platform/upstream/efl.git
7 years agoedje: unset timer pointer after delete.
Amitesh Singh [Sat, 4 Feb 2017 07:48:55 +0000 (13:18 +0530)]
edje: unset timer pointer after delete.

Signed-off-by: Amitesh Singh <amitesh.sh@samsung.com>
7 years agoevas render - drop eo overhead by using ptrs we already have
Carsten Haitzler (Rasterman) [Sat, 4 Feb 2017 02:31:55 +0000 (11:31 +0900)]
evas render - drop eo overhead by using ptrs we already have

several calls, specifically evas_object_change_reset,
evas_object_cur_prev, and evas_object_clip_changes_clean that are
called directly or indirectly as part of evas render on at least every
active object if not more, were doing full eo obj lookups when their
calling functions already all had the eo protected data looked up.
tha's silly and just adds overhead we don't need. my test dropped
_eo_obj_pointer_get overhead in perf profiles from 4.48% to 2.65%. see:

   4.48%  libeo.so.1.18.99            [.] _eo_obj_pointer_get
   4.23%  libevas.so.1.18.99          [.] evas_render_updates_internal
   2.61%  libevas.so.1.18.99          [.] evas_render_updates_internal_loop
   1.68%  libeo.so.1.18.99            [.] efl_data_scope_get
   1.57%  libc-2.24.so                [.] _int_malloc
   1.42%  libevas.so.1.18.99          [.] evas_object_smart_changed_get
   1.09%  libevas.so.1.18.99          [.] evas_object_clip_recalc.part.37
   1.08%  libpthread-2.24.so          [.] pthread_getspecific
   1.05%  libevas.so.1.18.99          [.] efl_canvas_object_class_get
   1.01%  libevas.so.1.18.99          [.] evas_object_cur_prev
   0.99%  libeo.so.1.18.99            [.] _efl_object_event_callback_legacy_call
   0.87%  libevas.so.1.18.99          [.] _evas_render_phase1_object_ctx_render_cache_append
   0.82%  libpthread-2.24.so          [.] pthread_mutex_lock
   0.81%  libevas.so.1.18.99          [.] _evas_render_phase1_object_process
   0.79%  libc-2.24.so                [.] _int_free

vs now the improved:

   4.82%  libevas.so.1.18.99          [.] evas_render_updates_internal
   3.44%  libevas.so.1.18.99          [.] evas_render_updates_internal_loop
   2.65%  libeo.so.1.18.99            [.] _eo_obj_pointer_get
   2.22%  libc-2.24.so                [.] _int_malloc
   1.46%  libevas.so.1.18.99          [.] evas_object_smart_changed_get
   1.04%  libeo.so.1.18.99            [.] _efl_object_event_callback_legacy_call
   1.03%  libevas.so.1.18.99          [.] _evas_render_phase1_object_ctx_render_cache_append
   0.97%  libeina.so.1.18.99          [.] eina_chained_mempool_malloc
   0.93%  libevas.so.1.18.99          [.] evas_object_clip_recalc.part.37
   0.92%  libpthread-2.24.so          [.] pthread_mutex_lock
   0.91%  libevas.so.1.18.99          [.] _evas_render_phase1_object_process
   0.84%  libc-2.24.so                [.] _int_free
   0.84%  libevas.so.1.18.99          [.] evas_object_cur_prev
   0.83%  libeina.so.1.18.99          [.] eina_chained_mempool_free
   0.80%  libeo.so.1.18.99            [.] efl_data_scope_get

of course other things "increase their percentage" as oe overhead now
dropped, and things seem to move around a bit, but it does make sense
to do this with no downsides i can see as we already are accessing the
protected data ptr in the parent func.

7 years agoecore_evas: use ecore_timer_loop_add for idle timeout update
Derek Foreman [Fri, 3 Feb 2017 21:54:48 +0000 (15:54 -0600)]
ecore_evas: use ecore_timer_loop_add for idle timeout update

This shouldn't need "exact" time, so use a loop time based timer
to potentially save a syscall on archs that don't have a vdso.

@optimize

7 years agoecore: poller could actually be related to the main loop timing.
Cedric BAIL [Fri, 3 Feb 2017 19:12:28 +0000 (11:12 -0800)]
ecore: poller could actually be related to the main loop timing.

Reduce call to ecore_time_get which is a syscall to gettimeofday.

7 years agoelput: add event for relative motion
Mike Blumenkrantz [Fri, 3 Feb 2017 20:14:01 +0000 (15:14 -0500)]
elput: add event for relative motion

@feature

7 years agoelm_code: fix bugs in the highlighting of full line comments
Andy Williams [Fri, 3 Feb 2017 17:17:05 +0000 (17:17 +0000)]
elm_code: fix bugs in the highlighting of full line comments

@fix

7 years agoelm_code: Add syntax highlighting for python
Andy Williams [Fri, 3 Feb 2017 17:16:12 +0000 (17:16 +0000)]
elm_code: Add syntax highlighting for python

7 years agoelm_code: update syntax symbols for C
Andy Williams [Fri, 3 Feb 2017 17:13:27 +0000 (17:13 +0000)]
elm_code: update syntax symbols for C

Also don't hard code preprocessing parse

7 years agocmake: sync EFL_EO_API_SUPPORT definition with autotools.
Gustavo Sverzut Barbieri [Fri, 3 Feb 2017 14:17:05 +0000 (12:17 -0200)]
cmake: sync EFL_EO_API_SUPPORT definition with autotools.

in autotools we always define that symbol in config.h (thus now done
in common.cmake), while the Efl_Config.h that is generated never have
it, what's left is to enable/disable the legacy and beta APIs.

This fixes building eldbus_suite with cmake.

7 years agoifdef RUN_IN_TREE logic.
Gustavo Sverzut Barbieri [Fri, 3 Feb 2017 13:49:38 +0000 (11:49 -0200)]
ifdef RUN_IN_TREE logic.

This logic is only needed for autotools, cmake will replicate the
installation file structure and thus eina_prefix works out of box.

7 years agocmake: add ecore_input
Marcel Hollerbach [Fri, 3 Feb 2017 10:14:45 +0000 (11:14 +0100)]
cmake: add ecore_input

7 years agocmake: add eeze
Marcel Hollerbach [Fri, 3 Feb 2017 09:39:23 +0000 (10:39 +0100)]
cmake: add eeze

there is currently a bug in the testsuite which needs to change,
otherwise modules cannot be loaded. The code acutally runs when its
installed and EFL_RUN_IN_TREE is not set.

7 years agoeeze: move binaries into subdirectories
Marcel Hollerbach [Thu, 2 Feb 2017 23:04:19 +0000 (00:04 +0100)]
eeze: move binaries into subdirectories

in order to move to cmake we are moving every binary into its own
subdirectory.

7 years agoecore_con/efl_net - dedicated threads for network i/o not from the pool
Carsten Haitzler (Rasterman) [Fri, 3 Feb 2017 05:57:49 +0000 (14:57 +0900)]
ecore_con/efl_net - dedicated threads for network i/o not from the pool

so ecore_con/efl_net were using the standard ecore_thread thread pool
for doing things like dns lookups (that can take multiple minutes
until timeouts) and actual http transactions. similarly they can block
thread workers for long periods or indefinitely thus basically
blocking the whole eocre_thread pool and stopping others from sharing
it. the best solution we have right now is to bypass the thread pool
queue and have dedicated threads for these actions. what we should
have is a dedicated thread pool with each thread taking on N
connections (via select etc.) and the ability to create and destroy
thread pools for specific tasks so you can separate the work out from
other work.  but that is basically a redesign of our thread pool infra
so let's do the quick solution here until that day comes.

this partially addresses D4640

a dedicated thread per image load though is going to be a lot nastier...

7 years agoedje - fix crash on loading ancient edje files
Carsten Haitzler (Rasterman) [Fri, 3 Feb 2017 05:39:57 +0000 (14:39 +0900)]
edje - fix crash on loading ancient edje files

ancient edje files dont put hash strings in the file dictionary but
instead inline and thus the strings cannot be direct added... silly
ancient files.

this is a horrible ugly workaround this but there isn't really
anything better than trying to detect such files (which dont seem to
have an older version in them i can find?) so detect by pointer
address and mapping the file.

this fixes T5138

@fix

7 years agoecore: implement new requirement in the test suite.
Cedric BAIL [Fri, 3 Feb 2017 00:52:38 +0000 (16:52 -0800)]
ecore: implement new requirement in the test suite.

7 years agoecore: avoid calling ecore_time_get everytime we create a timer.
Cedric BAIL [Fri, 3 Feb 2017 00:52:10 +0000 (16:52 -0800)]
ecore: avoid calling ecore_time_get everytime we create a timer.

This is particularly useful for ecore_timer_loop_add which are not
supposed to trigger any syscall. It does add a new requirement
on the EFL_LOOP_TIME_CLASS to always have an interval defined during
construction.

7 years agoCpumonitor theme - Set an aspect preference other than both so we can properlyl size...
Stephen 'Okra' Houston [Thu, 2 Feb 2017 21:02:15 +0000 (15:02 -0600)]
Cpumonitor theme - Set an aspect preference other than both so we can properlyl size the gadget.

7 years agocmake: add ecore_file.
Gustavo Sverzut Barbieri [Thu, 2 Feb 2017 19:04:44 +0000 (17:04 -0200)]
cmake: add ecore_file.

7 years agocmake: add ecore_ipc.
Gustavo Sverzut Barbieri [Thu, 2 Feb 2017 16:43:40 +0000 (14:43 -0200)]
cmake: add ecore_ipc.

7 years agoelm_code: Fix empty selection checks
Andy Williams [Thu, 2 Feb 2017 18:47:39 +0000 (18:47 +0000)]
elm_code: Fix empty selection checks

7 years agoevas: simplify waiting logic for image decoding as we do not need lock and cond with...
Cedric BAIL [Thu, 2 Feb 2017 18:29:59 +0000 (10:29 -0800)]
evas: simplify waiting logic for image decoding as we do not need lock and cond with Ecore_Thread infra.

7 years agoevas: make sure we always wait in the correct thread.
Cedric BAIL [Thu, 2 Feb 2017 18:29:39 +0000 (10:29 -0800)]
evas: make sure we always wait in the correct thread.

7 years agoelm_code: add support for redo
YeongJong Lee [Thu, 2 Feb 2017 18:07:02 +0000 (18:07 +0000)]
elm_code: add support for redo

Summary:
now, elm_code support redo.
shortcut is <Ctrl> + <y>

Test Plan:
1. elementary_test - Code Editor
2. Check that undo and redo are work correctly.

Reviewers: ajwillia.ms

Reviewed By: ajwillia.ms

Subscribers: cedric, jpeg

Differential Revision: https://phab.enlightenment.org/D4642

7 years agoevlog - fix "remove setting stack var to 0" that broke evlogs
Carsten Haitzler (Rasterman) [Thu, 2 Feb 2017 01:13:24 +0000 (10:13 +0900)]
evlog - fix "remove setting stack var to 0" that broke evlogs

this fixes b38c610c7e60cbdcb45fd6d27450e39e3a9c9dd1 - recent commit by
me. bad me! spank!

7 years agoevas grid smart - fix divide by 0
Carsten Haitzler (Rasterman) [Wed, 1 Feb 2017 23:50:11 +0000 (08:50 +0900)]
evas grid smart - fix divide by 0

if object happens to be 0x0 then we can get a divide by 0... fix!

@fix

7 years agoecore_wl2: Make www events opt-in
Derek Foreman [Wed, 1 Feb 2017 22:21:34 +0000 (16:21 -0600)]
ecore_wl2: Make www events opt-in

Now we must set EFL_WAYLAND_ENABLE_WWW to turn on www.  (Note:
this does NOT re-enable any visual effect, it merely controls
the event infrastructure that can be used to let a client
know when its window is mapped or moving)

This wasn't really disabled properly earlier, so disable it now
for a slight performance win when dragging windows (and a major
reduction in log chatter when debugging wayland)

7 years agoedje: change seat signals
Bruno Dilly [Wed, 1 Feb 2017 19:14:39 +0000 (17:14 -0200)]
edje: change seat signals

Use seat names as prefix, not as suffix, following
a top-down approach and avoiding issues
with applications that may be receiving
doubled signals (legacy + suffixed).

So instead of "mouse,in,seat1" signal will be
"seat,seat1,mouse,in".

7 years agoedje: fix CID 1368019 - Null pointer dereferences
Bruno Dilly [Thu, 26 Jan 2017 20:26:23 +0000 (18:26 -0200)]
edje: fix CID 1368019 - Null pointer dereferences

It only should check if it's allowed to focus a part
if a part was passed (if rp is null such check should
be skipped).

7 years agoefl_net_dialer_http: check for finished handles after unpause.
Gustavo Sverzut Barbieri [Wed, 1 Feb 2017 18:23:11 +0000 (16:23 -0200)]
efl_net_dialer_http: check for finished handles after unpause.

After we call curl_multi_socket_action() we must call
_efl_net_dialer_http_curlm_check() in order to call
curl_multi_info_read() and be notified of handles that were
finished. Otherwise the handle is gone and we'll be waiting for an
action that will never happen.

Fixes T5079

7 years agobuild: Fix screw up with wayland header auto generation
Derek Foreman [Wed, 1 Feb 2017 17:29:40 +0000 (11:29 -0600)]
build: Fix screw up with wayland header auto generation

7 years agocmake: handle 'FindXXX.cmake' standard with 'find-XXX' libraries.
Gustavo Sverzut Barbieri [Wed, 1 Feb 2017 14:25:43 +0000 (12:25 -0200)]
cmake: handle 'FindXXX.cmake' standard with 'find-XXX' libraries.

One can use virtual 'find-XXX' entries in LIBRARIES or
PUBLIC_LIBRARIES to use XXX_LIBRARIES and XXX_INCLUDE_DIR.

7 years agocmake: handle support libraries in .pc
Gustavo Sverzut Barbieri [Wed, 1 Feb 2017 14:25:08 +0000 (12:25 -0200)]
cmake: handle support libraries in .pc

support libraries (src/static_libs), such as rg_etc, must not be
handled as a regular target, after all these are usually not shared
library and if they are, they have no .pc associated, then we use "-l"
as for system libraries, but using the installation path.

7 years agocmake: remove dependency on libjpeg in eet as it is handled in common
Stefan Schmidt [Wed, 1 Feb 2017 13:22:11 +0000 (14:22 +0100)]
cmake: remove dependency on libjpeg in eet as it is handled in common

7 years agoRevert "edje_edit: fix segmentation fault(Array index is out of bound)"
Vyacheslav Reutskiy [Wed, 1 Feb 2017 13:08:13 +0000 (15:08 +0200)]
Revert "edje_edit: fix segmentation fault(Array index is out of bound)"

This patch need to revert because it just a temporary solution for load
'inherit_only' groups. Problem with these groups is solved in patch
c3733a213ba2de3acc2496a381e8e3f591f2c10a

This reverts commit 88f5ac22c1dcd2ddaf4fda4f170fb5c0ad9b3afd.

7 years agoedje_cc: remove 'inherit_only' groups from result edj file
Vyacheslav Reutskiy [Wed, 1 Feb 2017 13:02:46 +0000 (15:02 +0200)]
edje_cc: remove 'inherit_only' groups from result edj file

The 'inherit_only' groups are helpers and no sense to save it to edj
file.

@fix

7 years agocmake: emile is depending on efl
Marcel Hollerbach [Wed, 1 Feb 2017 11:19:24 +0000 (12:19 +0100)]
cmake: emile is depending on efl

there is some .eot.h file included from the efl interfaces, so we need a
dependency here

7 years agocmake: use the jpeg module of cmake to find the library
Marcel Hollerbach [Wed, 1 Feb 2017 11:18:33 +0000 (12:18 +0100)]
cmake: use the jpeg module of cmake to find the library

finding libjpeg can be a bit more complex, there are systems that are
not delivering a .pc for it. So we should just use the cmake module to
find it.

7 years agocmake: fix efl version number
Stefan Schmidt [Wed, 1 Feb 2017 10:42:03 +0000 (11:42 +0100)]
cmake: fix efl version number

We are in the merge window for 1.19 so our number is actually 1.18.99

7 years agocmake: downgrade minimum cmake version to 3.4
Stefan Schmidt [Wed, 1 Feb 2017 10:40:25 +0000 (11:40 +0100)]
cmake: downgrade minimum cmake version to 3.4

There seems really no reason to have such a new version as 3.7 as minimum
here. MAybe even go down to 3.1 as suggested on the mailing list.

7 years agoelm_code: add copy/paste/cut undo
YeongJong Lee [Wed, 1 Feb 2017 09:08:14 +0000 (09:08 +0000)]
elm_code: add copy/paste/cut undo

Summary: The patch for supporting copy,paste and cut undo

Test Plan:
1. elementary_test - Code Editor
2. Check that undo work correctly when copy,paste and cut

Reviewers: ajwillia.ms

Reviewed By: ajwillia.ms

Subscribers: cedric, jpeg

Differential Revision: https://phab.enlightenment.org/D4636

7 years agoDnD/Container: fix drag during animation
Daniel Zaoui [Fri, 27 Jan 2017 10:08:55 +0000 (12:08 +0200)]
DnD/Container: fix drag during animation

If the user icons are set but animation time is set to 0, the icons are
not deleted, resulting in ghosts images in the application.
Additionally, behavior was badly handled when animation time or timeout
before drag were set to 0.

@fix

7 years agoecore_input: support Xbox One joystick
Shinwoo Kim [Wed, 1 Feb 2017 07:33:52 +0000 (16:33 +0900)]
ecore_input: support Xbox One joystick

Add joystick mapper for Microsoft X-Box One pad (Covert Forces)

7 years agoevas gl common - reduce data structure sizes got gl common context
Carsten Haitzler (Rasterman) [Wed, 1 Feb 2017 07:08:04 +0000 (16:08 +0900)]
evas gl common - reduce data structure sizes got gl common context

this reduced pipes to 32 as default max pipes is 32 and we never
actually go  up beyond 32 unless oyu use env vars to expand... and no
one will... and actual testing shows us top out at about 30 pipes
usage in anything i can find/test. this packs more bool flags together
as bits instead of itns or chars to save memory.

@optimize

7 years agoEldbus: add test suite for eldbus (eldbus_object*)
Mykola Solyanko [Wed, 1 Feb 2017 04:10:18 +0000 (13:10 +0900)]
Eldbus: add test suite for eldbus (eldbus_object*)

Summary:
add tests for next API:
eldbus_object_bus_name_get
eldbus_object_connection_get
eldbus_object_method_call_new
eldbus_object_path_get
eldbus_object_ref
eldbus_object_send
eldbus_object_get
eldbus_object_introspect
eldbus_object_peer_ping
eldbus_object_peer_machine_id_get
eldbus_object_get

Reviewers: cedric, NikaWhite, myoungwoon, raster

Subscribers: m.solyanko, jpeg, artem.popov

Differential Revision: https://phab.enlightenment.org/D4635

7 years agoelm_code: fix that selected code can't undo
YeongJong Lee [Tue, 31 Jan 2017 19:04:33 +0000 (19:04 +0000)]
elm_code: fix that selected code can't undo

Summary: Change some undo logic and remove unused function

Test Plan:
1. elementary_test - Code Editor
2. Select lines.
3. Remove selected lines.
4. undo removed lines(<Ctrl> + <z>)

Reviewers: ajwillia.ms

Reviewed By: ajwillia.ms

Subscribers: cedric, jpeg

Differential Revision: https://phab.enlightenment.org/D4631

7 years agoelput: Improve debug output for added devices
Chris Michael [Tue, 31 Jan 2017 15:55:55 +0000 (10:55 -0500)]
elput: Improve debug output for added devices

Small patch which does not print out debug info for devices we do not
handle, and adds the 'type' of device to the debug output (pointer,
keyboard, etc).

@fix

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoelput: Don't add devices which we are not going to use
Chris Michael [Tue, 31 Jan 2017 15:54:09 +0000 (10:54 -0500)]
elput: Don't add devices which we are not going to use

If an input device is not a pointer, keyboard, or touch device then we
are not going to be handling events for it, so do not create a new
Elput device for it.

@fix

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoRevert "elput: Change return of device_add/remove functions"
Chris Michael [Tue, 31 Jan 2017 15:53:30 +0000 (10:53 -0500)]
Revert "elput: Change return of device_add/remove functions"

Reverting this in favor of a better way to improve debug output

This reverts commit cee7ee39932c31a21d3dd90e19703371a946a0c0.

7 years agoelput: Change return of device_add/remove functions
Chris Michael [Tue, 31 Jan 2017 15:34:22 +0000 (10:34 -0500)]
elput: Change return of device_add/remove functions

Minor change on the return value of the device_add and device_remove
functions so that we can avoid debug output for devices which we do
not deal with.

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoecore-evas-wayland: Minor formatting fix
Chris Michael [Tue, 31 Jan 2017 15:29:50 +0000 (10:29 -0500)]
ecore-evas-wayland: Minor formatting fix

NB: No functional changes, just formatting

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoelput: Add more robust debugging output for input devices
Chris Michael [Tue, 31 Jan 2017 15:29:25 +0000 (10:29 -0500)]
elput: Add more robust debugging output for input devices

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoeina log bt - make it an eaven easier copy & paste to get a bt now
Carsten Haitzler (Rasterman) [Tue, 31 Jan 2017 10:30:41 +0000 (19:30 +0900)]
eina log bt - make it an eaven easier copy & paste to get a bt now

just copy & paste into a terminal and hit enter.

7 years agoevas gl common - have a gl array buffer pool to avoid hitting malloc
Carsten Haitzler (Rasterman) [Tue, 31 Jan 2017 09:04:47 +0000 (18:04 +0900)]
evas gl common - have a gl array buffer pool to avoid hitting malloc

this improves performance a bit by avoding libc's malloc and the heap
by using a custom buffer pool thathas basically immediate O(1) lookup
time (in real life) that baloons out and contracts back as time goes on.

@optimize

7 years agoevas draw context - reduce size and allocations and frees
Carsten Haitzler (Rasterman) [Tue, 31 Jan 2017 05:50:52 +0000 (14:50 +0900)]
evas draw context - reduce size and allocations and frees

so a little perf fun shows malloc/free/realloc/etc. are, combined a
reasonable overhead. this reduced malloc overhead for draw contexts so
whne we duplicate them or create new ones, we re-use a small cache of
8 of them to avoid re-allocation. just take the first one from the
list as it really is that simple. mempool would not have helped more
here and cost more overhead.

@optimize

7 years agoevlog - remove setting stack var to 0 until it is needed later
Carsten Haitzler (Rasterman) [Tue, 31 Jan 2017 05:48:36 +0000 (14:48 +0900)]
evlog - remove setting stack var to 0 until it is needed later

we set stack var to 0 even if evlog was off and thus didn't use it.
this cleans up the evlog func a bit and also moves locking until later
so it's locked for the minimum period to punt something into the log
buffer. it's an improvement, but no bug fix.

7 years agoevas async preload - fix valgrind issue in accessing image after free
Carsten Haitzler (Rasterman) [Tue, 31 Jan 2017 05:47:36 +0000 (14:47 +0900)]
evas async preload - fix valgrind issue in accessing image after free

cache flish and async end ended up dropping some cacneled images that
were then accessed further in the cancel func for the thread. this
fixes that

@fix

7 years agoedje_pick: implement fonts copy
Vyacheslav Reutskiy [Tue, 31 Jan 2017 08:11:36 +0000 (10:11 +0200)]
edje_pick: implement fonts copy

Previously this functional was miss. Edje_Pick was tried to copy fonts
from data block "edje_source_fontmap" what always empty because edje_cc
never fill it.

This patch rework the fonts copy from input files.

@fix

7 years agoeina_evlog: Don't call get_time prematurely
Derek Foreman [Mon, 30 Jan 2017 21:28:29 +0000 (15:28 -0600)]
eina_evlog: Don't call get_time prematurely

If we're not logging events this generates a lot of wasted system
calls.  They probably don't amount to much, but it's trivial to
get rid of them, and they make a mess when logging with strace.

7 years agoelementary: Ensure pointer hotspot update and reduce duplicated code
Chris Michael [Mon, 30 Jan 2017 17:59:29 +0000 (12:59 -0500)]
elementary: Ensure pointer hotspot update and reduce duplicated code

When we mouse_in on a window, the cursor hotspot may need updating for
the new pointer image, so utilize an existing function we already have
which will set the pointer image and calculate updated hotspot. This
patch also uses the same function call in _elm_win_frame_cb_move_stop
in order to reduce duplication.

ref T4987

@fix

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoelementary: Don't hide pointer window on mouse out
Chris Michael [Mon, 30 Jan 2017 17:37:44 +0000 (12:37 -0500)]
elementary: Don't hide pointer window on mouse out

This patch fixes an issue where pointers would occasionally disappear
when running EFL apps in a Wayland compositor. This was occuring
because we would hide the pointer window on mouse_out (and thus attach a NULL
buffer to the pointer surface), but then when we mouse_in again on the
window, it still have a NULL buffer attached to that pointer surface.
This patch fixes the issue.

ref T4987

@fix

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoelementary: No need to reset pointer surface on resize
Chris Michael [Mon, 30 Jan 2017 17:24:27 +0000 (12:24 -0500)]
elementary: No need to reset pointer surface on resize

As the pointer canvas will already be visible at this point, there is
no need to redo the wl_pointer_set_cursor call as we can just set the
pointer image we desire on the surface (_elm_theme_object_set) and
when the pointer canvas renders, it will have the proper cursor

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoelementary: No need to set null pointer surface on mouse out
Chris Michael [Mon, 30 Jan 2017 17:21:49 +0000 (12:21 -0500)]
elementary: No need to set null pointer surface on mouse out

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoecore-evas: Remove unused 'just_mapped' flag
Chris Michael [Mon, 30 Jan 2017 16:50:05 +0000 (11:50 -0500)]
ecore-evas: Remove unused 'just_mapped' flag

As this flag is not being used anywhere for anything, remove it.
Useless to have it.

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoecore-evas-wayland: Only fetch framespace values if we have a window
Chris Michael [Mon, 30 Jan 2017 16:07:07 +0000 (11:07 -0500)]
ecore-evas-wayland: Only fetch framespace values if we have a window

Minor optimization to avoid always fetching framespace values even if
we were not going to use them.

@optimize

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoelementary: Create wayland pointer before frame
Chris Michael [Mon, 30 Jan 2017 16:03:57 +0000 (11:03 -0500)]
elementary: Create wayland pointer before frame

If we create the pointer After the frame, then we end up setting
elm_object_cursor on the frame object which we do not need so this
patch just slightly modifies the order of things during finalize so
that _elm_win_frame_add does not end up calling
_elm_object_part_cursor_set on the frame object.

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agoelementary: Don't set wayland window pointer if pointer ee not visible
Chris Michael [Mon, 30 Jan 2017 15:58:55 +0000 (10:58 -0500)]
elementary: Don't set wayland window pointer if pointer ee not visible

This small patch avoids resetting ecore_wl2_window_pointer (and thus
wl_pointer_set_cursor) if the actual pointer canvas is not visible.

Signed-off-by: Chris Michael <cp.michael@samsung.com>
7 years agomem gadget: document all the values received in the edje message
Dave Andreoli [Sun, 29 Jan 2017 13:51:57 +0000 (14:51 +0100)]
mem gadget: document all the values received in the edje message

So that themers are aware of the unused fields

7 years agoelm_code: Add a safety code to prevent arithmetic exception
YeongJong Lee [Sat, 28 Jan 2017 12:38:37 +0000 (12:38 +0000)]
elm_code: Add a safety code to prevent arithmetic exception

Test Plan:
1. elementary_test - Code Editor
2. Chack there is not arithmetic exception

Reviewers: ajwillia.ms

Reviewed By: ajwillia.ms

Subscribers: cedric, jpeg

Differential Revision: https://phab.enlightenment.org/D4621

7 years agocmake: add ecore_con.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 19:25:12 +0000 (17:25 -0200)]
cmake: add ecore_con.

7 years agocmake: pkg-config names uses dashes and are not uniform.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 19:22:24 +0000 (17:22 -0200)]
cmake: pkg-config names uses dashes and are not uniform.

ecore_con provides ecore-con, but ethumb_client provides
ethumb_client. That means we need replace '_' with '-' and allow to
override that, so ethumb_client will work.

7 years agocmake: create and install 'checkme' files for libraries.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 18:54:11 +0000 (16:54 -0200)]
cmake: create and install 'checkme' files for libraries.

7 years agocmake: use CMAKE_INSTALL_LIBDIR instead of 'lib'
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 18:53:14 +0000 (16:53 -0200)]
cmake: use CMAKE_INSTALL_LIBDIR instead of 'lib'

this matches the platform preference whenever lib, lib32 or lib64.

7 years agocmake: EFL_BIN() now checks for EFL_UTILITY and installs to proper location.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 17:27:51 +0000 (15:27 -0200)]
cmake: EFL_BIN() now checks for EFL_UTILITY and installs to proper location.

utility binaries shouldn't be in bin ($PATH), but in a
library-specific directory (lib/${libname}/utils/v-${maj}.${min})

7 years agocmake/eo: expose HAVE_EO_ID to build via config_gen.h
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 17:26:07 +0000 (15:26 -0200)]
cmake/eo: expose HAVE_EO_ID to build via config_gen.h

7 years agoecore_con: add missing string nul byte terminator to efl_net_ip_address.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 17:25:23 +0000 (15:25 -0200)]
ecore_con: add missing string nul byte terminator to efl_net_ip_address.

7 years agoeolian gen: imply -gT if -o T:f is specified
Daniel Kolesa [Fri, 27 Jan 2017 16:11:43 +0000 (17:11 +0100)]
eolian gen: imply -gT if -o T:f is specified

If you specify type-explicit output filename, this automatically
adds it so that it's generated even without specifying -gT.

This is meant to help user friendliness. Also add a note about
this to help listing.

7 years agocmake: remove comment, it was my misunderstanding, not a bug.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 15:57:10 +0000 (13:57 -0200)]
cmake: remove comment, it was my misunderstanding, not a bug.

7 years agocmake/eolian: generate source, headers and legacy in one command.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 15:53:15 +0000 (13:53 -0200)]
cmake/eolian: generate source, headers and legacy in one command.

actually wasn't a bug, I need to specify -gchl to generate legacy as
well.

7 years agocmake: fix make-clean as add_custom_command() deletes output files.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 15:34:25 +0000 (13:34 -0200)]
cmake: fix make-clean as add_custom_command() deletes output files.

there is no way to mark output file as "precious", then cmake's
suggestion is to use add_custom_target() instead.

However that will always execute, so our generator script needs to be
smarter and only touch stuff when actually needed.

7 years agocmake: add EFL_SUPPORT_LIB() and simplify/speedup its usage.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 14:38:52 +0000 (12:38 -0200)]
cmake: add EFL_SUPPORT_LIB() and simplify/speedup its usage.

generate a static library for src/static_libs and use that as
LIBRARIES for the actual library, for those such as rg_etc that are
used multiple times will even speed up the final build by compiling
only once.

Although not used, they can be made into shared libraries that would
go inside /usr/lib/efl/support/v-1.19/libname.so

7 years agocmake: add ector.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 12:45:20 +0000 (10:45 -0200)]
cmake: add ector.

7 years agocmake/tests: improve naming and add missing include directories.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 12:34:58 +0000 (10:34 -0200)]
cmake/tests: improve naming and add missing include directories.

Some tests were relocated to subdirectory but need shared files, or
may need files from src/tests by including them relatively (../) as
done in ector.

With subdirectories we may end with src/test/libname/suite, to make it
more user friendly and backward compatible call these libname_suite.

7 years agocmake: auto-include subdirectories with generated .eo.*
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 12:13:34 +0000 (10:13 -0200)]
cmake: auto-include subdirectories with generated .eo.*

7 years agocmake: if path is absolute, include as-is.
Gustavo Sverzut Barbieri [Fri, 27 Jan 2017 12:03:44 +0000 (10:03 -0200)]
cmake: if path is absolute, include as-is.

7 years agocmake: add embryo.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 18:36:21 +0000 (16:36 -0200)]
cmake: add embryo.

7 years agocmake: add missing files (oops)
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 18:22:49 +0000 (16:22 -0200)]
cmake: add missing files (oops)

7 years agocmake: enable systemd/upower ecore modules.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 18:21:17 +0000 (16:21 -0200)]
cmake: enable systemd/upower ecore modules.

eldbus was a dependency.

7 years agocmake: add eldbus.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 18:16:15 +0000 (16:16 -0200)]
cmake: add eldbus.

7 years agocmake: improve options summary with per-library values.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 17:50:05 +0000 (15:50 -0200)]
cmake: improve options summary with per-library values.

segment the options based on libraries, makes it easier to read.

7 years agocmake: fix strlcpy() detection and usage.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 17:35:35 +0000 (15:35 -0200)]
cmake: fix strlcpy() detection and usage.

Linux usually ship with strlcpy(), but it's inside bsd/string.h and
needs -lbsd.

On BSD it's inside string.h and libc.

7 years agocmake: more headers and function checks.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 17:25:17 +0000 (15:25 -0200)]
cmake: more headers and function checks.

7 years agodocgen: link to the method an overriden one overrides
Daniel Kolesa [Thu, 26 Jan 2017 16:50:16 +0000 (17:50 +0100)]
docgen: link to the method an overriden one overrides

7 years agodocgen: autodocument inherited methods/properties
Daniel Kolesa [Thu, 26 Jan 2017 16:40:16 +0000 (17:40 +0100)]
docgen: autodocument inherited methods/properties

7 years agocmake/eolian: fix EOLIAN_BIN type.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 16:48:20 +0000 (14:48 -0200)]
cmake/eolian: fix EOLIAN_BIN type.

So EOLIAN_BIN had no type, which resulted in BOOL and thus ON/OFF. It
should be labeled as FILEPATH and then we can remove the special case
in EflMacros.cmake.

7 years agocmake: define EFL_{LIB}_BUILD and DLL_EXPORT.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 16:20:52 +0000 (14:20 -0200)]
cmake: define EFL_{LIB}_BUILD and DLL_EXPORT.

These are used on Windows builds to force dllexport/dllimport.

7 years agocmake: add eet.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 16:02:21 +0000 (14:02 -0200)]
cmake: add eet.

7 years agocmake: auto-detect binaries in src/bin/libname if no CMakeLists.txt
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 15:58:26 +0000 (13:58 -0200)]
cmake: auto-detect binaries in src/bin/libname if no CMakeLists.txt

similar to tests, binaries should also follow the same rule and allow
single-binaries with single-source.

7 years agocmake: auto-detect tests in src/tests/libname if no CMakeLists.txt
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 15:56:49 +0000 (13:56 -0200)]
cmake: auto-detect tests in src/tests/libname if no CMakeLists.txt

previously we were only auto-detecting sources in src/tests/libname/
subdirectories, but we should also check the parent directory if no
subdirectories were processed.

7 years agocmake: fix test include and defines.
Gustavo Sverzut Barbieri [Thu, 26 Jan 2017 15:54:19 +0000 (13:54 -0200)]
cmake: fix test include and defines.

TESTS_WD is prefixed to TESTS_SRC_DIR and should be empty (no idea why
it's like that, maybe to allow relative paths).

TESTS_SRC_DIR must be defined to test source directory and a typo was
preventing that.