Summary:
Prerequisite: Partial rendering ON + Image Prealoding + Triple surface buffer of GL.
Previously, evas trys to draw of an image which didn't prepare of image data yet (in case of preloading)
This time, it will draw a solid color onto the dest sufrace 1,
But luckily, preloading is finished just after, it draws proper image data onto next surface 2 and 3.
Now, triple buffer is filled with the image data but only first frame is still empty. That's a problem.
This patch skips to draw image if it doesn't prepare data yet, but once the preloading is finished,
it starts to draw images.
@fix
Reviewers: #committers
Subscribers: kimcinoo, cedric, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6739
Summary:
Unexpected image shows if image data is not ready.
Even though there is a change to check the 'preloading' in pre_render phase,
evas_object_image_render is called. So we need to check here as well.
Reference: https://phab.enlightenment.org/D6739
It seems that the 'preloading' is not enough. The 'preloading' could be reset
to FALSE by _evas_image_load_async_cancel > _image_preload_internal.
If the following step happens, then this patch set is neccessary.
(1) evas_object_image_pre_render
(2) _evas_iamge_load_async_start
(3) evas_object_image_render
I could not find out what the correct step, but it actullay happens.
The evas_object_image_render could be called with the 'preloading' TURE.
Reviewers: Hermet
Reviewed By: Hermet
Subscribers: cedric, #reviewers, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6778
Summary:
If a smart class overrides Evas_Smart_Class.move as below,
then original behavior must not be used for the smart class.
Evas_Smart_Class sc = EVAS_SMART_CLASS_INIT_NAME_VERSION("MyClass");
evas_object_smart_clipped_smart_set(&sc);
sc.move = &myMove;
But current implementation makes original behavior work.
So before using the original method, this patch is checking if the original
method is changed or not.
Reviewers: zmike, devilhorns
Reviewed By: zmike
Subscribers: woohyun, jypark, cedric, raster, jpeg, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6468
Summary:
mouse coordinate is set by seat.
Therefore, mouse_in state should be set based on seat.
As a result, mouse_in state of all pointers in the seat should be
updated at the same time.
Reviewers: eagleeye, devilhorns, zmike
Reviewed By: eagleeye, devilhorns, zmike
Subscribers: devilhorns, Hermet, kimcinoo, cedric, iscaro, zmike, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6699
Summary:
To take screenshots, Enlightenment makes a new snapshot object, performs
a manual render, and uses the snapshot results.
Turns out if this happens while an async render is in progress, the
async render's completion triggers a render post callback on the snapshot
object even though it's never been involved in a render.
We need to defer new render post callbacks until any currently running
render completes, then add them during that render's post.
Fix T7156
Reviewers: devilhorns, zmike
Reviewed By: devilhorns, zmike
Subscribers: devilhorns, cedric, #committers, zmike
Tags: #efl
Maniphest Tasks: T7156
Differential Revision: https://phab.enlightenment.org/D6711
Summary:
When testing partial rendering on TV, found this bug.
Probably this was caused because of wrong copy/paste.
When region1 completely subsumes region2, region2 should be ignored.
But currently region2 is copied to region1 which makes region1 rect completly wrong.
This was causing lot of visual defects because of wrong rects.
This patch will fix those visual defects when partial rendering is enabled.
Test Plan: Tested with the patch on TV after enabling partial rendering.
Reviewers: raster, cedric, Hermet, ManMower
Reviewed By: ManMower
Subscribers: bu5hm4n, ajwillia.ms, ManMower, devilhorns, cedric, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6696
Summary:
Here comparision is logically wrong because it compares
the coordinates(cur, prev) of the different spaces.
As you can see just above the conditions,
Cur coordinates could be transformed into the map space (if objs have)
on the otherhands, prev is not. These wrong condition could occur
embarassing event trigger.
@fix
Reviewers: #committers, zmike
Reviewed By: #committers, zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6674
Summary:
there is no point to feeding mouse events
if clipped object is invisible because
previsou/current situation won't be different.
fyi, move events won't be triggered if prev/cur coordinates
are same.
Reviewers: #committers, zmike
Reviewed By: #committers, zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6645
A null font set can potentially be stored and passed on. There is no
specific fallback in our codebase to roll-back if something like that
occurs.
This fixes a case where an entry of a null font was not removed from the
fonts_cache and was leaked to subsequent tests (the null font entry was
not removed from the fonts_cache).
Note that this was only apparent when EFL was built with
`--disable-fontconfig`, likely due to fontconfig succeeding to always
retrieve some non-null font.
@fix
Summary:
Basically, render_pre/post should be called only if rendering happens.
Unfortunately, this behavior had been broken by some mistakes.
As far as I understand after reading history carefully,
this issue introduced for fixing unpair of pre/post cbs.
In case of async render, the post callback would be triggered in
evas_render_pipe_wakeup(), it means POST callbake shouldn't be tiggered in
evas_render_updates_internal(). If that post callback were removed in
evas_render_updates_internal(), then in case of sync_render,
POST callback won't be triggered. So previously, the PRE/POST couldn't get
paired.
I guess cedric put a mistake that nothing2rendering which brought
the pre/post pair issue, even he said that patch fix the unpair of them.
(But actually that patch brought the issue!)
See this: dc841ed9b2
Whatever it was intented or not, that patch totally insane, even
non-descripting.
If we just trigger post cb only if sync render or no_rendered case,
We could make satisfication for both cases.
@fix
Reviewers: #committers, devilhorns, zmike
Reviewed By: #committers, zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6573
Summary:
an evas may or may not have a parent; this is legacy api and it's all
confusing
Reviewers: bu5hm4n, devilhorns
Reviewed By: bu5hm4n
Subscribers: cedric, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6486
Summary:
this should only happen if the user has made a mistake regarding the
existence or type of an object, so ensure that an error message occurs to
help debug any failures which result
fix T6326
Reviewers: bu5hm4n, Hermet, woohyun, devilhorns
Reviewed By: Hermet
Subscribers: cedric, #committers
Tags: #efl
Maniphest Tasks: T6326
Differential Revision: https://phab.enlightenment.org/D6322
Summary:
It was only happened when legacy newline is enabled. By default,
legacy newline is enabled. As I know, legacy newline option has
to change textblock's internal behavior. But, it shouldn't change
the given original text. It fixes T3399.
Test Plan: A Test case is included in Evas test suite.
Reviewers: Jaehyun_Cho, z-wony, tasn, woohyun, herdsman, Blackmole, devilhorns
Subscribers: #committers, zmike, raster, cedric, jpeg
Tags: #efl
Maniphest Tasks: T3399
Differential Revision: https://phab.enlightenment.org/D3874
The ellipsis item was tracked so it can be inserted at the correct
position. It was not cleared properly, leading to a randomly failing
test suite.
Actually, it seems that there is no need to even store the ellipsis item
as the object data, so it was also moved to be in the layout context,
instead.
A special shoutout to ManMower for lending me his machine, as the test
suite didn't fail on mine.
Fixes T6986
@fix
Summary:
fallbacks string also has to be compared to load proper fdesc.
If a font which does not have fallback fonts is loaded,
fallback fonts can't be appended to the same font.
@fix
Reviewers: tasn, woohyun, herdsman, zmike, devilhorns
Reviewed By: herdsman
Subscribers: #committers, zmike, raster, Blackmole, z-wony, cedric, jpeg
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D3707
Summary:
When multiple fallback fonts was passed to evas_font_load(),
the while loop could run forever. I think it is never tested.
@fix
Test Plan:
Set the following textblock style and set it to a textblock object.
"font=Sans font_fallbacks=Ubuntu,Inconsolata,BlahBlah font_size=..."
Run and see application is in stuck.
Reviewers: tasn, woohyun, herdsman, devilhorns
Subscribers: #committers, zmike, raster, Blackmole, z-wony, cedric, jpeg
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D3703
Summary:
Ok, this was started from a bug that canvas getting not be updated.
If map is just disabled, at least one frame in the map region should be redrawn
So I added a condition 'map changed' in the render even though map is off
status. Now, I got a performance regression issue because it makes dirty
region is always true for the map object.
That is a corner case acutally, that object is not rendered but map still
have changed status.
I replaced the condition only if object is changed + map is changed.
At least, my test case works better with this patch.
@fix T6975
Reviewers: #committers, ManMower, devilhorns
Reviewed By: #committers, ManMower
Subscribers: ApB, ManMower, cedric, #committers, zmike
Tags: #efl
Maniphest Tasks: T6975
Differential Revision: https://phab.enlightenment.org/D6429
Summary:
these separate inits and shutdowns make it impossible to effectively control
ecore's lifetime which makes evas_shutdown unreliable as objects may be
destroyed at any point
ref T7052
Depends on D6475
Reviewers: ManMower, devilhorns
Reviewed By: ManMower, devilhorns
Subscribers: cedric, #committers
Tags: #efl
Maniphest Tasks: T7052
Differential Revision: https://phab.enlightenment.org/D6476
Summary:
ecore_shutdown will trigger object deletions which require common
components to still be active in order to avoid crashes
ref 3433be343779424c5e030ace30e211298cd060f8
ref T7052
Reviewers: ManMower, devilhorns
Reviewed By: ManMower, devilhorns
Subscribers: cedric, #committers
Tags: #efl
Maniphest Tasks: T7052
Differential Revision: https://phab.enlightenment.org/D6475
Summary:
The variable "clean_them" can only ever be EINA_FALSE for much of this
function, but using it as a return value ensures that anyone not
intimately familiar with the code will have to read a lot of code
to figure out that this is so.
Instead, return EINA_FALSE up until the point clean_them can actually
be something else.
No functional change.
Reviewers: devilhorns, zmike
Reviewed By: zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6464
this resolves some invalid read/write operations between threads without locking
and also attempts to improve thread-related behavior after fork() calls
ref T7027
Differential Revision: https://phab.enlightenment.org/D6370
Summary:
parent_container should have eina_false as given null parent
in order to avoid call vg_node_changed() for parent.
Reviewers: #committers, devilhorns, zmike
Reviewed By: #committers, zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6387
Summary: this callback is triggered on the parent object. invalidated objects have no parent.
Reviewers: bu5hm4n, devilhorns, Hermet
Reviewed By: Hermet
Subscribers: Hermet, cedric, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6335
Summary:
the previous patch which improved this code for x86 archs broke compiling
for non-x86 and, coincidentally, for windows builds on x86 due to some
unusual #ifdef blocks
this attempts to restore handling on non-x86 and adds additional #ifdefs for
functions which did not build on windows due to removed code
ref 6b1ab3cd9c
Reviewers: ManMower, devilhorns
Reviewed By: ManMower
Subscribers: cedric, #committers
Tags: #efl
Maniphest Tasks: T7062
Differential Revision: https://phab.enlightenment.org/D6368
Summary:
this requires the parent of the deleted object to be reachable, so use the
invalidate callback as this is the last point during which it's possible to
retrieve a parent object
ref T7030
Depends on D6320
Reviewers: bu5hm4n, Hermet, woohyun, devilhorns
Reviewed By: bu5hm4n
Subscribers: cedric, #committers
Tags: #efl
Maniphest Tasks: T7030
Differential Revision: https://phab.enlightenment.org/D6321
Since usage has been changed under the beta version,
Some users may keep the old-usage that breaks the vg behavior now.
For their information, vg prints messages in case of breaks.
Summary:
To determine if a system supports SIMD instructions, the cpuid facility
should be used. However, for 15+ years EFL has been trapping SIGILL,
then attempting to execute these intstructions.
Continuing after SIGILL is explicitly undefined behaviour and can never
safely be relied upon - it is possible the CPU will respond to the
unknown instruction in an upredictable way and the program will not
continue correctly. Even if it hasn't caused problems before, there's
no reason to believe a processor released in the future won't behave
differently.
Lately we've had a couple of bug tickets where SIGILL appears to cause
problems at a system level as well, but there seems little point in
chasing those problems down as we shouldn't even be doing this in the
first place.
ref T6711
ref T6989
We still rely on SIGILL in a few configurations where eina_cpu doesn't
know how to query features properly (powerpc, sparc, and non linux
ARM configurations). Hopefully someone with expertise on those
platforms can follow up and we can remove this entirely.
Note: MMX2 appears to not really be a thing, and is instead provided by
both 3DNow! and SSE. We already conflate it with SSE in other parts of
evas, so I've just used SSE here to test for its presence.
Depends on D6313
Reviewers: devilhorns, zmike
Reviewed By: zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Maniphest Tasks: T6989, T6711
Differential Revision: https://phab.enlightenment.org/D6314
Summary:
This reverts commit cd6b890c73.
this resolves complaints from a safety check in the debug profile,
but it seems more likely that the safety check is bogus and should be
removed rather than making this code more complex
Reviewers: devilhorns
Reviewed By: devilhorns
Subscribers: cedric, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6283
Summary:
what here was done was fundamentally wrong, deleting the pd->object
field of a evas object after a efl_del / evas_object_del is completly
wrong. evas object lifetimes are controller with eo_manual_free, this
means, they are still alive, even after you called evas_object_del on
them. removing pd->object results in eo_menual_free calls to NULL
objects and leaking the object carrying the private data. Overall,
breaking this pd->object field and unsetting it is a very bad idea, as
its the only way that evas cleansup the object correctly.
This brings down the number of ui related leaked objects on shutdowns to
0. (YEY :))
This also fixes weird error messages on app shutdown.
fixes T6964
Reviewers: devilhorns, zmike
Reviewed By: zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Maniphest Tasks: T6964
Differential Revision: https://phab.enlightenment.org/D6290
tbh, current vg interfaces a little bit bad... here is one scenario to this
stupid case.
efl_parent_set() and evas_object_vg_root_node_set() both do re-parent
job. They could be conflicted if user calls both apis in either way.
efl_parent_set(root_node, NULL); but Vg Object still keeps the root node
which is just a dangling pointer that occurs a crash while rendering.
efl_gfx_path itself took care of efl_gfx_shape data but its heirarchy was
conceptually wrong. Even efl_gfx_shape is mixing the efl_gfx_path...
Damend design...
Some of derived classes of efl_gfx_path (i.e. Vg.Node and Vg.Container) are
none of the Path acutally. They are just mixing Path's interpolation interface.
So, Here patch changes VG.Node to stop calling the super's interpolate method
and Vg.Shape to call both super -gfx_shape and vg_node- interpolate method.
@fix T6996
Summary:
At some point this code stopped being necessary. The image data is already
rotated and width and height match it, so no need to rotate it again
(which produced warnings and incorrect behaviour)
Fixes T5841
Test Plan: This fixed examples/evas_images{2,4,5} which were failing before.
Reviewers: bu5hm4n, zmike, devilhorns
Reviewed By: zmike
Subscribers: cedric, #committers
Tags: #efl
Maniphest Tasks: T5841
Differential Revision: https://phab.enlightenment.org/D6259
Summary:
enabling write on a COW which is already writing is a user error,
which can occur during object invalidate due to reuse of a helper function.
by adding an extra param containing the COW data, this scenario can be avoided
fix T7005
Reviewers: bu5hm4n, segfaultxavi, devilhorns
Subscribers: cedric, #committers
Tags: #efl
Maniphest Tasks: T7005
Differential Revision: https://phab.enlightenment.org/D6273
Summary:
sscanf("%ms", &encoding) allocates memory which must be freed.
It is done in the general case, but not if the image data map fails and the
code jumps to the no_pixels: label.
Reviewers: zmike, bu5hm4n
Reviewed By: zmike
Subscribers: cedric, #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6258
So, here is the first-aid, but the last patch for recovering basic vector
rendering. All vg nodes duplicated by container would be constructed without
a parent, they need to get unref() by container when container is freed.
Here code takes deal with it on the invalidation time.
Cached roots are constructed out of canvas.
Means, they have no parents, generated efl_add_ref()
which requests to call efl_unref() instead of explicit deletion.
This fixes vg to connect node tree properly on copying nodes.
Vector rendering was broken when vg cache tries to reconstruct the scene-graph
tree on copying from the original cached one. Exactly, nodes lost parents on
copying. Here it fixes the lost links of the scene-graph tree.
@fix T6993
This is an alternative patch for 9fcd03952e.
Since canvas map rendering sequence is quite complex,
Not easy to estimiate the logic sequence by all scenarios.
The brings me to fix the code case by case.
This reverts commit b2f85bb9ed.
this is a subsquent patch for the previous revert:
77684dc53a
ps. JackDaniel saved efl world from a guy who broke the system and ran away...
This reverts commit 9fcd03952e.
Just noticed from ManMower, this caused a side-effect that drops
enlightenment performance. I need to check it seriously then
decide again how this patch to be.
Summary:
It has been widely used for quite some time now.
Fixes T6889
Test Plan:
Manually built the evas-vg-simple.c and evas-vg-batman.c examples after
removing the manual define of EFL_BETA_API_SUPPORT and EFL_EO_API_SUPPORT
that they have at the top.
Reviewers: ajwillia.ms, zmike
Reviewed By: zmike
Subscribers: cedric, #committers
Tags: #efl
Maniphest Tasks: T6889
Differential Revision: https://phab.enlightenment.org/D6230
Summary:
the same change as the one for evas table, just for box.
Depends on D6228
Reviewers: cedric, zmike, stefan_schmidt
Reviewed By: zmike
Subscribers: #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6229
Summary:
when focus events have been emitted the smart parent was persistent, now
its not anymore, lets restore this behaviour
Depends on D6227
Reviewers: cedric, zmike, stefan_schmidt
Subscribers: #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6228
Summary:
Before the parent was persistent when the intersept function and events
for clipper changes have been called. Lets get back to this behaviour.
Depends on D6226
Reviewers: cedric, zmike, stefan_schmidt
Reviewed By: zmike
Subscribers: #committers
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6227
Summary:
EFL_EVENT_DEL is wrong here, as the callback uses a parent, which is
gone at the time of EFL_EVENT_DEL.
Reviewers: zmike, cedric, stefan_schmidt
Reviewed By: zmike
Subscribers: cedric, #committers, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6225
This is clearly a workaround to a bad situation, but there is no case
that I can find that solely lead to object being NULL nor is there any
code that should do that, but still it does in some case...
Technically I do not thing it is a correct behavior to force destroy
reference that evas do not hold, but evas_object are deaply tied to
the canvas they are build on and even after invalidate it is hard
to not have function call that would lead to crash. Making the pointer
incorrect thanks to eo indirection seems safer here.
When map of a smart object is turned off, redraw-updation area could be incorrect.
This results in a mirror image problem.
for fixing this, let smart object request update-area in render_pre() properly only in case of map changed.
Thanks to akanad for reviewing together.
@fix
Summary:
All events must have a type now, otherwise bindings don't know how to handle
the event_info field.
Most of the missing event types were actually "void" (no event_info present).
Some struct definitions had to be moved to eo instead of h files, so they
are available to bindings. Some have not, and are marked with FIXME.
Some namespaces have been fixed (like Efl_Event_Cb -> Efl.Event_Cb).
In general, there are hundreds of changed files, but mostly to add a type which
was not present before, so there's no harm done.
Also, A lot of FIXMEs have been added which should be, like, fixed.
For example, some events can send different types of event_info, which is
very inconvenient (and error prone).
Test Plan: make with c# bindings works, make check and make examples work too.
Reviewers: cedric, q66, lauromoura
Subscribers: zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6169
Summary:
This appears to be called from a delete callback that takes place well
after the eo parent relationship is deleted, however
efl_input_device_get_seat() finds the seat by finding the parent. That
will always be NULL during this callback, so we'll leak the data.
Instead, search all seats for the pointer.
Depends on D6181
Reviewers: zmike
Reviewed By: zmike
Subscribers: cedric, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6182
Summary:
This delete callback is shared for mouse, pointer, touch, and seat. We
should only delete pointer data when we're called on a pointer.
Note that (at least under wayland) this particular bug is irreproducible
right now because, due to a ref leak, this deletion callback can never
fire.
Depends on D6180
Reviewers: zmike
Reviewed By: zmike
Subscribers: cedric, zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6181
Summary:
Calling efl_canvas_text_style_set() with empty key means
setting a default style to object. But, it counldn't store style
as default properly. It caused a crash issue from elementary_test.
@fix
Test Plan:
New test case is included. Run test suite. Or,
1. Run elementary_test
2. Find and launch "Image Zoomable animation" test.
3. Close the image test window.
4. See the crash issue.
Reviewers: raster, herdsman, jpeg, cedric, zmike
Subscribers: zmike
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D5692
Committer's note: rebased and removed unrelated test.
Summary:
A style tag among a text has to be replcaed by its matched tag when
a format node is created. If the matched tag is changed, format nodes
should be updated.
But, if a style text for Evas Textblock Style is changed,
related format nodes are not updated without setting new text.
The patch changes to update format nodes when new style text is set.
@fix
Test Plan: Included in Evas Test Suite.
Reviewers: raster, tasn, herdsman, subodh6129, zmike
Subscribers: zmike, cedric, z-wony, Blackmole
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D4697
Committer's note: rebased.
Summary:
this is on the one side error prone and race, as reacting to a focus out
with a new focused object leads to weird cases. On the other side we are
then using eina hash wrong which leads to later weirdness as the element
that is assosiated with &key is broken.
Reviewers: raster, zmike, stefan_schmidt, cedric
Reviewed By: zmike
Subscribers: cedric
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D6158
Initial results of our static analysis showed a bunch of unused
imports or imports used only for documentation references. In the
first case, remove entirely, in the second case, change to 'parse'
in order to keep references working.
The static analysis is not perfect and yields false negatives for
certain cases, so there will be a second batch later.
Summary:
c->paragraphs couldn't be NULL if it is created by
_layout_paragraph_new() well. So, NULL checking should be
moved to after _layout_paragraph_new().
Test Plan: N/A
Reviewers: jpeg, tasn, raster, herdsman, cedric
Subscribers: zmike, stefan_schmidt, jpeg
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D4300
Summary:
hnj_hyphen_hyphenate2() needs properly encoded text based on the given
dictionary. Each dictionary contains its encoding information at the head
of file. So, text will be converted to proper encoding before calling
the function. It fixes T3221.
@fix
Test Plan: Included in Evas test suite.
Reviewers: z-wony, tasn, woohyun, herdsman, Blackmole, minudf
Subscribers: zmike, stefan_schmidt, raster, cedric, jpeg
Tags: #efl
Maniphest Tasks: T3221
Differential Revision: https://phab.enlightenment.org/D3863
Summary: The Textblock Style which is created for user style was managed
application side. It is created and free'd from application - outside
of Evas Textblock. Recently, evas_object_textblock_style_user_push/pop
start to call efl_canvas_text_style_set() instead of legacy code. The
problem is efl_canvas_text_style_set() is always going to call free()
when a style is going to be deleted. It makes conflicts(double free
issue) with application which is used to call
evas_textblock_style_free(). So, the issue will be fixed by this patch.
The patch also revise push/pop/peek code to make clean and avoid
meaningless calculation/events.
@fix
Test Plan:
A test case is Included in this patch.
The test case try to trigger double free.
Reviewers: herdsman, raster, cedric
Subscribers: zmike, woohyun
Tags: #efl
Differential Revision: https://phab.enlightenment.org/D5812
Committer's note: formatted commit summary (80 width).
Change name of animation event info to synchronize with other event
info names.
Change data name in animation event info to synchronize with data type
name, Efl.Event.Description.
In commit 8367f497b8,
type checking code was added for legacy image functions
in order to reduce failing calls of efl_xxx function.
The default return values of some functions for none-legacy
objects (including NULL), however, were not equal to eo-based
default return value, which broke compatibility.
This commit fixes it.
@fix
Summary: Need to check whether Evas_Public_Data is null or not before dereferencing it.
Test Plan: Execute test suite
Reviewers: raster, Hermet, cedric, jpeg, stefan_schmidt, Jaehyun_Cho
Differential Revision: https://phab.enlightenment.org/D5987
Reviewed-by: Cedric Bail <cedric@osg.samsung.com>
This changes a lot of things all across the EFL. Previously,
methods tagged @const had both their external prototype and
internal impl generated with const on object, while property
getters only had const on the external API. This is now changed
and it all has const everywhere.
Ref T6859.
Summary:
move evas_async_events_shutdown: to out of EVAS_CSERVE2 ifdef block
to make it reachable.
Reviewers: cedric, woohyun
Differential Revision: https://phab.enlightenment.org/D5926
Reviewed-by: Cedric BAIL <cedric@osg.samsung.com>
since this can take an extension as well as a file path (extension
being .gif or .jpeg etc.) it would skip if passed a small extension
only (5 chars or less). fix and this fixes e's thumbnailing too for
some files.
@fix
Summary:
The fribidi couldn't reorganize paired brackets (Ex. '(', ')')
when there is RTL + LTR text. According to TR9(http://www.unicode.org/reports/tr9/),
it has to be shown properly without LRM or RLM.
Also, from the fribidi 1.0.0, fribidi_get_par_embedding_levels() was deprecated.
It is replaced with fribidi_get_par_embedding_levels_ex() which is including
paired brankets rules from TR9.
@feature
Test Plan:
1. Create a elm_entry.
2. Set a text by calling text_set.
elm_entry_entry_set(entry, "مرحبا Hello (40)");
3. Run and see the results.
- Without this patch or fribidi 1.X.X, it will show text like this...
"(Hello (40 مرحبا"
- With this patch and fribidi >= 1.0.0
"Hello (40) مرحبا"
Reviewers: raster, cedric, herdsman, woohyun
Reviewed By: herdsman
Differential Revision: https://phab.enlightenment.org/D5921
Change the parameter name of property event_animation from "event_type"
to "desc" to synchronize with other APIs.
(The parameter type is Efl.Event.Description)
Previously, beginning state of animation is applied after animation is
finished if final_state_keep_get() is false.
Now, map is reset (no animation is applied) after animation is finished
if final_state_keep_get() is false.
To synchronize other types of event info,
Efl_Canvas_Animation_Player_Running_Event_Info is changed to
Efl_Canvas_Animation_Player_Event_Running.
Remove unused struct.
Old version algorithm was imperfection a bit, quality was poor at some specific
degrees, specifically, when pixel increment pattern on the diagonal lines is
unstable.
This revised version was better than old one even source code is much cleaner
and simpler.
See belows.
*NonAA vs AA:
https://ibb.co/bCNfMc
*Compare the worst case aa in the old version:
https://ibb.co/bEJsZx
*Test video:
https://youtu.be/Wn20Tym5lfg
Summary:
this commit add null check.
In EVAS_OBJECT_DATA_SAFE_GET, if eo_obj is null, obj can also be null.
and In _efl_canvas_output_async_block() can return null.
Test Plan: N/A
Reviewers: woohyun, kimcinoo, cedric
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D5870
Reviewed-by: Cedric BAIL <cedric@osg.samsung.com>
[Dereference after null check]
(1) src/lib/ecore/ecore_main.c
- _efl_loop_handler_efl_object_finalize checks if pd->loop_data is NULL.
After that, _handler_reset > _handler_clear > _ecore_main_fd_handler_del >
_ecore_main_fdh_pool_del is directly dereferencing pd->pool_data.
- _efl_loop_handler_efl_object_parent_set checks if pd->loop_data as well.
Then it calls _handler_reset as well.
(2) src/lib/ecore_wayland/ecore_wl_dnd.c
- ecore_wl_dnd_selection_set checks if t - result of wl_array_add - is NULL.
And it is dereferecing t directly for wl_data_source_offer.
(3) src/lib/elementary/efl_ui_dnd.c
- Third parameter const char *data could be NULL.
In this case strlen dereferences NULL. The data should be non NULL value.
I have checked this with Mr. Thiep Ha.
(4) src/lib/evas/canvas/evas_object_inform.c
- _efl_canvas_object_efl_gfx_stack_stack_below checks if obj->layer is NULL.
So it could call evas_object_inform_call_call_restack which is dereferencing
obj->layer directly.
Evas map supports anti-alias(aa) rendering on sw backened.
When aa is toggled on, map forcely turns alpha channel on while it draws on the surface.
Actually, it was intended to blend polygon edges with destination,
but it breaks one case if the original source image alpha channel were turned off.
Simply, it fixed the issue, new implmentation removes the alpha channel switching,
instead fill the alpha values with 255 when map + aa + alpha_off is drawing on it.
@fix T1975
Evas map anti-aliasing haven't worked at all if the smooth scaling were disabled.
evas map rendering has a lot of corner-cases, previous call-position was wrong,
(by mistake maybe) shouldn't be in a certian case.
Let aa post-processing function be performed in universally.
Ector renderer are created by a factory pattern which return a reference.
The possible logic improvement from here would be to rely on setting the
parent of the object to the Ector_Surface and manage a cache by intercepting
efl_noref to trigger an efl_del of the object or a temporary cache of it.
That's for later.
Summary:
this commit add null check logic on evas_object_smart_attach
because a segmentation fault occurs once the argument is not valid (eg. null).
Test Plan:
1. invoke evas_object_smart_add(obj, NULL)
2. see the application crashes
Reviewers: woohyun, cedric
Differential Revision: https://phab.enlightenment.org/D5817
Signed-off-by: Cedric Bail <cedric@osg.samsung.com>
ok. i can't find the root cause because all i have is a backtrace from
ApBBB and he says he can't reproduce it and i know im->cache is
null... if i could reproduce ... i'd be hunting the root cause. but
the best i can do is check for null im->cvache ptrs and be safe.
crashes are bad. especially for end users.
so the MAIN loop is actually an efl.app object. which inherits from
efl.loop. the idea is that other loops in threads will not be efl.app
objects. thread on the creator side return an efl.thread object.
inside the thread, like the mainloop, there is now an efl.appthread
object that is for all non-main-loop threads.
every thread (main loop or child) when it spawns a thread is the
parent. there are i/o pipes from parnet to child and back. so parents
are generally expected to, if they want to talk to child thread, so
use the efl.io interfaces on efl.thread, and the main loop's elf.app
class allows you to talk to stdio back to the parent process like the
efl.appthread does the same using the efl.io interfaces to talk to its
parent app or appthread. it's symmetrical
no tests here - sure. i have been holding off on tests until things
settle. that's why i haven't done them yet. those will come back in a
subsequent commit
for really quick examples on using this see:
https://phab.enlightenment.org/F2983118https://phab.enlightenment.org/F2983142
they are just my test code for this.
Please see this design document:
https://phab.enlightenment.org/w/efl-loops-threads/
This reverts commit 135154303b.
Revert "efl: move signal events from efl.loop to efl.app"
This reverts commit 3dbca39f98.
Revert "efl: add test suite for efl_app"
This reverts commit 3e94be5d73.
Revert "efl: create Efl.App class, the parent of Efl.Loop"
This reverts commit 28fe00b94e.
Go back to before efl.app because I think this should be done with
superclassing here not a parent object. reasons?
1. multiple loops per single thread make no sense. so if multilpe loop
objects they wont be contained in a single app object and then deleted
like this.
2. the app object is not really sharable in this design so it cant be
accessed from other threads
3. it makes it harder to get the main loop or app object (well 2 func
calls one calling the other and more typing. it is longer to type and
more work where it is not necessary, and again it can't work from
other threads unless we go duplicating efl.app per thread and then
what is the point of splittyign out the signal events from efl.loop
then?)
etc.
also implement Efl_Canvas.objects_at_xy_get
note that any function which returns an iterator cannot be @const since
it's necessary to wref the object to ensure the iterator's lifetime
this file contains canvas methods which should be subject to compiler
type checking, requiring that they be included in the file which contains
the eolian generated code
this reveals existing breakage which went unnoticed due to the compiler
being unable to do type checks since the generated eolian code was not included
Summary: As the gfx_path mixin class destructor is not called hence explicit call of reset.
Reviewers: jpeg, SanghyeonLee
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D5803
Signed-off-by: Cedric Bail <cedric@osg.samsung.com>
every canvas object has had this leak for some time:
==25217== 64 bytes in 1 blocks are definitely lost in loss record 1,669 of 3,515
==25217== at 0x4C2FB6B: malloc (vg_replace_malloc.c:299)
==25217== by 0x688D839: eina_hash_new (eina_hash.c:755)
==25217== by 0x688D940: eina_hash_string_superfast_new (eina_hash.c:791)
==25217== by 0x6265F8C: _efl_gesture_manager_efl_object_constructor (efl_gesture_manager.c:66)
Summary:
Calling evas_object_polygon_add(NULL) cause segmentation fault,
because there has been removed null check logic in evas_object_polygon_add(0ff7429)
This patch fix the problem.
Test Plan:
1. Call evas_object_polygon_add(NULL)
2. Check segmentation fault happen
Reviewers: cedric, jpeg, woohyun, jypark
Differential Revision: https://phab.enlightenment.org/D5801