Next step: state handling in the GStreamer backend
Reviewers: cedric
CC: cedric
Differential Revision: https://phab.enlightenment.org/D431
Signed-off-by: Cedric Bail <cedric.bail@free.fr>
At least on the gstreamer1 version in Fedora 19 this include is needed. Glima
reported it as well and I think he also uses Fedora.
modules/emotion/gstreamer1/emotion_gstreamer.c:643:4: error: unknown type name
'GstNavigationCommand'
Even if other distros or gstreamer1 versions do not need this it should be safe
to add it here.
in SOME cases a wm may deny a resize request entirely and keep the
window size the same. it does so with a configurenotify of the SAME
size as before. problem is we never passed this onto the app with a
resize callback of the original size before request to the new size.
this does that now.
cherry-pick me!
This code was disabled and did not work properly. If this should be
re-added it will have to be rewritten anyway and currently it only
clutters the existing code.
Reviewers: cedric
CC: cedric
Differential Revision: https://phab.enlightenment.org/D414
Signed-off-by: Cedric BAIL <cedric.bail@samsung.com>
Being annoyed by different types of eina critical macros - CRI, CRIT,
CRITICAL -, I concluded to unify them to one. Discussed on IRC and
finally, CRI was chosen to meet the consistency with other macros -
ERR, WRN, INF, DBG - in terms of the number of characters.
If there is any missing bits, please let me know.
Some commits to port emotion to GStreamer 1.0 and implement some missing features,
clean up code a bit and fix some bugs on the way.
This works as good as the 0.10 code for me now with the emotion examples,
just the Samsung hardware specific code is commented out. This should be
ported by someone who has such hardware, and also in a clean way now that
GStreamer since 1.0 has features to handle all this properly.
There's still a lot of potential to clean things up and fix many bugs, and also
to implement zerocopy rendering. But those are for later if there's actual
interest in this at all.
Commits:
- Update configure checks and ecore example to GStreamer 1.0
- Initial port of emotion to GStreamer 1.0
- Samsung specific code commented out, should be ported by someone
with the hardware.
- Return GST_FLOW_FLUSHING when the sink is unlocked
- Remove unused GSignal from the sink
- Use GstVideoInfo to store the format details inside the sink
- Add support for pixel-aspect-ratio
- Store video format information in GstVideoInfo for the different video streams
- Use GstAudioInfo to store the audio format information
- Remove some unused defines
- Header cleanup
- Implement initial support for GstNavigation interface
- Implement setting of audio/video channel
Reviewers: cedric
CC: cedric
Differential Revision: https://phab.enlightenment.org/D387
Signed-off-by: Cedric BAIL <cedric.bail@samsung.com>
- create permissions as 0700, this should be private to the running user.
- do shm_unlink() on error, prevents resource leaks;
- apply close() on error, prevents resource leaks;
- handle strerror() properly.
- shm name uses process id, object name and ecore's time, should be
cleaner and avoid possible clashes
- ecore_exe do not need to read from slave stdout/err, this removing
the flags allows messages to come visible at terminal.
- fix typo, it's slave not slayer... :-D
- errno: you can't use errno to report error from failed mmap() if
you do another call that sets errno after that, in that case
fprintf(). So have a single print so we can use the errno value of
the proper call.
- although the slave generic player is not that important if it leaks
its single shmfd, let's be pedantic and close it. Manpage says it
is safe to close filedescriptor after mmap() worked.
Git log should be enough for the credit :)
This code was merged into efl upstream by devilhorn's aggressive synchronization between efl upstream and tizen.
But that commit didn't have the author's credit.
However We will never forget the contribution of Gwanglim Lee <gl77.lee@samsung.com>.
when the window alpha mode is changed, the x window is re-created.
but if the window is going on the show state(not visible yet but waiting for visible from xserver) this window never be visible.
so here it make the window visible also if the window is going on the visible state.
parsing problem with opengl_strtok() which would free the previous
token "p", but in some cases it would be a const string. this should
fix CID 1039653
this fixes CID 1039884 which isn't a real problem as the callback del
never dereferences the data pointer - just uses it as a value, but
this is really to ensure that it doesn't come back if the code were to
change.
stable release - cherry-pick me!
the evas gif loader used way too much cpu to decode animated gifs
because in the rewrite that made it correct, it did not store the
current gif file handle and state, thus each frame it would have to
decode all frames before that one before finally decoding the final
one. that means to decode frame 200, it decoded frame 1, 2, 3, 4 etc.
all the way up to 199 THEN decoded 200 on top, so decode cost became
progressively more then further through the animation you were.
this fixes that by storing state and file handle and allowing you to
iterate through.
There's no need to call it on text_input_leave too, otherwise this would
be called twice, the one from text_input_leave possibly being called
after the focus was regain already by a text input, causing the bug
described in T237.
This fixes T237.
In my config, running terminology with the GL engine and under
cserve2, some image could not be loaded. The tex argument
in evas_gl_preload_target_[un]register was NULL, leading to
an immediate crash.
This fixes Phab #T611. Previously, windows which were opaque would
become transparent after being hidden and then shown. We fix that by a
call to ecore_wl_window_alpha_set when the window gets shown. This
patch also brings the wayland_egl hide code more inline with the shm
hide code by testing if the surface does not match the existing one.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Properly fix efl wayland elm window resize problem.
This adds support for min, max, step, aspect, and base size properties
when resizing a canvas under EFL Wayland.
This Also Properly fixes raster's report for EFL Wayland elm windows not
resizing properly. Previously, when resizing an elm window in wayland,
a portion of the window would draw outside the frame.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Before fixing this issue, the cursor of preedit appears the in front of preedit string.
The cursor of preedit string will appear in the proper position.
outside the window).
Be sure that the EEs requested geometry gets updated in our
common_resize function After min/max have been taken into account.
Elm is using ecore_evas_request_geometry_get in it's resize_job code
(Why...I have no clue lol. Does not seem like a good thing to check).
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Ecore_wl_window_resize essentially tells the shell to start the resize
process. We should have the evas engine info's resize_edge updated
Before we start that process so that Evas knows where the resize is
occuring from
Signed-off-by: Chris Michael <cp.michael@samsung.com>
This callback must be manually destroyed and removed on
ecore_evas_hide(), because it won't be delivered anymore after the
surface is destroyed. If the callback still exists, the engine will find
it and avoid doing a new redraw until it is finally called.
Maybe the correct thing to do is to keep this callback in the
Ecore_Wl_Window struct, and have some functions to set/unset it, so it
gets destroyed when the window is hidden. Or when the surface is
destroyed.
We must unset it, so the Evas engine knows that it has to do all the
setup to destroy the EGL surfaces associated with it. And in the case of
an ecore_evas_show, recreate everything again.
When an Ecore_Evas is hidden, it will destroy the buffer swapper. When
it's shown again, it will try to attach a new buffer, that can be same
buffer. If that global var is still pointing to the old buffer, it can
match to it again and avoid sending a new buffer. So, just put this sent
buffer var in the buffer swapper, and it will get set to NULL when the
swapper is destroyed and created again.
This should fix an intermitent problem of ecore_evas_show() not always
working after an ecore_evas_hide() on the wayland_shm engine.
Fix phab T392.
Notice that it should reopen T359, as it wasn't really fixed, but a
rotation with resize was being used when a non-resized rotation was
requested. The cause of the "protruding surfaces" is likely the fact
that Elementary is setting the opaque regions manually, instead of
leaving it to Ecore_Evas. This must be fixed either inside Elementary
itself, or adding the "surface extents" (shadow/non-visible surface
parts) info to Ecore_Evas too.
This reduces code duplication for ecore_evas resize routines inside
wayland engines by using a common _ecore_evas_wl_common_resize
function as the resize code for both engines was the same anyway.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
[Problem] When glTextureDelete is called in image_cache_flush(), it sometimes doesn't work.
[Cause] glTextureDelete is called with the wrong eglContext.
[Solution] Call eng_window_use() in image_cache_flush() and image_cache_set() to use the correct eglContext.
Change-Id: Id7ab1aaeb456be6dbc5f09cb2731ace5399a5dce
Signed-off-by: Cedric Bail <cedric.bail@samsung.com>
In evas_gl_api_ext_def.h there're calls such as:
_EVASGL_EXT_DRVNAME(EGL_KHR_image_base)
The macro is defined in evas_gl_api_ext.c as:
(strstr(glexts, #name) != NULL || strstr(glueexts, #name) != NULL)
if (_EVASGL_EXT_CHECK_SUPPORT(name)) *ext_support = 1;
But EGL_KHR_image_base is itself a macro, which is defined
in EGL/eglext.h like this:
Thus, the _EVASGL_EXT_CHECK_SUPPORT macro will unwrap into:
(strstr(glexts, "1") != NULL || strstr(glueexts, "1") != NULL)
instead of intended:
(strstr(glexts, "EGL_KHR_image_base") != NULL ||
strstr(glueexts, "EGL_KHR_image_base") != NULL)
This patch fixes this by applying stringification earlier in
_EVASGL_EXT_DRVNAME
Bugfix reported by jinhyung.jo@samsung.com
These two APIs will save the title and class_name inside
Ecore_Wl_Window, so if they are called before the shell surface is
created, the stored names will be used later when the window is finally
shown (shell surface is created).
This way we are also hiding the shell surface from ecore_evas modules.
This is a configure event info, so put it in the right place. Some
places adding edges info were also removed, which means that they were
redundant.
Tested with <Meta> + middle click resize, and with window border resize,
on all the edges. Apparently, nothing breaks.
It's mostly only used inside ecore_wayland library anyway.
The only bit needed outside of the library is the "seat" pointer, but a
new function was added to retrieve such pointer from Ecore_Wl_Input.
This struct is only used inside ecore_wayland itself, so there's no real
reason to expose it. Apparently, hiding it doesn't break anything,
except for the ecore_imf wayland module, which was easily fixed.
If anyone notices a breakage, please let me know.
Previously, a "maximize" of an efl client would not actually resize to
the provided size as it would account for framespace width & height.
This fixes that so now "maximize" is Actually maximized.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
needed.
NB: Currently, this will only work for fullscreen or maximized states
as other ones like sticky, withdrawn, iconic, etc, etc are not
supported via wayland protocol yet.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
after several days of beating head on desk, i gave up trying to find
the exact cause of some gifs not rendering right as animated gifs due
to the loader. it had something to do with dispose mode handling and
which frame it was applied to. i noticed the structure made it also
hard to fix, so this re-structures the entire thing with cleaner code,
less code AND more comments, with a limited memory cache (512k) for
previous frames per image (to avoid eating up huge memory blobs for
big/long animations - though at the expense of cpu), and with some
notes for future fixes - like fixing the "load 2 copies of the same
animated gif" issue... that's another day.
this does fix https://phab.enlightenment.org/T443 along with many
other things.
At the end of the ecore_wl_init() function, send a sync request to the
server, and add a callback listener to the "done" event. When this event
is received, we are sure that all the registry bind requests done so
far were processed already, and that the registry and globals are
available and can be used.
Now, on the functions that request interfaces or registry, we call
_ecore_wl_init_wait(), which will check if the callback was received
already (that means that all requests inside the init were processed).
If it was not yet, then we wait until receiving that callback, before
returning the requested data.
Simply call the appropriate cache2 functions when possible
and check for usage of cache2 whenever an evas_cache_ function
is called.
This effectively adds cserve2 image (data) load support for the
GL engines. Fonts were already working out-of-the-box.
Let's reuse the logic from scalecache and call cserve2
functions when the scalecache should be used.
So, now, cserve2 server will not scale any image... This is
too computationally intensive for the server's main thread.
This is not optimal but makes a hell of a lot more sense for
the moment. (since cserve2 manages the SHM segments)
Pass around "animated" flag for images that can be animated.
Fallback to local cache if the image is animated.
Implementing support for animated images in cserve2 does
not seem to make a lot of sense considering each frame must
be requested independently in real time,... and to be honest
there doesn't seem to be any valid use case anyway :)
cserve2 can't handle virtual files (mmap-only), by design.
Proper support can be added later on, but for now we might want
to just fallback to the normal cache functions.
Fixes photocam test
Evas GL direct rendering mode didn't properly take into account
the image object's clipping information and clip the region that
it was directly rendering to. Hence there were issues with the
direct rendering region drawing over the objects that are sitting
on top of it.
Also, cleaned up the direct rendering coordinate computation code
and a nasty dependency with image object that should have been
removed a long time ago. Basically the evas-gl engine was directly
accessing the image object data structure for its data when it
really should have just passed along necessary information.
This reverts commit 9bfb730ff5.
wrong! on deiconify this will now FORCE a ACTIVATE request to the wm
(which asks the wm to focus the window/hilight it and make it active.
might switch desktops or whatever). this is wrong. de-icoifying a
windows does NOT mean ALSO activating it!. the log message also doesnt
say what "go back to normal state"... means.... do they expect/INSIST
windows MUSt be focused after de-iconification? that's wrong as it's a
window management FOCUS policy, not something to do in the lib.
Is there a title for the commiter with the longest revert of a revert list?
I screwed this one up. Had other local changes that sneaked in.
Resetting and doing it right now.
This reverts commit ee155b771d.
You can bet on reverting makes people speed up the process in fixing it.
Revert the revert here now that Cedric fixed it in eina.
This reverts commit 875e7cf74d.
This patch brings in constant E crash for me. It seems to be timing
related which somehow leads to a mem corruption. Revert it for now
to allow people using E while Cedric looks into it.
This reverts commit 12d34309c7.
This add finally support for JPEG 2000, but be aware that libopenjpeg
is very badly managed. There is currently only version 1.5.x that does
provide the right files, is usable by a third party and portable. You
can seriously forget any other version.
NB: We will still create all new canvases as alpha by setting it in evas,
however this allows the 'state' of ecore_evas_alpha_set/get to retain
it's original functionality and thus accurately report what a client
application May have set alpha value too.
NB: Fixes Phab Ticket T350
Signed-off-by: Chris Michael <cp.michael@samsung.com>
For EvasGL direct rendering, EvasGL does a make_current to the
surface that evas is holding on to. When EvasGL was shutting down
it was wrongly deleting evas' surface. This issue was temporarily
fixed by Raphael before but the proper fix was added.
Remove calls to ecore_wl_window_update_size inside the alpha &
transparent_do functions (this call not needed here anymore).
NB: This is needed due to recent frame "fixes". Basically, even tho we
set the opaque region correctly on a surface now, if the canvas itself
is not alpha, then we will end up with a black "spacer" around the
window.
NNB: Tested Both engines with terminology (both translucent and not) and also the window state test in elementary.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
There was already a surface created by _evgl_tls_resource_create(). If
we assign a new one here, the wrong one will be destroyed at
_evgl_tls_resource_destroy(), and later the GL window will be destroyed
before the surface, causing invalid access errors.
This fixes https://phab.enlightenment.org/T326
If the egl_surface is different from the current one, it may be that the
it has been destroyed already. Removing the below check (and just
checking for different contexts) will avoid calling makecurrent when
destroying a window. That was always failing anyway.
Should fix https://phab.enlightenment.org/T311 for gl_x11 too.
If the egl_surface is different from the current one, it may be that the it has
been destroyed already. Removing the below check (and just checking for
different contexts) will avoid calling makecurrent when destroying a window.
That was always failing anyway.
Should fix https://phab.enlightenment.org/T311 for wayland_egl.
duplicated code.
Resize the frame object before we update the window saved size.
Remove (again) call to _ecore_evas_wayland_resize and set the resize
edge of the window.
NB: The call to _ecore_evas_wayland_resize ends up sending duplicate
configure events here, hence whey it is removed.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
At least on recent mesa (since commit 9f07ca11c17), it will find the
mentioned symbols but they won't really work, leading to error messages,
and possibly some other errors. So far, I just ifdef'ed the
glGenFramebuffer and glBindFramebuffer functions, but it may require
others to be ifdef'ed too.
This is just a workaround, to fix https://phab.enlightenment.org/T246.
NB: Not sure how/why this was here, but it's entirely Not needed and
leads to duplicate calls of wl_shell_surface_resize.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
It's always enabled as it's a dbus module and links to nothing,
actually the daemon doesn't need to be running -- in that case it will
do nothing. In the case the daemon becomes active then it will get the
OnLowBattery property and keep it in sync.
NOTE: I couldn't test the property change as my laptop takes many
hours to get to that situation... let's hope it works :-)
Ecore will now load "system modules" on ecore_init(). The "systemd"
module will use DBus to monitor localed, hostnamed and timedated and
add system events related to those changes.
We don't want to execute any of the function below if t is NULL. Its not
only about strdup but alsa ecore_x_netwm_name_set which might dereference it.
CID 1039400
NB: These calls are NOT needed here (they setup and tear down the
actual wayland display fd connections, and are not necessary in order
to use ecore_wl functions).
Signed-off-by: Chris Michael <cp.michael@samsung.com>
erroneous calls to engine_setup code if we check surface validity.
NB: Short Version: Don't reassign surface if we don't need to.
NB: Prior to this, during a call to ecore_evas_show we would always
Reset the evas engine data (which was basically triggering code to
tear down and reconstruct all the engine internals including the
buffers and the swapper). By checking surface validity first, we can
potentially avoid all that teardown and reconstruction Thus resulting
in faster "show" speed.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
NB: When trying to attach a buffer to a surface, there was a corner
case that could fail in that if we just had 'rects' but count was 0,
then the surface would not get damaged (essentially a call to
wl_surface_damage would get passed 0,0 as the size). This fixes that
problem in that if count is <= 0 now, then we'll damage the whole
surface.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
functions on it.
NB: Wayland (internal) functions typically do not do Any object
validity checking, so lets be proactive and do it here first.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
ecore_wl_globals_get before trying to use them.
After we have bound a text input manager, there is no need to keep
looping the globals, so break out.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Porting commit f10c108283d9b76708418ba172b9ff0b28a09cd6 to wayland
engine.
Added a feature to dump out a png image before calling SwapBuffers.
To turn on the feature, use the following env vars.
EVAS_GL_SWAP_BUFFER_DEBUG_DIR="dir_name" sets the name of the directory
that the files will be output to and enables the debug mode.
EVAS_GL_SWAP_BUFFER_DEBUG_ALWAYS=1 will enable the writing of the
files every frame. This in effect turns on the swap_buffer_debug
variable, which can be set on or off in gdb for debugging. This will
allow dumping of certain frames without having to dump out a ton of
files to track down one frame.
Added a feature to dump out a png image before calling SwapBuffers.
To turn on the feature, use the following env vars.
EVAS_GL_SWAP_BUFFER_DEBUG_DIR="dir_name" sets the name of the directory
that the files will be output to and enables the debug mode.
EVAS_GL_SWAP_BUFFER_DEBUG_ALWAYS=1 will enable the writing of the
files every frame. This in effect turns on the swap_buffer_debug
variable, which can be set on or off in gdb for debugging. This will
allow dumping of certain frames without having to dump out a ton of
files to track down one frame.
- ecore_evas_pixmap_visual_get
- ecore_evas_pixmap_colormap_get
- ecore_evas_pixmap_depth_get
Assign interface functions for above when creating the new interface.
When creating the pixmap, store the visual and colormap for later
retrieval.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Since we take the lock at the start of the function, we should release
the mutex lock if we are going to return.
NB: Fixes Coverity CID1039381
Signed-off-by: Chris Michael <cp.michael@samsung.com>
- Add "pixmap" structure to engine_data.
- Free the front and back pixmaps when the ecore_evas gets freed.
- On resize, drop the backing pixmap if size is different.
- Before we render, check pixmap size vs EE size and create new
backing pixmap if needed.
- After we have rendered into the backing pixmap, flip it to the
front so that any calls to "pixmap_get" will return the proper one.
- Add pixmap_new_internal code (the actual guts for all this)
NB: These can be used to create an internally double-buffered
ecore_evas that renders to a pixmap (not a window).
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Carsten Haitzler
Cedric Bail
Chris Michael
Sebastian Dröge
Jihoon Kim
Stefan Schmidt
Vostokov Sergey
Daniel Juyung Seo
Dennis Schridde
Gustavo Sverzut Barbieri
Cedric BAIL
Guillaume Friloux
ChunEon Park
Rafael Antognolli
Jean-Philippe Andre
Chris Michael
Stanislav Vorobiov
Sebastian Dransfeld
Wonsik Jung
Sung W. Park
Cedric BAIL
discomfitor
Michal Jagiello
Vincent Torri
Davide Andreoli
José Roberto de Souza
Mike Blumenkrantz