
already set to the same value. Avoid doing any work then.

This adds a missing intercept call (size_set as it's overriden
by Evas.Text).

7 years agoelm: Make clock widget more resilient to invalid config
Jean-Philippe Andre [Tue, 21 Feb 2017 07:14:20 +0000 (16:14 +0900)]
elm: Make clock widget more resilient to invalid config

I was in a situation where my elm_config had the proper version
but did not contain a proper module name for the clock widget.
This resulted in not being able to create any field object in
the clock widget... in other words it made for a very useless
clock :)

Ideally the config should be updated at this point. I'm not
sure how to proceed exactly, though :(

7 years agoevas: Fix module memleak during evas_shutdown
Jean-Philippe Andre [Tue, 21 Feb 2017 05:46:02 +0000 (14:46 +0900)]
evas: Fix module memleak during evas_shutdown

This fixes a minor memory leak during shutdown.

Note: This does NOT perform the dlclose (and still leaks the
Eina_Module descriptor). Calling dlclose leads to a whole lot
of other issues, so we avoid it.

7 years agoevas: Fix make check (evas_suite compilation)
Jean-Philippe Andre [Tue, 21 Feb 2017 06:11:32 +0000 (15:11 +0900)]
evas: Fix make check (evas_suite compilation)

This fixes issues where Evas_Object_Intercept_Cb_Type was not
defined, as this is defined in Evas_Legacy.h (unfortunately...
as it's used by elementary), but the private headers defining
EFL_CANVAS_OBJECT_PROTECTED were included after Evas_Legacy.h.

7 years agoevas: Fix compilation warning (unused variable)
Jean-Philippe Andre [Tue, 21 Feb 2017 06:11:42 +0000 (15:11 +0900)]
evas: Fix compilation warning (unused variable)

7 years agoevas: Avoid data_scope_get in internal function
Jean-Philippe Andre [Tue, 21 Feb 2017 03:24:56 +0000 (12:24 +0900)]
evas: Avoid data_scope_get in internal function

7 years agoevas: Inline part of clip_dirty
Jean-Philippe Andre [Mon, 20 Feb 2017 14:14:38 +0000 (23:14 +0900)]
evas: Inline part of clip_dirty

clip_dirty is called A LOT. Unfortunately this patch does not
result in very measurable improvements.

**********************
Note about this merge:

After this series of patches, the CPU usage for a certain test
case has significantly gone down:

Based on c0e6a8d698c17fc16f9b67fc9 (upstream before git push):
NS since frame 2 = 28910806786 , 2937 frames = 9843652 / frame

After this patch:
NS since frame 2 = 19218592951 , 2928 frames = 6563727 / frame

1.18:
NS since 2 = 13105584220 , 2961 frames = 4426066 / frame

As we can see, 1.18 remains *much* better than 1.19. I'm still
struggling trying to figure out why (clip_recalc is called more,
but the call tree is hard to decypher).

The test case is:
EINA_FREEQ_BYPASS=1 ELM_TEST_AUTOBOUNCE=100 \
  elementary_test -to "Scroller 2"

EFL was compiled with GCC 6.3.1 with -O3 -g

@optimization

7 years agoevas: Inline part of clip_recalc
Jean-Philippe Andre [Mon, 20 Feb 2017 11:24:21 +0000 (20:24 +0900)]
evas: Inline part of clip_recalc

This function was moved out of inline (see d7c6fca6c00a0bfb05) but
unfortunately the early checks at its beginning are likely
to result in an early return. Inline this part so we get back a
better performance. Inlining the whole function does not improve
the performance, as GCC simply gives up with inlining.

Note: Between 1.18 and master the number of calls to clip_recalc
has simply blown out. It is thus crucial to find out where those
calls come from but also micro-optimize the function itself. This
patch does the latter only.

@optimize

7 years agoevas: Optimize out some calls to data_scope_get
Jean-Philippe Andre [Mon, 20 Feb 2017 10:34:58 +0000 (19:34 +0900)]
evas: Optimize out some calls to data_scope_get

Within evas when we know an object is marked as is_smart, then
we know it's a smart object. :)

@optimize

7 years agoevas: Avoid more calls to efl_isa in clip_set
Jean-Philippe Andre [Mon, 20 Feb 2017 09:26:40 +0000 (18:26 +0900)]
evas: Avoid more calls to efl_isa in clip_set

No need to verify that the clip is an evas object unless we're
actually going to change said clipper. This has a minor effect,
removing ~10% of all calls to efl_isa (0.1% of everything).

After all those efl_isa patches the number of calls has dropped
from 1.9M (~3.05%) to ~570k (0.76%). More can still be done,
surely. Now efl_isa accounts for ~0.05% of my CPU usage according
to perf top (down from ~0.7%).

If edje uses EO APIs instead of legacy image APIs, then many of
the remaining calls will vanish.

@optimization

7 years agoevas: Assume object is valid and avoid call to efl_isa
Jean-Philippe Andre [Mon, 20 Feb 2017 09:07:00 +0000 (18:07 +0900)]
evas: Assume object is valid and avoid call to efl_isa

In an internal function we should be able to safely assume an
object is indeed a smart object.

@optimization

7 years agoevas: Optimize out more calls to efl_isa
Jean-Philippe Andre [Mon, 20 Feb 2017 08:59:50 +0000 (17:59 +0900)]
evas: Optimize out more calls to efl_isa

In relative smart move, we only need to check once that the object
is indeed a valid smart object. This drops the call count from ~1.02M
to ~785k.

@optimization

7 years agoevas: Avoid call to EAPI inside internal code
Jean-Philippe Andre [Mon, 20 Feb 2017 08:25:29 +0000 (17:25 +0900)]
evas: Avoid call to EAPI inside internal code

This avoids calling efl_isa and locking the async mutex.

In callgrind analysis, this reduces the count of calls to
efl_isa from 1.96M to 1.02M.

@optimization

7 years agoevas: Avoid useless calls to eo_data_scope_get
Jean-Philippe Andre [Fri, 17 Feb 2017 09:46:09 +0000 (18:46 +0900)]
evas: Avoid useless calls to eo_data_scope_get

Use an internal variant of the function for evas when we already
have the protected data pointer.

7 years agoeo: Make _eo_obj_pointer_done an inline function
Jean-Philippe Andre [Fri, 17 Feb 2017 09:37:03 +0000 (18:37 +0900)]
eo: Make _eo_obj_pointer_done an inline function

@optimization

7 years agoevas: Remove useless call to evas_object_async_block
Jean-Philippe Andre [Fri, 17 Feb 2017 09:15:57 +0000 (18:15 +0900)]
evas: Remove useless call to evas_object_async_block

It's already called from the intercept cb.

7 years agoevas: Inline checks for post-event cb call
Jean-Philippe Andre [Fri, 17 Feb 2017 09:06:10 +0000 (18:06 +0900)]
evas: Inline checks for post-event cb call

This avoid excessive calls to a real function.

@optimization

7 years agodocs: Fix typos and some wrong expressions in Eina API reference doxygen.
Myoungwoon Roy, Kim [Mon, 20 Feb 2017 06:18:21 +0000 (15:18 +0900)]
docs: Fix typos and some wrong expressions in Eina API reference doxygen.

Summary: I had fixed some typos and some wrong expressions, such as capital letters, singular, and orders of groups in Eina API reference doxygen.

Test Plan: Doxygen Revision

Reviewers: stefan, cedric, raster, Jaehyun_Cho, jpeg

Reviewed By: jpeg

Subscribers: conr2d

Differential Revision: https://phab.enlightenment.org/D4674

7 years agoexamples: Use printf instead of fprintf(stdout, ...)
Bryce Harrington [Mon, 20 Feb 2017 05:33:04 +0000 (14:33 +0900)]
examples: Use printf instead of fprintf(stdout, ...)

Summary:
Also fix warning about lack of a format string:

    warning: format not a string literal and no format arguments [-Wformat-security]
         fprintf(stdout, commands);

Signed-off-by: Bryce Harrington <bryce@osg.samsung.com>
Reviewers: cedric, jpeg

Reviewed By: jpeg

Subscribers: cedric, jpeg

Differential Revision: https://phab.enlightenment.org/D4673

7 years agorelease: Update NEWS and bump version for 1.19.0-beta2 release
Stefan Schmidt [Mon, 20 Feb 2017 21:45:31 +0000 (22:45 +0100)]
release: Update NEWS and bump version for 1.19.0-beta2 release

7 years agoeina utils: fix build without getpwent.
Mykyta Biliavskyi [Mon, 20 Feb 2017 09:29:56 +0000 (11:29 +0200)]
eina utils: fix build without getpwent.

If HAVE_GETPWENT isn't defined - the closing brace was missed.
Also prevent situation when strdup() tried to duplicate NULL
pointer, because that could cause segfault.

@fix

7 years agointerface_scrollable: Improve gravity_set api to support pan changing
Wonki Kim [Mon, 20 Feb 2017 02:31:12 +0000 (11:31 +0900)]
interface_scrollable: Improve gravity_set api to support pan changing

Summary:
When you set gravity 1 on scroller, scroller sticks to the bottom
even content is changed.
however, scroller don't work like above, if size of pan is changed.

this commit uses pan_pos_max rather than w/h of content_info
because pan_pos_max is related with both content_size and pan size.

gravity_set will work properly even if both size of content and pan are
changed simultaneously.

Test Plan:
1. Select 'scroll3' in the elementary_test
2. Append enough items so that scroll bar appears (about 30 items)
3. Go to the bottom and Set gravity 1.0
4. Check that scroller sticks to the bottom once you append another item
   (it works)
5. Check that scroller sticks to to bottom once you resize window(pan)
   (it doesn't work without this patch)

Reviewers: eagleeye, jpeg, cedric, woohyun, z-wony, herdsman

Differential Revision: https://phab.enlightenment.org/D4665

7 years agoevas: do not error for smart class objects
Marcel Hollerbach [Mon, 20 Feb 2017 02:11:34 +0000 (11:11 +0900)]
evas: do not error for smart class objects

Summary: this should fix some spamming in e

Reviewers: jpeg

Reviewed By: jpeg

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D4675

Signed-off-by: Jean-Philippe Andre <jp.andre@samsung.com>
7 years agoecore_evas_wayland_common: fix bug that window is not rotated when app request rotate.
Jiyoun Park [Mon, 20 Feb 2017 01:28:27 +0000 (10:28 +0900)]
ecore_evas_wayland_common: fix bug that window is not rotated when app request rotate.

In landscape mode, width & height of evas are different with ecore's.
so diff is checked differently according to the  portrait & landscape mode.

7 years agoelm_test: Add autobounce test for "Scroller 2"
Jean-Philippe Andre [Fri, 17 Feb 2017 07:59:41 +0000 (16:59 +0900)]
elm_test: Add autobounce test for "Scroller 2"

7 years agoCanvas text: fix clearing existing style
Daniel Hirt [Sun, 19 Feb 2017 16:10:24 +0000 (18:10 +0200)]
Canvas text: fix clearing existing style

Fixes T5187.
This fixes a leak in evas_textblock_style_user_pop, that made use of the
new key-based efl_canvas_text_style_set api.

@fix

7 years agoelm_code: Another Coverity fix, better this time.
Andy Williams [Sun, 19 Feb 2017 14:15:29 +0000 (14:15 +0000)]
elm_code: Another Coverity fix, better this time.

@fix CID1371322

7 years agoreadme - systemd is default on now... --disable if you dont want it
Carsten Haitzler (Rasterman) [Sun, 19 Feb 2017 12:11:57 +0000 (21:11 +0900)]
readme - systemd is default on now... --disable if you dont want it

7 years agoelm code - fix warning about sign comparisons
Carsten Haitzler (Rasterman) [Sun, 19 Feb 2017 06:48:41 +0000 (15:48 +0900)]
elm code - fix warning about sign comparisons

7 years agoelm_code: Fix Coverity issues.
Andy Williams [Sat, 18 Feb 2017 21:41:53 +0000 (21:41 +0000)]
elm_code: Fix Coverity issues.

@fix CID1356612 CID1368335 CID1368489 CID1371123 CID1371124 CID1371126 CID1371128

7 years agoelm config base profiles - fixselect allow to be orrectly iniitted
Carsten Haitzler (Rasterman) [Fri, 17 Feb 2017 08:34:44 +0000 (17:34 +0900)]
elm config base profiles - fixselect allow to be orrectly iniitted

fixes previous commit...

7 years agofix entry selection by having proper cfg update and def fields
Carsten Haitzler (Rasterman) [Fri, 17 Feb 2017 08:21:20 +0000 (17:21 +0900)]
fix entry selection by having proper cfg update and def fields

this should fix T5183

7 years agoeo: ensure the generation is correctly clamped
Marcel Hollerbach [Thu, 16 Feb 2017 21:37:16 +0000 (13:37 -0800)]
eo: ensure the generation is correctly clamped

Summary:
when a few recursive event emissions are happening, and in some deep
recursive level a subscription to the same object is happening, the
subscription would just be executed when the complete recursion is done.
that is wrong. The subscription needs to be executed when the event is
called after the subscription is added, undepended from any recursive
level. That fixes that and adds a regression test for it.

This was discovered in e, since e gives a lot of error messages about a eo object
that is already freed. It turned out this object is returned from evas, and exactly
the above happened to the EFL_EVENT_DEL subscription of that object.

Test Plan: make check

Reviewers: tasn, cedric, stefan_schmidt

Subscribers: stefan_schmidt, netstar, zmike, raster, jpeg

Differential Revision: https://phab.enlightenment.org/D4656

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
7 years agodoxygen: fix typos and some wrong expressions in Edje and Eet API reference documenta...
Myoungwoon Roy, Kim [Thu, 16 Feb 2017 18:16:49 +0000 (10:16 -0800)]
doxygen: fix typos and some wrong expressions in Edje and Eet API reference documentation.

Summary: I had fixed some typos and some wrong expressions, such as capital letters, singular, and orders of groups in Edje and Eet API reference doxygen.

Test Plan: Doxygen Revision

Reviewers: stefan, cedric, raster, Jaehyun_Cho, jpeg

Subscribers: conr2d

Differential Revision: https://phab.enlightenment.org/D4666

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
7 years agodocgen: add -p to print what exactly is being generated currently
Daniel Kolesa [Thu, 16 Feb 2017 16:29:23 +0000 (17:29 +0100)]
docgen: add -p to print what exactly is being generated currently

7 years agodocgen: make sure all class retrievals are cached
Daniel Kolesa [Thu, 16 Feb 2017 16:20:00 +0000 (17:20 +0100)]
docgen: make sure all class retrievals are cached

7 years agodocgen: cache function instantiations
Daniel Kolesa [Thu, 16 Feb 2017 16:12:43 +0000 (17:12 +0100)]
docgen: cache function instantiations

7 years agodocgen: cache class retrievals for performance
Daniel Kolesa [Thu, 16 Feb 2017 15:06:56 +0000 (16:06 +0100)]
docgen: cache class retrievals for performance

7 years agogengrid: Remove duplicated smart callbacks
Jean-Philippe Andre [Thu, 16 Feb 2017 13:05:58 +0000 (22:05 +0900)]
gengrid: Remove duplicated smart callbacks

WRN... evas_smart.c:219 evas_smart_cb_descriptions_fix() duplicated
  smart callback description with name 'item,focused' and type ''

7 years agowidget: Reduce WRN to DBG for on_focus_region message
Jean-Philippe Andre [Thu, 16 Feb 2017 13:00:19 +0000 (22:00 +0900)]
widget: Reduce WRN to DBG for on_focus_region message

Many widgets don't implement this function, and this case is
perfectly well handled by the caller. No need to WRN here.

7 years agoevas: Silence useless WRN message on canvas del
Jean-Philippe Andre [Thu, 16 Feb 2017 12:56:11 +0000 (21:56 +0900)]
evas: Silence useless WRN message on canvas del

7 years agoecore_evas: Fix mouse event in inlined windows
Jean-Philippe Andre [Thu, 16 Feb 2017 12:51:47 +0000 (21:51 +0900)]
ecore_evas: Fix mouse event in inlined windows

This fixes the following ERR message:
ERR<10589>:eina_safety /home/jpeg/e/core/efl/src/lib/ecore_evas/ecore_evas.c:3149
  _ecore_evas_mouse_move_process_internal() safety check failed: cursor == NULL

7 years agowin: Avoid magic check failed in ecore with inline win
Jean-Philippe Andre [Thu, 16 Feb 2017 12:45:12 +0000 (21:45 +0900)]
win: Avoid magic check failed in ecore with inline win

Fixes T4663

7 years agoelm_test: Fix invalid API use (EO)
Jean-Philippe Andre [Thu, 16 Feb 2017 12:32:39 +0000 (21:32 +0900)]
elm_test: Fix invalid API use (EO)

7 years agowin: Avoid malloc in icon_object_set
Jean-Philippe Andre [Thu, 16 Feb 2017 11:40:22 +0000 (20:40 +0900)]
win: Avoid malloc in icon_object_set

Also support both Evas.Image and EO Efl.Canvas.Image classes.
Add a test case in elm_test (under "Icon").

I'm not so happy about this patch... it shows that the API
barrier between legacy and EO implemented for images may not
be such a great idea after all :(

7 years agoevas: Flush post-event callbacks only when needed
Jean-Philippe Andre [Thu, 16 Feb 2017 11:17:12 +0000 (20:17 +0900)]
evas: Flush post-event callbacks only when needed

This avoids flushing the post-event cb queue when processing
recursive events. See 54e5841b2f6386765dd5e2831aa3481b952e5.

7 years agowin: Fix window icons with object set
Jean-Philippe Andre [Thu, 16 Feb 2017 11:05:04 +0000 (20:05 +0900)]
win: Fix window icons with object set

Fixes T4981

7 years agoevas: Re-fix post-event cb by supporting recursion
Jean-Philippe Andre [Thu, 16 Feb 2017 10:34:08 +0000 (19:34 +0900)]
evas: Re-fix post-event cb by supporting recursion

The previous patch (b184874fa5a0c3177c) was preventing
post-event callbacks from triggering any form of input event,
including side-effects due to mouse,in. In fact by tracking
which exact events we want to post-process we can support
proper recursion. This fixes crashes in Bryce.

I'm not changing the documentation as this is still a dubious
code design.

Fixes T3144
Fixes T5157

7 years agoevas: Strengthen post-event callbacks
Jean-Philippe Andre [Thu, 16 Feb 2017 07:47:57 +0000 (16:47 +0900)]
evas: Strengthen post-event callbacks

See T3144 that I marked as Wontfix.

Bryce in E manually feeds events from a post-event callback
resulting in Evas going insane and leading to frequent crashes.
The ideal solution (for E) would be to ensure that everything works
smoothly, the input event data is valid up until the post-event cb
is called, etc... Unfortunately, with recursive events the exact
order of operations may be messed up: the post-event

I don't want to add yet more complexity to Evas events here (it's
already spaghetti all over the place) so I'm simply blocking any
new event feed when running the post-event callback list.

It's not possible to just freeze the events (I tried, it failed).

**********************
Some more explanation:

post-event callbacks are used to implement reverse-order logic
where the on-hold flag of an input event may be set by an event
listener that does not come first.

Here's a situation to illustrate: scroller A inside scroller B.

As events are propagated from children to parents (assuming the
propagate flag is set), we'd assume the events to go first to A
and then to B, which means a mouse wheel event would make the
inner-most scroller (A) scroll, and the outer-most scroller (B)
wouldn't budge.

But as things are designed, A and B are not simple evas objects,
and the actual event-catching object is a top-most transparent
rectangle (top-most in Z stack order). Since A is inside B, B's
rectangle BR is over A's rectangle AR, thus catches the wheel
event first. But in terms of UX we still want A to scroll, not B.

The solution then is to reverse the event processing order and
post-event callbacks are the way to do that. This comes with the
consequence that the event_info needs to remain valid until the
post-event is called, and stay the same (so that the on-hold flag
set by A can be read by B).

Recursive events (by explicit feed or modifying the canvas so
that mouse,in or mouse,out are triggered) mess with this logic,
and trigger the post-events too early (event is not fully
processed) or too late (event_info is not valid anymore... and
crash!).

Thanks @raster for explaining the goal of post-event callbacks!

7 years agoevas: Ensure post-event cb push is called from an event cb
Jean-Philippe Andre [Thu, 16 Feb 2017 07:26:42 +0000 (16:26 +0900)]
evas: Ensure post-event cb push is called from an event cb

This rejects calls to evas_post_event_callback_push() that don't
originate from inside an input event callback.

7 years agoevas/doc: Add warning about post_event_callback_push
Jean-Philippe Andre [Thu, 16 Feb 2017 07:25:50 +0000 (16:25 +0900)]
evas/doc: Add warning about post_event_callback_push

7 years agoRevert "Revert "evas textblock: keep previous size when the calculation is skipped""
Carsten Haitzler (Rasterman) [Thu, 16 Feb 2017 01:18:54 +0000 (10:18 +0900)]
Revert "Revert "evas textblock: keep previous size when the calculation is skipped""

This reverts commit b8beb6834b01f9688cb5e7aa177bacbc713318b9.

this now actually works... for some mysterious reason... ? :/ i am
baffled. go back in until we can find the issue then...

7 years agodoxygen: fix typos and some wrong expressions in API reference.
Myoungwoon Roy, Kim [Thu, 16 Feb 2017 00:30:18 +0000 (16:30 -0800)]
doxygen: fix typos and some wrong expressions in API reference.

Summary: I had fixed some typos and some wrong expressions, such as capital letters, singular, and orders of groups in API reference doxygen.

Test Plan: Doxygen revision

Reviewers: stefan, cedric, raster, Jaehyun_Cho, jpeg

Subscribers: conr2d

Differential Revision: https://phab.enlightenment.org/D4658

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
7 years agoefl: fix memory leak in edje and evas.
Godly T.Alias [Thu, 16 Feb 2017 00:29:05 +0000 (16:29 -0800)]
efl: fix memory leak in edje and evas.

Summary:
Signed-off-by: Godly T.Alias <godlytalias@yahoo.co.in>
Signed-off-by: Ayush Srivastava <ayush.sriv@samsung.com>
Test Plan: EFL apps

Reviewers: cedric, raster

Subscribers: rajeshps, ayush.sriv, jpeg, prince.dubey

Differential Revision: https://phab.enlightenment.org/D4660

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
7 years agoecore_drm2: ERR once when page flip fails with EBUSY
Derek Foreman [Wed, 15 Feb 2017 15:51:32 +0000 (09:51 -0600)]
ecore_drm2: ERR once when page flip fails with EBUSY

On systems where this happens it'll probably happen a lot, so
we don't want to continuously log this, but since it's definitely
showing a bug somewhere (efl or kernel) it probably should be an ERR.

7 years agoedje: Fix object leak with Efl.Part
Jean-Philippe Andre [Wed, 15 Feb 2017 11:55:58 +0000 (20:55 +0900)]
edje: Fix object leak with Efl.Part

A call to efl_reuse was missing.

7 years agoevas_loader: add since tag to newly added task_cancelled API
Stefan Schmidt [Wed, 15 Feb 2017 11:07:02 +0000 (12:07 +0100)]
evas_loader: add since tag to newly added task_cancelled API

7 years agoeina_strbuf / eina_ustrbuf: add since tags for newly added APIs
Stefan Schmidt [Wed, 15 Feb 2017 10:13:36 +0000 (11:13 +0100)]
eina_strbuf / eina_ustrbuf: add since tags for newly added APIs

These six have been added in this cycle but missed the since tags.

7 years agoeina_lock: add since tag to now exposed eina_lock_debug API
Stefan Schmidt [Wed, 15 Feb 2017 10:07:42 +0000 (11:07 +0100)]
eina_lock: add since tag to now exposed eina_lock_debug API

This function was static inline before but form 1.19 onwards its EAPI so give
it a since tag.

7 years agoeina: binbuf: add since tags new append_slice and insert_slice APIs
Stefan Schmidt [Wed, 15 Feb 2017 09:57:11 +0000 (10:57 +0100)]
eina: binbuf: add since tags new append_slice and insert_slice APIs

7 years agofileselector: Add missing call to group_del
Jean-Philippe Andre [Wed, 15 Feb 2017 11:07:38 +0000 (20:07 +0900)]
fileselector: Add missing call to group_del

Remove unused refcount thingy and enforce call to smart del.

Fixes T4598

7 years agogroup: Track calls to group_del
Jean-Philippe Andre [Wed, 15 Feb 2017 11:07:11 +0000 (20:07 +0900)]
group: Track calls to group_del

After a long search I found that fileselector was not calling
super.group_del on deletion, leading to the use of dangling pointers.
So let's verify that group_del is properly called.

See T4598

7 years agoeo: Fix shadow variable warning
Jean-Philippe Andre [Wed, 15 Feb 2017 07:23:44 +0000 (16:23 +0900)]
eo: Fix shadow variable warning

7 years agogenlist: remove uncessary _item_cache_zero call
Amitesh Singh [Wed, 15 Feb 2017 10:18:38 +0000 (15:48 +0530)]
genlist: remove uncessary _item_cache_zero call

_item_cache_zero() is already getting called in _mirrored_set

Signed-off-by: Amitesh Singh <amitesh.sh@samsung.com>
7 years agoelm_genlist_item: update docs for 'next_get'
Vyacheslav Reutskiy [Fri, 10 Feb 2017 08:50:35 +0000 (10:50 +0200)]
elm_genlist_item: update docs for 'next_get'

7 years agotextblock: Fix dangling data ref
Jean-Philippe Andre [Wed, 15 Feb 2017 06:43:20 +0000 (15:43 +0900)]
textblock: Fix dangling data ref

In this case data_scope_get is more appropriate as the data is
indeed stored on the stack (function scope) and not somewhere else.

After this last fix I see no eo_debug error logs in elementary test.
Yay! eo_debug is now usable :)

7 years agoevas & elm: Fix invalid uses of efl_data_ref(obj, NULL)
Jean-Philippe Andre [Wed, 15 Feb 2017 06:04:20 +0000 (15:04 +0900)]
evas & elm: Fix invalid uses of efl_data_ref(obj, NULL)

The data class should be specified for debug purposes.
Also, this fixes invalid uses inside the smart object
implementation where it assumed that the smart data was part
of the eo data. It may not (legacy objects).

7 years agoevas_render: Remove data_unref that has no data_ref
Jean-Philippe Andre [Wed, 15 Feb 2017 06:32:04 +0000 (15:32 +0900)]
evas_render: Remove data_unref that has no data_ref

7 years agoeo_debug: Remove some abusive goto where not needed
Jean-Philippe Andre [Wed, 15 Feb 2017 06:25:15 +0000 (15:25 +0900)]
eo_debug: Remove some abusive goto where not needed

goto was used for micro-optimization. There is absolutely no
need for those if we're using the slow path with eo_debug.

Simplify the code.

7 years agoeo_debug: Some more log improvements
Jean-Philippe Andre [Wed, 15 Feb 2017 06:19:08 +0000 (15:19 +0900)]
eo_debug: Some more log improvements

Nothing fancy here...

7 years agoeo_debug: Improve dangling xref debug logs even more
Jean-Philippe Andre [Wed, 15 Feb 2017 05:43:35 +0000 (14:43 +0900)]
eo_debug: Improve dangling xref debug logs even more

In case of manual free, as is heavily used by Evas, we can't really
print an ERR if there are still references before free has been
called.

This may not be ideal from a pure EO point of view but considering
how Evas uses manual free this is the best solution to avoid
polluting debug logs.

7 years agoeo_debug: Improve one log
Jean-Philippe Andre [Wed, 15 Feb 2017 04:42:29 +0000 (13:42 +0900)]
eo_debug: Improve one log

This changes the following message when the object is referencing
its own data. Also lowers from ERR to WRN and adds the class
name for the referenced data.

ERR<17450>:eo /home/jpeg/e/core/efl/src/lib/eo/eo_private.h:337
  _efl_unref_internal() in /home/jpeg/e/core/efl/src/lib/eo/eo.c:620:
  func '_efl_object_call_end' Data of object 0x400000021008db58 is
  still referenced by object 0x400000021008db58

Note that evas and elm have a few calls to efl_data_ref(obj, NULL)
which are imho quite ugly: not using the return value and not
specifying the data class. I'm keeping them as-is for now.

7 years agoelm_widget: Fix order of operations (isa before data_get)
Jean-Philippe Andre [Wed, 15 Feb 2017 04:32:15 +0000 (13:32 +0900)]
elm_widget: Fix order of operations (isa before data_get)

This fixes some of the occurences of the following error message
when using eo_debug:

ERR<23101>:eo /home/jpeg/e/core/efl/src/lib/eo/eo.c:1955
  efl_data_scope_get() Tried getting data of class 'Elm.Widget'
  from object of class 'Edje.Object', but the former is not a
  direct inheritance of the latter.

If we're going to call efl_isa, then let's at least do it before
efl_data_scope_get.

7 years agoelc_multibuttonentry: input panel should be shown whenever mbe is clicked.
Woochan Lee [Wed, 15 Feb 2017 04:06:49 +0000 (13:06 +0900)]
elc_multibuttonentry: input panel should be shown whenever mbe is clicked.

Summary:
When mbe got focus the input panel will be shown.
Otherwise, input panel will be hidden.

If the mbe still got focus then input panel hide by user action.
there is no way to reactivate the input panel in case.

Reviewers: woohyun, jpeg, cedric, Hermet

Reviewed By: jpeg

Subscribers: cedric, jpeg

Differential Revision: https://phab.enlightenment.org/D4661

7 years agoecore_input: add API to set/get deadzone of joystick event for an axis.
Shinwoo Kim [Wed, 15 Feb 2017 04:00:13 +0000 (13:00 +0900)]
ecore_input: add API to set/get deadzone of joystick event for an axis.

Summary:
The axis type joystick event could occur without user's control if joystick is too sensitive.
The deadzone prevents this unnecessary event. The default value is 200.
The event value for an axis is a signed integer between -32767 and +32767.

Test Plan: Using example

Reviewers: raster, cedric, jpeg

Reviewed By: jpeg

Subscribers: stefan_schmidt

Differential Revision: https://phab.enlightenment.org/D4654

7 years agobuild: Add our wayland protocols to the dist
Derek Foreman [Tue, 14 Feb 2017 23:00:34 +0000 (17:00 -0600)]
build: Add our wayland protocols to the dist

Oops, without these in the distribution tarball you can't
build wayland support.

7 years agoemotion: add since tag to new emotion_file_meta_artwork_get API
Stefan Schmidt [Tue, 14 Feb 2017 20:42:20 +0000 (21:42 +0100)]
emotion: add since tag to new emotion_file_meta_artwork_get API

7 years agoefl: add since tag to new invalid object error type
Stefan Schmidt [Tue, 14 Feb 2017 20:33:33 +0000 (21:33 +0100)]
efl: add since tag to new invalid object error type

7 years agoeo: add missing since tags for functions added during 1.19 cycle
Stefan Schmidt [Tue, 14 Feb 2017 20:31:59 +0000 (21:31 +0100)]
eo: add missing since tags for functions added during 1.19 cycle

7 years agoecore_x: add singce tags for ecore_x_stack_* functions added during 1.19 cycle
Stefan Schmidt [Tue, 14 Feb 2017 20:16:00 +0000 (21:16 +0100)]
ecore_x: add singce tags for ecore_x_stack_* functions added during 1.19 cycle

7 years agoecore_evas: add brief docs and since tags for new ecore_evas_shadow_geometry_*
Stefan Schmidt [Tue, 14 Feb 2017 20:06:38 +0000 (21:06 +0100)]
ecore_evas: add brief docs and since tags for new ecore_evas_shadow_geometry_*

These have been missing when introduced during the 1.19 cycle.

7 years agoecore-wl2: Fix setting input region on a window
Chris Michael [Tue, 14 Feb 2017 16:31:57 +0000 (11:31 -0500)]
ecore-wl2: Fix setting input region on a window

Previous code here would store incorrect values into the
window->input rectangle. These values should be what is actually
getting set in the wl_region. This code also fixes an issue when
setting input region for transparent/alpha windows by clearing out
any pending input region (wl_surface_set_input_region(surface, null)).

Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
@fix

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoecore-wl2: Fix setting opaque region on a window
Chris Michael [Tue, 14 Feb 2017 14:58:49 +0000 (09:58 -0500)]
ecore-wl2: Fix setting opaque region on a window

Previous code here would store incorrect values into the
window->opaque rectangle. These values should be what is actually
getting set in the wl_region. This code also fixes an issue when
setting opaque region for transparent/alpha windows by clearing out
any pending opaque region (wl_surface_set_opaque_region(surface, null)).

Reviewed-By: Derek Foreman <derekf@osg.samsung.com>
@fix

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoRevert "evas textblock: keep previous size when the calculation is skipped"
Carsten Haitzler (Rasterman) [Tue, 14 Feb 2017 08:32:26 +0000 (17:32 +0900)]
Revert "evas textblock: keep previous size when the calculation is skipped"

This reverts commit c39855a8ac6f86c2dd8b1856ed70f85e0e521f3e.

This actually breaks 1 dialog in e (app exited with error exit code).
it worked everywhere else so i thought it was good. seemingly not
after i saw one of these. revert D3595

7 years agoEdje calc: Fix textblock size calculation logic
Youngbok Shin [Tue, 14 Feb 2017 07:16:45 +0000 (16:16 +0900)]
Edje calc: Fix textblock size calculation logic

Summary:
In singleline textblock, using "text.min: 1 0" and min, max width,
Edje allows to use expandable text with ellipsis. It shows ellipsis
when only text's width reach the max width.
But, Edje couldn't support same feature on multiline textblock.
Edje dose not use max height or text.max properly if ellipsis is enabled.
This feature is very useful to make a layout with dynamically aligned text.
@fix

Reviewers: cedric, tasn, woohyun, raster, herdsman

Subscribers: z-wony, eagleeye, jpeg

Differential Revision: https://phab.enlightenment.org/D3595

7 years agoevas textblock: keep previous size when the calculation is skipped
Youngbok Shin [Tue, 14 Feb 2017 07:16:26 +0000 (16:16 +0900)]
evas textblock: keep previous size when the calculation is skipped

Summary:
When a size calculation is skipped because of some reasons,
Evas Textblock should keep same size with the previous size.
@fix

Test Plan: N/A

Reviewers: raster, herdsman, cedric, jpeg

Differential Revision: https://phab.enlightenment.org/D4659

7 years agoecore_evas: Avoid safety check error log
Jean-Philippe Andre [Tue, 14 Feb 2017 06:09:36 +0000 (15:09 +0900)]
ecore_evas: Avoid safety check error log

If an engine can not be used (eg. GL inside a standard Xephyr),
some ERR logs will be printed out by ecore_evas_x. This patch
avoids extra eina_safety error logs from using a NULL pointer.