this is generally unnecessary noise as the majority of the time i see
this it's because the message was canceled by the caller... thats not
an error though then... it's intended. less debug noise to sift through.
basically ecore_con's test suite was broken. ipv6 checks were broken.
you now need to set an env var to get ipv6 tested. using eocre_con to
detetc it didnt init eina or ecore_con and if you do init these then
the other tests that check return values from these fail... so make it
an env var if you want to test ipv6. now it's reliable and works!
min value bitshifts are negative and thus not portable. just tax max
(as its 1 more than min it if we do -max) as the limit as its within
range. this should fix it
@fix
found by PVS studio
no point - all we are doing is having a final fallback of any tag that
starts with ! that isnt a special one like !DOCTYPE, !-- comment and
![CDATA stuff... analysers dont like these pointless calls.
found by PVS studio
extra check already done in if above - no need. looks like copy &
paste of the checks from above thus why it's there, but not needed.
not a bug, byt analysers dont like it so remove
found by PVS studio
test for non null already done above, so test must obviously be null
here so no need to check. analysers dont like this but it's not a bug.
make them happy thought
found by PVS studio
one if condition is always true by virtual of previous if statements
and drop-through so can remove. not actually any bug but analysers
don't like it
found by PVS studio
analysers like PVS studio dont know the eina_list_free returning NULL
design pattern for consistency and so dislike the var being set twice
in a row. this will make them happy without any skin off our backs.
not a bug at all.
found by PVS studio
only the else changes finfo so reset inside there. not really any bug
at all byt style-wise a bit better and analysers don't like it
found by PVS studio
eina_list_free always returns NULL. it's meant as a design pattern to
follow like the append/prepend that ensures a freed list is NULL, but
it can be confusing to analysers so remove the assing. not a bug
actually.
found by PVS studio
so a type we handle earlir inan if we re-handle as invalid later. this
wouldnt lead to a crash or bugs as the if's would ned to be evaluated
in order normally, but it's good to get it right.
found by PVS studio
surprising this wasn't caught when testing fallbacks. etc2 is used
hyper-rarely but still... basically it'd skip pixels and read stack
junk (which really would not crash but we'd have junk in image data).
or should.
found by PVS studio
@fix
this fixes a bit wraparound in the shift as the 1 is an int (32bit)
type that then gets shifted .. then after that cast to 64bit.
found by PVS studio
@fix
this actually wasn't a bug that would cause a crash. cloning an array
access would fail as the magic check would find its an accessor not an
array. indeed a bug... but we never used this anywhere i can find.
this was cast to the correct func ptr callabck in the accessor struct
as the clone method though.. thus everyhting was happy with it
seemingly.
found by PVS studio
@fix
libinput won't actually give us this info, and logs an error instead.
We can't synthesize based on pointer position due to pointer
warping.
So we must track abs reports and use previous abs reports to
synthesize rel events.
So vmware's graphics driver reports the MONOTONIC drm cap, yet uses
CLOCK_REALTIME instead. This leaves us with a gigantic offset between the
gpu timestamp and the times ecore_time_get() gets from CLOCK_MONOTONIC.
Since ticking screws directly with loop time this results in some long
distance clock jumping.
This commit fixes drm/gl_drm operation under vmware.
We can't depend on vblank waits being implemented by the driver, but we
can count on page flips functioning, so add a fallback that does a page
flip and waits for it.
This lets us do a blocking wait for a vsync. Something we should try to
do as infrequently as possible, but in some cases we need it one time at
startup to catch graphics driver bugs.
This patch sets some Ecore_Con flags that were missing after
the EO migration. These flags must be set in order
maintain the Ecore_IPC behaviour before Ecore_Con EO was implemented.
Fixes T5722
according to q66 this is correcter than ptr(..).
Anyhow this still leads to a bug in cxx that crashes compilation in cxx
examples, people are notified!
usecase:
show -> rotation -> hide -> show
ecore_evas_wayland didn't check the rotation.
when ee is landscapemode, it cannot update the right area of evas.
Our close callback tells logind we're done with a device, but it should
also actually close the fd it's passed, or we end up leaking piles of
fds on VC switch.
see weston commit 8f5acc2f3a29c3831af4ddd6bed57f703c98dc77
and subsequent regression in commit 72dea06d7952e3ce8dd8057f7106186da4fa2678
and pending fix in https://patchwork.freedesktop.org/patch/168992/
When destroying any object, its parent class destructor should
be called after the subclass destructor has been called. Only
some extremely limited work may be done after the super call.
This commit makes sure that all efl_destructor() overrides in
elementary are doing operations in the right order.
Also, remove a return void.
with this we can whipe out the focus.manager field in elm.widget so for
the case that something goes wrong we only get the error message where
actually something went wrong, and not the whole bunch of follow ups
where the code assumes its registered but it isnt.
When I added the code to probe drm devices to ensure they're
modeset capable (ref 414d406b3b)
I didn't realize elput didn't allow us to open and close more than
one drm device at startup without blowing up libinput.
This is a somewhat dirty hack to rough that in.
The problem is that open/close the device during startup will
result in an async "gone" callback from logind, which then kicks
off an input shutdown. We need to try harder to only do that
shutdown when it makes sense.
if files size was 0... which would basically make a lot fail before...
it could be a divide by zero. fix and just assume minimum size of 1.
fix CID 1371127
Ecore_Audio.h had #ifdef HAVE_PULSE ... and that is just so wrong as
this is something an ap[p may or may not set in its config.h ... but
this certainly hase no place in our public headers. the api's there
should always be there... symbols always... just imtplementation may
be empty ... though ecore_audio doesnt build at all without pulse...
so it's moot.
so fix this build issue i found when fixing e build issues with meson
changes etc.
@fix
Some systems have dri devices that can't mode set, and if they're first in
the directory they'll get picked by our code and things fall apart later.
So, open the potential device and ensure it has basic functionality before
selecting it.
This is a little inefficient as it gets the device via elput twice before
it can be used - this will be addressed later as the changes are a little
more invasive to optimize.
csd attempts to create sw cursors if the pointer surface doesn't exist,
so create the pointer surface before the csd to avoid duplicate cursors
#TheDisappointer
this should be unmapped so edje stops trying to render to it, though
at some point it could be optimized to not hide and just reuse the buffer
when showing again
#IGot99TicketsBut1x1AintOne
Just because the #define is present doesn't mean the extension is, so we're
BAILing on egl completely on some systems for no good reason at all. (saw
this on an SGX stack)
This is still wrong. I don't want to try too hard until after the upcoming
release, though.
We should actually be testing for the presence of client extensions before
attempting to do any of this. It's entirely possible that a gl stack will
return bogus functions for these from eglGetProcAddress
asking for an x- field from a desktop shouldnt result in null safety
checks IF the x hash is null - it's fine if it's null - that field
just doesn't eixst. if other fields were in the x hash but the desired
one wasn't it'd not complain, so why complain if there is just no x
hash?
fname would have been leaked if we failed to get a file_handler.
Also make sure we always create a new fname strbuf even without
a output_dir present to fix a potential unbalance between creating
and freeing.
CID: 1378112
it seems like the visibility is evalulated from different parent
relations, Just because the box (that is a container of a element) is
not visible, does not mean that the element in it is also not visible,
this somehow comes back to the box beeing deleted, the children beeing
reparented in evas, but not in elm. This is for now just a quickfix,
since those calls are just for optimization.
This fixes a load of errors once the elm_test window is closed
@fix
This fixes a crash in make check when --profile=dev is explicitely
enabled. eina_list_init() is called by the standard eina_init() loop
and by eina_debug_init() as well.
Honestly I'm not sure why it doesn't crash for other people as
well...
The crash was in eolian_suite during the second eina_init (called
from eolian_init).
Ooops! When mapping for writing we can't use the same code
path as when tofree is true. This restores the path for image
writing. This basically fixes rage with GL engine.
See also 45c8e5e983
Summary:
Actual slider value is not changed.
So when user release mouse or touch, slider value is returned
to the value by user set.
This is a partial revert of f41e276160
Fixes T5745
Test Plan: elementary_test -> slider
Reviewers: jpeg
Subscribers: cedric
Maniphest Tasks: T5745
Differential Revision: https://phab.enlightenment.org/D5031
Some objects don't go through render_pre (unchanged, child
of mapped parents), even though they will go through
evas_render_mapped. Those were marked as pre_render_done
inside evas_render_mapped since it seemed to fix some issues
a long time ago.
Unfortunately, if those objects are changed their flag may
not be reset to false, which means they never go through
render_pre, leading to render issues.
I believe simply restoring the value of the pre_render_done
flag should be good enough. I don't know why it is set to
true inside evas_render_mapped but I also don't want to find
out :)
See 9ac13e4aec (old)
See 87e5e70a9d (older)
Patch made with @jiin.moon
If an object fails to call efl_destructor() on all the parent
classes, then it may never be removed from the layer object
inlist and never would its data be deleted either (eo keeps
it safely alive).
Here's how to test: add a "return;" statement inside an object's
destructor (eg. the window class).
Fixes T5680
A recent commit broke texture_from_pixmap for NVIDIA EGL
(again), because eglCreateImage is a symbol in libEGL.so
but isn't in fact implemented by the driver.
That's because eglCreateImage() is exposed by libglvnd but
the underlying EGL implementation is NVIDIA and its version
is only 1.4, not 1.5 (where the API was introduced as core).
Instead of reverting the patch, it's better to cover our
bases properly and use dlsym() only if the version is right.
Note that GetProcAddress() may return garbage function
pointers for ALL functions as dynamic virtual functions may
be created on the fly by libglvnd. So it is absolutely
necessary to check the extension string as well.
See 0255f14dc2
eglCreateImage is objectively better than eglCreateImageKHR - it allows
attributes large enough to hold pointer values. We should use it when
available and only use the older extension version as fallback.
Also, eglCreateImage is core EGL functionality so don't depend on extensions
to be present to use it. Theoretically we should be testing for
EGL version >= 1.5 but it's probably safe not to.
This was only necessary due to bugs in the wayland_egl and gl_drm engine
that have been corrected.
Wayland has no bizarre requirements making this necessary.
Commit 2e6587a14b adds a gl extension string to glsym_evas_gl_symbols()
to prevent using functions that are provided by extensions that aren't
available in the current gl context.
However, we can't query the extension strings until after we create an egl
context.
Split the regular symbol lookup stuff from the gl symbol lookup stuff so
we can query at the appropriate time.
These aren't a thing. the GL_OES_EGL_image_base extension doesn't exist
and the GL_OES_EGL_image extension doesn't extend egl, because it's a gl
extension.
So let's stop being the only google match for these nonsense strings.
Explicit lock / unlock of the shared mutex mixed with implicit
lock / unlock when accessing the internal Eo_Object data lead
to uncaught issues such as these.
This was found by trying to run E with gfx filters under eo_debug.
evas_object_image_data_get() is a legacy API that I made work
with snapshot objects (evas_object_image_snapshot_set()). Some
changes in the engine broke the behaviour and this patch fixes
it.
When getting the pixels from an FBO, in read-only mode, we need
to create a temporary image (pixels surface) that contains the
copy of the pixels we get from glReadPixels. This image needs
to be deleted afterwards. It is thus stored by the image object
and freed upon _image_data_set() (good) or object deletion (bad).
FBO + read-write is not supported by this API (it is supported
through buffer_map as the filters had to use that).
Fixes T5754
Summary:
There are four cases where items are deleted.
1. Select item and push back button.
2. Delete "X" part of item.
3. Calling elm_object_item_del().
4. MBE deleted.
Previous logic is can't support for all cases.
below problems were there.
eina_list_remove() called twice.
list data free twice.
@fix
Test Plan: elementary_test -> multibuttonentry test.
Reviewers: jpeg, cedric
Subscribers: cedric, jpeg
Differential Revision: https://phab.enlightenment.org/D5030
so emotion was using a much older (pre 1.0) buffer map mechanism for
getting video data. a new frame map api was added to gst in 1.0 but we
didnt use it. gst broke the old buffer direct mapping for some codecs
paths between 1.10 and 1.12. since we were using a very old method/api
this broke us. this also happens to fix some vaapi issues (except
where the buffer is simply never mappable ever). so this is kind of a
fix (updating us to as more modern api) and a workaround.
@fix
This exposes an elm_win sizing bug. To test (under X):
ELM_DISABLE_EXTERNAL_MENU=1 elementary_test -to "Main Menu"
The layout is not completely wrong 100% of the time, but it is
exactly correct 0% of the time. Window sizing is our flying
spaghetti monster.
On OSX the framespace and CSD (Client-Side Decorations) are not
supported at all... I am not able to test this case. This patch
restores the main menu functionality based on pre 1.19 themes,
where it was located inside win.edc (app content) and not in
border.edc (framespace).
Note that the initial size of a window may be wrong, eg as in
elementary_test -to "Main Menu"
Fixes T5734 (hopefully!)
Changes cursor handle name from 'Efl.Text.Cursor.Cursor_Data' to
'Efl.Text.Cursor.Cursor'.
Also, replace all usages of Efl_Canvas_Text_Cursor
with Efl_Text_Cursor_Cursor as the handle for the cursor.
Because there might be hidden C implements defined, we need to
always generate the appropriate class initializer. This does not
hurt anything as the contents would have been called automatically
by Eo anyway.
@fix T5736
Summary:
- when deleted callback is found _edje_signal_callback_move_last() is called
in order to pack match array.
- during _edje_signal_callback_move_last() index skips when another deleted
callback is found, but does not reduce members_count.
- this duplicates a remaining callback and calls the callback twice.
Test Plan:
1. add multiple edje_signal_callback by edje_object_signal_callback_add()
which have the same source, signal, func but different data.
2. delete first and last callback by
edje_object_signal_callback_del/edje_object_signal_callback_del_full.
3. emit edje_signal.
4. observe one callback is called twice.
Reviewers: SanghyeonLee, conr2d, jpeg
Subscribers: cedric, jpeg
Differential Revision: https://phab.enlightenment.org/D4985
The API eina_slstr_ is meant to look just like const
strings, wihch means here there could instead be an internal
struct holding the name, and the API would behave the same.
Note: This API is new in 1.20
This will help to focus on creating efl_ui_widget class work.
And, we need to change all the Elm_Widget_Item related logics to
factory or something else.
this fixes an issue that has cropped up in the past few months - only
nvidia drivers with egl/gles in x11... and compositing won't work
(native surface) and the introduction of libglvnd
it's a combination of libglvnd lying that it has symbols it can't
later find, new features to get core functions via procaddress that we
hadn't migrated to use AND use preferring core functions that libglvnd
will expose, so switching to KHR extensions by preference. we also
need to symmetrically use destroy image khr too...
oddly enough using procaddress purely for create/destroy image makes
wayland fail ... sofor now i'm taking advantage of the fact that
wayland has no extensions string passed in at the moment and still
doing dlsym... this is odd though.
@fix
the normal usage of these is something like
if (!strcmp(engine, my_engine))
win = window_get(ee);
which is a waste of effort since the window_get() functions all check
the engine interface internally