platform/upstream/efl.git
8 years agomap: don't update surface unencessarily.
ChunEon Park [Thu, 10 Sep 2015 12:27:47 +0000 (21:27 +0900)]
map: don't update surface unencessarily.

smart object don't need to update map surface always
check the state more elaborately so minus one more surface drawing!

8 years agoefl - set thread names for internally created threads for debugging
Carsten Haitzler (Rasterman) [Thu, 10 Sep 2015 06:17:08 +0000 (15:17 +0900)]
efl - set thread names for internally created threads for debugging

@feature

8 years agoeina - thread naming - when naming thread, cut off at 15 chars
Carsten Haitzler (Rasterman) [Thu, 10 Sep 2015 06:15:01 +0000 (15:15 +0900)]
eina - thread naming - when naming thread, cut off at 15 chars

automatically cut off at 15 chars (and add 0 byte) to ensure we get a
name, but just truncate it if it is too long.

8 years agoeina thread - add wrapper for setting thread name and fix affinity
Carsten Haitzler (Rasterman) [Wed, 9 Sep 2015 03:45:47 +0000 (12:45 +0900)]
eina thread - add wrapper for setting thread name and fix affinity

added support to set thread names for debugging. this wraps the gnu
extension to set a thread name of up to 15 chars (16 inc nul byte).
eina_thread_name_set() is the new api.
@feature

in the process i found the autofoo wasnt enabling cpu affinity support
at all in reality, so i had to fix that at the same time.

8 years agoevas: fix memory leak in evas_object_image_render_pre.
Nak-Gyeong Kim [Tue, 8 Sep 2015 23:08:45 +0000 (01:08 +0200)]
evas: fix memory leak in evas_object_image_render_pre.

Summary:
Eina_Rectangles are removed from pixel_updates list without eina_rectangle_free during render pre stage.

@fix

Test Plan: check memory leak.

Reviewers: cedric, jypark

Subscribers: jypark, cedric

Differential Revision: https://phab.enlightenment.org/D3006

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
8 years agoevas: CID 1323091
perepelits.m [Tue, 8 Sep 2015 22:07:45 +0000 (00:07 +0200)]
evas: CID 1323091

Summary:
@fix CID 1323091 for Coverity

Reviewers: cedric, Hermet, raster

Subscribers: cedric, artem.popov

Differential Revision: https://phab.enlightenment.org/D3037

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
8 years agoecore x vsync - add in exynos checks and move kernel v checks inside
Carsten Haitzler (Rasterman) [Tue, 8 Sep 2015 13:00:33 +0000 (22:00 +0900)]
ecore x vsync - add in exynos checks and move kernel v checks inside

this adds asmung exynos driver as a whiteliested drm vsync driver.
this means 3.0 needs to work so move kernel ver checks inot each driver.

@feature

8 years agoEdje_cc: check value for text.source attribute.
Mykyta Biliavskyi [Tue, 8 Sep 2015 09:34:16 +0000 (18:34 +0900)]
Edje_cc: check value for text.source attribute.

Summary:
Add additional check for description.text.source
and description.text.text_source attributes in TEXT part.
For cases when as source uses non TEXT/TEXTBLOCK part
serialization process should be stopped. This will cause
segmentation fault on runtime. (see edje_text.c line 251)

Reviewers: Hermet, raster, cedric, jpeg

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D3026

Signed-off-by: Jean-Philippe Andre <jp.andre@samsung.com>
8 years agoevas/gl_common: Do not set minus value to unsigned int
Dongyeon Kim [Tue, 8 Sep 2015 08:41:15 +0000 (17:41 +0900)]
evas/gl_common: Do not set minus value to unsigned int

Summary:
cur_tex* are unsigned int, so minus value should not be set.
Also, add cur_texa, cur_texm here.

Test Plan: Local tests

Reviewers: jpeg

Subscribers: mer.kim, wonsik, cedric

Differential Revision: https://phab.enlightenment.org/D3036

8 years agoeet_example: Fixing memory leak.
Srivardhan Hebbar [Tue, 8 Sep 2015 08:23:42 +0000 (10:23 +0200)]
eet_example: Fixing memory leak.

Summary:
calloc is done 2times, so there is a memory leak. So fixing that.

Signed-off-by: Srivardhan Hebbar <sri.hebbar@samsung.com>
Reviewers: cedric, stefan_schmidt

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D3032

8 years agoecore_evas_convert: make sure we add all needed flags and deps for the build
Stefan Schmidt [Tue, 8 Sep 2015 08:16:21 +0000 (10:16 +0200)]
ecore_evas_convert: make sure we add all needed flags and deps for the build

We got a report where ecore_evas_convert linked against some old libs
installed in the system instead of the recent ones in-tree. While I was not
able to reproduce this the flags and deps for ecore_evas_convert have been
missing in any case and could be the culprit.

Ref T2716

8 years agoecore_evas_extn: Fix server rendering after restart
Shinwoo Kim [Tue, 8 Sep 2015 04:41:23 +0000 (13:41 +0900)]
ecore_evas_extn: Fix server rendering after restart

Summary:
The server can render, only after the server get the OP_SHOW from the client.
However, if the server relaunches while client is running, the server cannot get the OP_SHOW.
In this case, the client should send the OP_SHOW, when the server is added.

Test Plan: Relaunch a server especially the indicator, while client is running.

Reviewers: raster, cedric, Hermet, woohyun, jaehwan, jypark, jpeg

Subscribers: cedric, seoz

Differential Revision: https://phab.enlightenment.org/D3028

Signed-off-by: Jean-Philippe Andre <jp.andre@samsung.com>
8 years agoevas examples: avoid shadowing a global variable in a function
Stefan Schmidt [Mon, 7 Sep 2015 13:52:35 +0000 (15:52 +0200)]
evas examples: avoid shadowing a global variable in a function

8 years agoevas: gl_generic: fix leak if we return dst and not newdst
Stefan Schmidt [Mon, 7 Sep 2015 11:41:47 +0000 (13:41 +0200)]
evas: gl_generic: fix leak if we return dst and not newdst

In case we still retrun dst we should make sure we free up the newly
allocated newdst.

CID: 1323090

8 years agoecore_example: Fixing NULL deference.
Srivardhan Hebbar [Mon, 7 Sep 2015 11:30:48 +0000 (13:30 +0200)]
ecore_example: Fixing NULL deference.

Summary:
pd was dereferenced and then checked for NULL. If it was NULL, the program would crash much before. So fixed that.

Signed-off-by: Srivardhan Hebbar <sri.hebbar@samsung.com>
Reviewers: cedric, stefan_schmidt

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D3031

8 years agoecore_example: Fixing memory leak.
Srivardhan Hebbar [Mon, 7 Sep 2015 09:36:18 +0000 (11:36 +0200)]
ecore_example: Fixing memory leak.

Summary:
Fixed a memory leak in this example.

Signed-off-by: Srivardhan Hebbar <sri.hebbar@samsung.com>
Reviewers: cedric, stefan_schmidt

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D3010

8 years agoevil: remove executable bit from memcpy assembler file.
Stefan Schmidt [Mon, 7 Sep 2015 09:23:53 +0000 (11:23 +0200)]
evil: remove executable bit from memcpy assembler file.

Hopefully the last offender we have in tree with this.

8 years agoecore_example: Fixing NULL dereference.
Srivardhan Hebbar [Mon, 7 Sep 2015 09:13:48 +0000 (11:13 +0200)]
ecore_example: Fixing NULL dereference.

Summary:
cd was dereferenced and then was checked for NULL. If cd is NULL, it would crash before itself. So checking for NULL before dereferencing cd.

Signed-off-by: Srivardhan Hebbar <sri.hebbar@samsung.com>
Reviewers: cedric, stefan_schmidt

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D3009

8 years agoecore_example: Cleanup of ecore_compose_get_example.c
Srivardhan Hebbar [Mon, 7 Sep 2015 09:00:20 +0000 (11:00 +0200)]
ecore_example: Cleanup of ecore_compose_get_example.c

Summary:
Relating to D3010. This patch is code cleanup. Have done dos2unix. After this will update D3010 bug.

Signed-off-by: Srivardhan Hebbar <sri.hebbar@samsung.com>
Reviewers: cedric, stefan_schmidt

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D3025

8 years agobuild: add efl as internal dep for eio
Stefan Schmidt [Mon, 7 Sep 2015 08:43:35 +0000 (10:43 +0200)]
build: add efl as internal dep for eio

Thanks to Adrien Nader for the report and fix.

8 years agoupdate mailmap
Boris Faure [Sat, 5 Sep 2015 11:56:01 +0000 (13:56 +0200)]
update mailmap

8 years agoedje_cc: remove executable bit from source file
Stefan Schmidt [Fri, 4 Sep 2015 12:19:39 +0000 (14:19 +0200)]
edje_cc: remove executable bit from source file

8 years agoecore: win32: remove executable bit from source file
Stefan Schmidt [Fri, 4 Sep 2015 12:19:08 +0000 (14:19 +0200)]
ecore: win32: remove executable bit from source file

8 years agoector: software: remove executable bit from source files
Stefan Schmidt [Fri, 4 Sep 2015 12:15:57 +0000 (14:15 +0200)]
ector: software: remove executable bit from source files

This really shows what parts are developed under windows. While it does not
matter where you develop please make sure that you do not introduce CRLF
endings and normal source files as executables.

8 years agoector: software: convert dos to unix line endings
Stefan Schmidt [Fri, 4 Sep 2015 12:09:50 +0000 (14:09 +0200)]
ector: software: convert dos to unix line endings

8 years agoecore: cocoa: convert dos to unix line endings
Stefan Schmidt [Fri, 4 Sep 2015 12:07:52 +0000 (14:07 +0200)]
ecore: cocoa: convert dos to unix line endings

8 years agoevas: tests: convert man_mesh.obj from dos to unix line endings
Stefan Schmidt [Fri, 4 Sep 2015 12:05:38 +0000 (14:05 +0200)]
evas: tests: convert man_mesh.obj from dos to unix line endings

8 years agoRemove TODO-eo2 file.
Tom Hacohen [Fri, 4 Sep 2015 11:30:28 +0000 (12:30 +0100)]
Remove TODO-eo2 file.

This should never have gotten in.

8 years agoEvas: Move REND_DBG to evas_render.c
Jean-Philippe Andre [Fri, 4 Sep 2015 06:13:35 +0000 (15:13 +0900)]
Evas: Move REND_DBG to evas_render.c

Unlike what the comment claimed, it is not being used anywhere outside
this file.

8 years agoevas: Use malloc instead of calloc in context_dup
Jean-Philippe Andre [Fri, 4 Sep 2015 06:02:48 +0000 (15:02 +0900)]
evas: Use malloc instead of calloc in context_dup

Since we call memcpy right after, no need for calloc

8 years agoevas_render: Fix some clipping issues inside proxy
Jean-Philippe Andre [Thu, 3 Sep 2015 08:38:48 +0000 (17:38 +0900)]
evas_render: Fix some clipping issues inside proxy

Inside a proxy, clipping information might be wrong since the
source object may be at a different position than within
the proxy. If source_clip is not set, then we need to discard
all clips that are outside the proxy context.

So we just propagate the clip information inside the current
draw context, and even recurse from clipper to clipper to
find the final state of clipping.

Map and proxies and others (who said masks?) should definitely
rely more on the same model.

This code is not a mess. At all. You gotta love evas_render.

8 years agoevas_render: print out context clip with REND_DBG
Jean-Philippe Andre [Thu, 3 Sep 2015 07:36:32 +0000 (16:36 +0900)]
evas_render: print out context clip with REND_DBG

8 years agoevas_render: simplify masking and clipping in general
Jean-Philippe Andre [Thu, 3 Sep 2015 05:06:36 +0000 (14:06 +0900)]
evas_render: simplify masking and clipping in general

Use context_dup to inherit from previous contexts in a clean
manner. This removes the need for restoring the previous
clip information.

Plus, this commit removes lines of code so it must be good, right?

8 years agoevas_render: Cosmetic changes in evas_render_mapped
Jean-Philippe Andre [Thu, 3 Sep 2015 03:11:52 +0000 (12:11 +0900)]
evas_render: Cosmetic changes in evas_render_mapped

Use ENFN, ENDT macros to simplify code.
I believe it is more concise and readable like this.

8 years agoevas: Add internal context_dup function
Jean-Philippe Andre [Thu, 3 Sep 2015 03:03:23 +0000 (12:03 +0900)]
evas: Add internal context_dup function

This will simplify some code related to clipping and masking.

8 years agoevas_render: moar debug (REND_DBG)
Jean-Philippe Andre [Tue, 1 Sep 2015 11:47:40 +0000 (20:47 +0900)]
evas_render: moar debug (REND_DBG)

8 years agoEvas masking: Fix potential invalid access to mask image
Jean-Philippe Andre [Mon, 31 Aug 2015 08:28:58 +0000 (17:28 +0900)]
Evas masking: Fix potential invalid access to mask image

After clip_image_get, the old mask may be replaced by a new one,
and unref'ed, but it is later on set back as the context mask image.
Maybe it's possible that there was 0 reference and the image
got freed in between.

No idea how to test this.

@fix

8 years agoEvas masking: Make sure to check alloc before freeing old image
Jean-Philippe Andre [Mon, 31 Aug 2015 07:46:19 +0000 (16:46 +0900)]
Evas masking: Make sure to check alloc before freeing old image

Okay, I'm being paranoid and this can't possibly happen (calloc
fail? yeaaaah) but this makes sure we don't return NULL after
freeing the original image.

8 years agoEvas: Remove useless include
Jean-Philippe Andre [Thu, 27 Aug 2015 08:14:03 +0000 (17:14 +0900)]
Evas: Remove useless include

config.h is already included in the private headers, and should be
included first anyways.

8 years agoeolian: remove support for old style docs completely
Daniel Kolesa [Thu, 3 Sep 2015 14:35:53 +0000 (15:35 +0100)]
eolian: remove support for old style docs completely

8 years agoeolian: disable parsing of old style docs
Daniel Kolesa [Thu, 3 Sep 2015 14:08:08 +0000 (15:08 +0100)]
eolian: disable parsing of old style docs

8 years agoevas canvas: finish doc conversion
Daniel Kolesa [Thu, 3 Sep 2015 14:00:06 +0000 (15:00 +0100)]
evas canvas: finish doc conversion

8 years agoeio: unify deletion of eina_file_ls's iterator.
Guillaume Friloux [Wed, 2 Sep 2015 20:10:40 +0000 (22:10 +0200)]
eio: unify deletion of eina_file_ls's iterator.

Due to a bug with eio_file_stat_ls(), a previous patch
(73a23401ef007cd80297df6546484f3e8a065d2c) was written.

This patch change the way we delete returned eina iterator
from eina_file_ls in _eio_file_eina_ls_heavy() from a separate
thread, into eio_async_end() from the main thread.

To clarify the code, and avoid a useless call to
eina_iterator_free(), this patch also moves the deletion
of the same eina iterator from _eio_file_heavy() into
eio_async_end() as well.

8 years agoecore_psl1ght: Removed unused variable.
Srivardhan Hebbar [Thu, 3 Sep 2015 08:19:06 +0000 (10:19 +0200)]
ecore_psl1ght: Removed unused variable.

Summary: Signed-off-by: Srivardhan Hebbar <sri.hebbar@samsung.com>

Reviewers: cedric, stefan_schmidt

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D3011

8 years agoecore-con-url- split out the curl dleopener and resolver and headers
Carsten Haitzler (Rasterman) [Thu, 3 Sep 2015 03:22:01 +0000 (12:22 +0900)]
ecore-con-url- split out the curl dleopener and resolver and headers

this splits out out "manual" dlopen (eina_module_load) of curl into
its own .c file and special header out of ecore_con_url.c to tidy up
that code a bit and isolate our curl magic loading/handling

8 years agoevas canvas: some doc conversions
Daniel Kolesa [Wed, 2 Sep 2015 15:21:43 +0000 (16:21 +0100)]
evas canvas: some doc conversions

8 years agoevas_canvas: prepare for doc conversion + partial conversion
Daniel Kolesa [Wed, 2 Sep 2015 14:13:36 +0000 (15:13 +0100)]
evas_canvas: prepare for doc conversion + partial conversion

8 years agoefl_vg_base: fix wrong @since (was pushed after 1.15)
Daniel Kolesa [Wed, 2 Sep 2015 13:16:25 +0000 (14:16 +0100)]
efl_vg_base: fix wrong @since (was pushed after 1.15)

8 years agoefl_vg_base: convert docs
Daniel Kolesa [Wed, 2 Sep 2015 13:13:15 +0000 (14:13 +0100)]
efl_vg_base: convert docs

8 years agoevas object: finish doc conversion
Daniel Kolesa [Wed, 2 Sep 2015 13:09:04 +0000 (14:09 +0100)]
evas object: finish doc conversion

8 years agoevas object: more doc conversion
Daniel Kolesa [Wed, 2 Sep 2015 11:51:15 +0000 (12:51 +0100)]
evas object: more doc conversion

8 years agoevas object: partial doc conversion
Daniel Kolesa [Wed, 2 Sep 2015 11:21:59 +0000 (12:21 +0100)]
evas object: partial doc conversion

8 years agoevas - fix texture context bind reset in gl common using wrong texture
Subhransu Mohanty [Wed, 2 Sep 2015 09:55:43 +0000 (18:55 +0900)]
evas - fix texture context bind reset in gl common using wrong texture

evas gl common was simply resetting to the wrong texture id in the gl
context struct - it was using pipe[0] not state.current. why i don't
know. i know i wrote the pipe[0] code many years ago - really don';t
know. it may have been a transitional piece of code that just happened
t6o work 99% of the time that never got fixe when i added pipes.

@fix

8 years agoevas:textgrid: correctly check malloc failures
Boris Faure [Fri, 20 Mar 2015 15:14:50 +0000 (16:14 +0100)]
evas:textgrid: correctly check malloc failures

8 years agoevas:textgrid: rename text_props to text_props_index
Boris Faure [Fri, 20 Mar 2015 14:48:08 +0000 (15:48 +0100)]
evas:textgrid: rename text_props to text_props_index

8 years agoevas clip: add warning msg for wrong scenario.
ChunEon Park [Tue, 1 Sep 2015 07:32:46 +0000 (16:32 +0900)]
evas clip: add warning msg for wrong scenario.

8 years ago[Evas] Counting of tangent space (fix)
perepelits.m [Mon, 31 Aug 2015 10:04:26 +0000 (19:04 +0900)]
[Evas] Counting of tangent space (fix)

Summary: It is counting here, so we can avoid using glsl functions dFdx, dFdy.

Reviewers: cedric, raster, Hermet

Subscribers: Oleksander, cedric, artem.popov

Differential Revision: https://phab.enlightenment.org/D2923

8 years agoedje: remove un-used edje_container code
Stefan Schmidt [Mon, 31 Aug 2015 07:25:30 +0000 (09:25 +0200)]
edje: remove un-used edje_container code

The code have been ifdef'ed out for a along time. I checked back until
the merge of our libs before the 1.8 release. Maybe even before that.

If you have a use case for this code fell free to revert but also enable
the code and make sure it works when doing so.

8 years agoeina tests: disabling eina_test_matrix_quaternion which fails for over a week
Stefan Schmidt [Mon, 31 Aug 2015 06:58:19 +0000 (08:58 +0200)]
eina tests: disabling eina_test_matrix_quaternion which fails for over a week

This fails on nightly_efl_gcc_x86 (32bit) for over a week now blocking all other
nightly builds and tests. Inside the test case there is already something
disabled which meantions problems with rounding. MAybe the same problem here.

Reported as T2701 to keep track of it an enable again once fixed.

ref T2701

8 years agoevas - gl common - make yuv gl textures double-bufffered like yuy2
Carsten Haitzler (Rasterman) [Mon, 31 Aug 2015 01:25:02 +0000 (10:25 +0900)]
evas - gl common - make yuv gl textures double-bufffered like yuy2

this pulls yuv in line with yuy2 textures and dobule buffers them.
this is a workaround some driver bugs where either the driver doesnt
block and wait for the gpu to finish with a texture when updating OR
it does wait, and when it does it blocks and spins using cpu.

8 years agoecore win32: increase command buffer size.
ChunEon Park [Sat, 29 Aug 2015 07:59:19 +0000 (16:59 +0900)]
ecore win32: increase command buffer size.

PATH_MAX is not enough for command.
Even in this case, the buffer is not being used only for path.

8 years agoecore win32: modifiy to better meaningful exception.
ChunEon Park [Sat, 29 Aug 2015 07:46:22 +0000 (16:46 +0900)]
ecore win32: modifiy to better meaningful exception.

8 years agoeina: cpu - fix a typo in eina_cpu_count API doc
Amitesh Singh [Sat, 29 Aug 2015 02:31:10 +0000 (08:01 +0530)]
eina: cpu - fix a typo in eina_cpu_count API doc

8 years agoemile jpeg: more descriptive comment.
ChunEon Park [Fri, 28 Aug 2015 15:59:06 +0000 (00:59 +0900)]
emile jpeg: more descriptive comment.

8 years agoemile: fix can not parsing the EXIF info of the jpeg file which has IFD offset
jiin.moon [Fri, 28 Aug 2015 15:47:43 +0000 (00:47 +0900)]
emile: fix can not parsing the EXIF info of the jpeg file which has IFD offset

Summary:
IFD offset of jpeg is not fixed.
But emile support only 0x8 on now

Reviewers: jypark, cedric, Hermet

Reviewed By: Hermet

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D3000

8 years agoevas_canvas3d: Recalculation coordinates for pick object from scene
Oleksandr Shcherbina [Fri, 28 Aug 2015 11:32:45 +0000 (20:32 +0900)]
evas_canvas3d: Recalculation coordinates for pick object from scene

Summary:
Change recalculation coordinate for pick to depend on size of viewport, not from
scene data.
Due to not correct working pick in case resize view window.
@fix

Reviewers: cedric, Hermet

Reviewed By: Hermet

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D2950

8 years agoEina rectangle: Check for valid rectangle before taking intersection
Subodh Kumar [Fri, 28 Aug 2015 07:27:21 +0000 (16:27 +0900)]
Eina rectangle: Check for valid rectangle before taking intersection

Summary:
Check for valid rectangle

If any dst or src rectangle has zero width or height,
intersection should not return true.

@fix

Test Plan: Added test cases

Reviewers: cedric, herdsman, Hermet

Reviewed By: Hermet

Subscribers: shilpasingh, cedric

Differential Revision: https://phab.enlightenment.org/D2990

8 years agoeet - improve eet file storage when encoding by aligning data
Carsten Haitzler (Rasterman) [Fri, 28 Aug 2015 03:56:18 +0000 (12:56 +0900)]
eet - improve eet file storage when encoding by aligning data

this forces data when written to be aligned to 8 bytes. this makes
mapping keys directly far better as from now on you can assume all
data is aligned to 8 bytes allowing you, if you want, to use eet files
as caches for raw binary data you mmap directly and just use (not
portable though).

@feat

8 years agoEvas textblock: fix case of own_closer in style_set
Daniel Hirt [Thu, 27 Aug 2015 15:20:47 +0000 (18:20 +0300)]
Evas textblock: fix case of own_closer in style_set

Looks like it was assumed that an fnode->orig_format always ends with a
'/' character if the fnode is an own_closer.
The problem is that a paragraph separator ("ps" and "br" - the latter in
legacy newline mode) is also an own_closer, but might not have '/' at the
end, so decrementing the length is wrong.

This fixes T2654. The example markup had "br" read as "b", which led to
a mismatch with the "font_weight=Bold" tag. Coincidentally, "ps" was not
affected as there was no matching "p" in the style.

@fix

8 years agoevas/gl_common: add new header files to build to make them available for dist
Stefan Schmidt [Thu, 27 Aug 2015 14:42:18 +0000 (16:42 +0200)]
evas/gl_common: add new header files to build to make them available for dist

8 years agoecore_wayland: only add listener if recovery interface was binded correctly
Stefan Schmidt [Thu, 27 Aug 2015 14:19:14 +0000 (16:19 +0200)]
ecore_wayland: only add listener if recovery interface was binded correctly

In the case the bidn fails we could crash here by accessing it.

8 years agoedje: Fix double free scenario caused by static pointer.
Youngbok Shin [Thu, 27 Aug 2015 10:04:57 +0000 (11:04 +0100)]
edje: Fix double free scenario caused by static pointer.

Summary:
The result of evas_object_textblock_cursor_content_get() API has to be cleaned
by outside.  _edje_entry_cursor_content_get() is calling free() inside of the
function for handle the result using static pointer. But, the caller of
_edje_entry_cursor_content_get() is already handling the result using free().
It can cause double free problem.

The bigger issue is in elementary. See elm_entry_cursor_content_get() API's
document. The document advice developers to free the result when it is done.

@fix

Test Plan: N/A

Reviewers: tasn, raster, woohyun

Subscribers: cedric

Differential Revision: https://phab.enlightenment.org/D2991

8 years agoecore_pipe: add write fdset and except fdset.
Ji-Youn Park [Thu, 27 Aug 2015 07:58:10 +0000 (16:28 +0830)]
ecore_pipe: add write fdset and except fdset.

ecore_pipe_wait can be run not only select but also _ecore_glib_select__locked.
if write fdset is null, it make problem in FD_ISSET.

@fix

8 years agoecore_wayland: add wayland session recovery listener
Stefan Schmidt [Wed, 26 Aug 2015 12:08:25 +0000 (14:08 +0200)]
ecore_wayland: add wayland session recovery listener

Register and wait for the uuid event issued by the compositor once a new
surface is created.

8 years agoecore_wayland: bind and destroy session_recovery interface correctly.
Stefan Schmidt [Tue, 25 Aug 2015 12:31:09 +0000 (14:31 +0200)]
ecore_wayland: bind and destroy session_recovery interface correctly.

This was missing from the initial session recovery support patches. Bind
the interface so we can actually work with it on the client side and destroy
it at the end.

@fix

8 years agoEo base: print an error when an object with a parent is destructed.
Tom Hacohen [Wed, 26 Aug 2015 13:17:26 +0000 (14:17 +0100)]
Eo base: print an error when an object with a parent is destructed.

This should not happen. Objects with parents must have their parents
unset before they reach refcount == 0. That's because the parent is the
one holding the refcount. This means that if we get to the destructor
(object is deleted) while a parent is still set, we have an error
scenario.

8 years agoRevert "Revert "Eo base: Change parent_set to be an assignment of ref.""
Tom Hacohen [Wed, 26 Aug 2015 09:47:06 +0000 (10:47 +0100)]
Revert "Revert "Eo base: Change parent_set to be an assignment of ref.""

Had to revert it until I pushed the changes to elm, which I had issues
with. Now the patch can safely go back in.

This reverts commit 37abea3831f62ac612b377bb48ce6c90e076dd25.

8 years agoRevert "Eo base: Change parent_set to be an assignment of ref."
Tom Hacohen [Wed, 26 Aug 2015 09:32:32 +0000 (10:32 +0100)]
Revert "Eo base: Change parent_set to be an assignment of ref."

Damn, this breaks some things. Reverting until fixed.

This reverts commit 9c78ee0bf4125c095e85f7fcf9921586cda64a52.

8 years agoEo base: Change parent_set to be an assignment of ref.
Tom Hacohen [Wed, 26 Aug 2015 09:02:01 +0000 (10:02 +0100)]
Eo base: Change parent_set to be an assignment of ref.

After this change, parent_set assigns a ref, so for example:
  obj = eo_add(CLASS, parent); /* Ref is 1 */
  eo_do(obj, eo_parent_set(parent2)); /* Ref is 1 */
  eo_ref(obj); /* Ref is 2 */
  eo_do(obj, eo_parent_set(NULL)); /* Ref is 1, giving the ref to NULL */
  eo_do(obj, eo_parent_set(parent)); /* Ref is 1 */

This is following a discussion on the ML about commit
8689d54471aafdd7a5b5a27ce116bf2ab68c1042.

@feature

8 years agoedje_cc_out: Fix edje_cc compile error with script on Windows.
Jaehyun Cho [Wed, 26 Aug 2015 04:30:15 +0000 (13:30 +0900)]
edje_cc_out: Fix edje_cc compile error with script on Windows.

On Windows, PATH_MAX is 260 and PATH_MAX is used as string buffer
size in edje_cc compile. This causes edje_cc compile error when the edc
file contains "script" keyword and the length of file paths is
relatively long.
To resolve this problem, change the string buffer size in edje_cc
compile.

@fix

8 years agoRevert "Revert "ecore_win: fix the incorrect mouse cursor position.""
ChunEon Park [Tue, 25 Aug 2015 10:32:29 +0000 (19:32 +0900)]
Revert "Revert "ecore_win: fix the incorrect mouse cursor position.""

This reverts commit 3ce8860dab29ba6d0849ed1ecbbc72bbe0e11773.

Apply only to mouse wheel case. Button press/release wans't problem actually.
If I correct, this is caused because of different nature of window systems.
Anyway our Ecore_Event_Mouse values should keep consistency among the various systems.

8 years agoRevert "ecore_win: fix the incorrect mouse cursor position."
ChunEon Park [Tue, 25 Aug 2015 09:12:40 +0000 (18:12 +0900)]
Revert "ecore_win: fix the incorrect mouse cursor position."

This reverts commit 1f1542f82404fa2262526cc3eed7b2edc622c5ce.

coordinate has been corrupted. need to investigate more.

8 years agobuild: replace libsystemd-{daemon, journal} with libsystemd
Marcel Hollerbach [Tue, 25 Aug 2015 08:52:47 +0000 (10:52 +0200)]
build: replace libsystemd-{daemon, journal} with libsystemd

Summary:
since all the libs got merged into libsystemd in 209, we can just check
for libsystemd

Reviewers: cedric

Subscribers: stefan_schmidt, morlenxus

Differential Revision: https://phab.enlightenment.org/D2984

8 years agoecore_win: fix the incorrect mouse cursor position.
ChunEon Park [Tue, 25 Aug 2015 06:31:05 +0000 (15:31 +0900)]
ecore_win: fix the incorrect mouse cursor position.

Ecore_Event_Mouse_* x, y values are relative to the current window position
as well as the root x, y, values are relative to the root window.

previously, x,y is started from the root window and root x, y values are invalid.

fix them

@fix

8 years agosmall typo in docs, mounse is mouse
Thanatermesis [Tue, 25 Aug 2015 01:04:50 +0000 (03:04 +0200)]
small typo in docs, mounse is mouse

8 years agoeldbus: add since tag for now public eldbus_message_signal_new API
Stefan Schmidt [Mon, 24 Aug 2015 10:07:10 +0000 (12:07 +0200)]
eldbus: add since tag for now public eldbus_message_signal_new API

Thanks go to vtorri for bringing this to my attention.

8 years agoevas_3d: fix defines
Marcel Hollerbach [Mon, 24 Aug 2015 10:00:57 +0000 (12:00 +0200)]
evas_3d: fix defines

Summary:
evas_3d: removed unnessecary defines

Evas_Real was allready defined.

The typedefs of the Eo types can be avoided by fixing the include order

Reviewers: cedric, stefan_schmidt, tasn

Reviewed By: stefan_schmidt, tasn

Subscribers: stefan_schmidt, cedric

Projects: #efl

Maniphest Tasks: T2658

Differential Revision: https://phab.enlightenment.org/D2974

8 years agoevas/filter: Remove unused variable
Stefan Schmidt [Mon, 24 Aug 2015 08:15:11 +0000 (10:15 +0200)]
evas/filter: Remove unused variable

lib/evas/filters/evas_filter_parser.c:861:14: warning: unused variable 'ok' [-Wunused-variable]

8 years agoeldbus: remove now public eldbus_message_signal_new declaration from private.h
Stefan Schmidt [Mon, 24 Aug 2015 08:11:34 +0000 (10:11 +0200)]
eldbus: remove now public eldbus_message_signal_new declaration from private.h

The declaration is now on eldbus_message as EAPI so we can safely remove it here.

8 years agoeldbus: add eldbus_message_signal_new to public header
Lukasz Stanislawski [Mon, 24 Aug 2015 08:07:40 +0000 (10:07 +0200)]
eldbus: add eldbus_message_signal_new to public header

Test Plan: elementary tests

Reviewers: cedric

Subscribers: seoz, cedric

Differential Revision: https://phab.enlightenment.org/D2772

8 years agoefl - jp2k loader - guard against openjpeg bug tha causes an abort
Carsten Haitzler (Rasterman) [Mon, 24 Aug 2015 02:39:12 +0000 (11:39 +0900)]
efl - jp2k loader - guard against openjpeg bug tha causes an abort

so... if you load a non-jp2k file using openjpeg, you can get an abort
deep inside the openjpeg library that we can't do anything about. we
set all error handlers but literally the openjpeg code has ab assert
there that causes this bug. it shouldn't and newer opengjpeg libs have
it removed, but 1.5.2 has it and this causes an untrappable crash.
this is simply bad behavior in openjpeg not allowing it to be used
safely to loade image files. the relevant backtrace:

    w=w@entry=0x7fffffffb548, h=h@entry=0x7fffffffb54c,
    alpha=alpha@entry=0x7fffffffb556 "", map=map@entry=0x7fff29ac2000,
    length=<optimized out>, error=error@entry=0x7fffffffb5bc,
    opts=<optimized out>)
    at modules/evas/image_loaders/jp2k/evas_image_load_jp2k.c:111

the relevant code in openjpeg:

int cio_numbytesleft(opj_cio_t *cio) {
  assert((cio->end - cio->bp) >= 0);
        return cio->end - cio->bp;
}

so that assert is triggered. and nothing can be done about it which is
pretty poor.

so an upgrade of openjpeg should fix this as in newer versions have
dropped the assert line in that function, but until poeople have that from
their distro, this adds magic number checks for file headers that avoids
using openjpeg if it's not "apparently" a jp2k file. this does not
stop a corrupt file or a maliciously designed file still causing this
problem, but it does just result in an abort() and isnn't seemingly an
overflow isse that can be exploted, so if you still suffer, find a way to
upgrade openjpeg to 2.x. until then... this reduces inadvertent damage.

@fix

8 years agoeina: fix coverity issue CID 1317154.
Cedric BAIL [Sat, 22 Aug 2015 17:50:58 +0000 (19:50 +0200)]
eina: fix coverity issue CID 1317154.

8 years agoethum: fix compilation on Windows due to missing visibility setting.
ChunEon Park [Sat, 22 Aug 2015 07:09:25 +0000 (16:09 +0900)]
ethum: fix compilation on Windows due to missing visibility setting.

8 years agoevil: add missing since
ChunEon Park [Sat, 22 Aug 2015 07:01:45 +0000 (16:01 +0900)]
evil: add missing since

8 years agoEo base: clean up parent removal in destructor.
Tom Hacohen [Fri, 21 Aug 2015 16:19:47 +0000 (17:19 +0100)]
Eo base: clean up parent removal in destructor.

This cleans up the changes in 8689d54471aafdd7a5b5a27ce116bf2ab68c1042.
This commit reduces code duplication, and fixes broken handling of
overridden parent_set().

8 years agoecore_con: fix test to properly send the right amount of data (including the '\0').
Cedric BAIL [Fri, 21 Aug 2015 15:54:13 +0000 (17:54 +0200)]
ecore_con: fix test to properly send the right amount of data (including the '\0').

8 years agoeina: use new eina_file_path_join API
Vincent Torri [Thu, 13 Aug 2015 06:14:32 +0000 (08:14 +0200)]
eina: use new eina_file_path_join API

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
8 years agoeina: add API to join pahs, as well as path separators
Vincent Torri [Thu, 13 Aug 2015 06:11:15 +0000 (08:11 +0200)]
eina: add API to join pahs, as well as path separators

eina_str_join() is used a lot to contatenate paths, but the
separator should be '\' on Windows. So add 2 API and 2 defines for
more cross platform code

@feature

Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
8 years agoector: don't forget to reset cairo matrix to identity.
Cedric BAIL [Fri, 21 Aug 2015 14:07:16 +0000 (16:07 +0200)]
ector: don't forget to reset cairo matrix to identity.

8 years agoefl: fix Efl.Gfx.Shape.Dup to use const parameter.
Cedric BAIL [Tue, 28 Jul 2015 23:17:53 +0000 (01:17 +0200)]
efl: fix Efl.Gfx.Shape.Dup to use const parameter.