I think we're now at the point where the two paths are merged.
Still no atomic functionality because nothing assigned the primary plane,
so we have no atomic state to commit. The machinery should be in place
though.
We'll be doing tests as we build up plane state assignment. it's too late
to do anything about it if we fail here - failed tests will block plane
assignment in the first place so the scene graph knows it still has to
render those visual elements.
This will simplify a bunch of API that would otherwise have to pass in
both output and plane - and in some cases we might not have the output
handy anyway.
In cases where output monitors have different frequencies, we need to
be doing atomic commits on a per-output basis. This patch modifies the
ecore_drm2_fb_flip function to support doing atomic commits per output.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
As we need to do atomic commits on a per-output basis, these 2 newly
added API functions can go because these functions did one atomic
commit for all outputs
Signed-off-by: Chris Michael <cp.michael@samsung.com>
As there is nothing inside this function which requires any Atomic API
calls, this #ifdef can be removed and the function can then still be
used to assign Primary planes for non-atomic use cases.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
This commit fills in various output 'state' structures during creation
so that those state structures can be reused for pageflip handling
even if Atomic support is not enabled.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
This commit enables the ability to fill our state structures even if
atomic support is not enabled. This will allow us to reuse those state
structures for dealing with pageflip in both the atomic & non-atomic
use cases.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
As there is nothing 'atomic' specific in these structures, we can move
them outside the atomic ifdef and make use of them for handling
pageflip for both atomic and non-atomic use cases.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
There have been cases where the logic of _event_callback_call break'ed
too early in the event submission.
Reason for that was the line ((const unsigned char *) desc -
(const unsigned char *) it->desc) producing a overflow.
This means the if statement
if (!legacy_compare &&
((const unsigned char *) desc - (const unsigned char *) it->desc) < 0)
was true while the pointer desc was smaller than it->desc, which means
the event subscription got aborted, even if it should not.
This turned out on two 32 bit maschines. And led to not rendering apps
anymore.
It was introduced by commit in 605fec91ee.
@fix
Summary:
Cleans up grammar, simplifies wording, and elaborates on some details
for better clarity. Assume the reader will already have a basic
understanding of reference counting and 3D graphics in order to avoid
overexplaining these concepts.
Subscribers: cedric, jpeg
Differential Revision: https://phab.enlightenment.org/D4837
widget operations use NULL terminated strings, backend does not.
Refactor some selection code so all selection deletions trigger
undo events (except where they should not).
Summary: Added a new api to send the prediction hint string to IME.
Test Plan: Tested in Tizen device
Reviewers: woohyun, id213sin, jihoon
Reviewed By: jihoon
Subscribers: cedric, jsuya, z-wony, jpeg
Differential Revision: https://phab.enlightenment.org/D4805
The function parse_str returns allocated memory which should be freed
before we exit this function.
Fix Coverity CID1374644
Signed-off-by: Chris Michael <cp.michael@samsung.com>
The function parse_str returns allocated memory which should be freed
before we exit this function.
Fix Coverity CID1374647
Signed-off-by: Chris Michael <cp.michael@samsung.com>
This is needed by dmabuf engine fallback when it realizes it locally
allocated a buffer, has been rendering to it, but the compositor can't use
it.
So the engine copies its buffer contents into a new wl_shm buffer and
continues from there - however we need to make sure the async renderer
has finished first, so we don't copy a partial buffer.
This allows us to block for all previously submit actions in the render
queue to complete.
Summary:
when collecting the objects under a mouse pointer,
evas uses the geometry of an object to decide
whether the mouse pointer is inside the area of the object,
which is inappropriate for a mapped object.
so mapped objects don't receive mouse events when they should.
this patch fixes the issue.
Test Plan: A sample code will be added as comments
Reviewers: jpeg, raster
Subscribers: cedric, eunue
Differential Revision: https://phab.enlightenment.org/D4826
Previously the progressbar in fileselector use hardcoded style name
"wheel", that made unpossible to create different style for
fileselector. This commit made it possible.
@fix
Missing style name for sub cursor call the next error output:
ERR<24193>:elementary lib/elementary/els_cursor.c:734 elm_cursor_theme()
Could not apply the theme to the cursor style=(null)
Also this error call extra recalc for cursor hot spots.
Fixes T5408
@fix
Summary:
node_get() can return NULL sometimes. Add a missing check, and add dox
for the function to document the error return.
>>> CID 1374434: Null pointer dereferences (NULL_RETURNS)
>>> Assigning: "node" = null return value from "node_get".
@fix CID1374434
Subscribers: cedric, jpeg
Differential Revision: https://phab.enlightenment.org/D4824
Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
The following situation:
- A Box in a naviframe, with n children.
- All added children register to the focus graph with the box as parent,
order gets set correctly.
- Naviframe hides this item, so box property tree unfocusable gets set
to true, it gets unregistered from the focus graph, even every single
child gets unregistered.
- The item gets shown - every child and the table are getting
registered again.
- Order is not set again, since the box does not get changed
- Order of the children is mixed up.
This should fix this case since the order is flushed every time the box
gets registered.
There could be the case that the item gets freed due to some handling in
a event handler of the event EFL_UI_FOCUS_MANAGER_EVENT_FOCUSED.
So the code now sets the node to NULL after the event is called and
saves the fields that are rfom use later.
It's unlikely that we'll have other stuff under Ip namespace, also not
that likely to have other than Ip Addresses (to invert it to
Address.Ip), thus make a toplevel entry Ip_Address as suggested by
DaveMDS.
Test case:
elementary_test -to "Gesture Layer"
Just move the pictures around (eg. to the bottom). They could
disappear entirely.
This is because the geometry used was based on the smart
object "bounding box" rather than the mapped output.
@fix
This is the first step toward handling multi output. This patch
remove engine.data.output from Evas structure and use an Eina_List
for it instead. It also start moving code around to fetch an output
or an engine context (which are the same at the moment, but will be
split in a later patch).
Summary:
Fixes some grammar confusion for in that/this, that/which, to/at,
to/for, at/by, etc.
Subscribers: cedric, jpeg
Differential Revision: https://phab.enlightenment.org/D4806
Summary:
Rounding the sum of glyph's advance could cause inconsistency of
each glyph's positions. When Evas enables Harfbuzz library,
Each glyph's position has to be handled by only nearby glyphs.
But, currently, totally unrelated glyph's advacne could change
other glyphs positions.
ex) 1. "connect."
2. "Tap here to connect."
You can see different gap between "c" and "o" of word "connect".
It should be same even if there was a different text before the word "connect".
@fix
Test Plan: N/A
Reviewers: raster, herdsman, jpeg
Reviewed By: raster
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D4782
Lets say there is a box with the following ordered children:
|Button|Box|Button|Box|Button| the two boxes do not have any children
at the time of the setup. The logic of the order_update will only order
the children like that:
|Button|Button|Button| Which is correct by that time, the two boxes dont
have any children.
Now the two boxes are also getting children, the order will not
selfupdate or anything so the logical chain would be:
|Button|Button|Button|Box|Box|. Which is wrong. To solve that the
manager keeps the order that got set last, and reapplies the order again
if something gets added to the parent where the order was set.
This should fix strange next / prev operations in ephoto.
Putting the PAGE_FLIP_EVENT flag on the set rotation request resulted
in an extra event on the drm device fd that screwed up page flipping
badly from that point on.
@fix
since this code's creation it seems that the internal int size was set to use
short in order to micro-optimize memory usage, while the api function parameters
used Eina_Rectangle which had a larger int size. when initializing the internal
rect struct, this would lead to overflows which resulted in broken tilers which
returned iterators with no valid rects after having valid rects added
test case: run weston-subsurfaces
@fix
the api function requires this, but the unified handler for api+edje handler does
not, since edje singals are deferred and the button which triggered the move
may be released before the signal is processed
ref ea7bbfe47d
@fix
We don't need to keep this in eo files anymore because the APIs
using them are now fully in C. This also allows removal of the
event callback builtin from Eolian.
welm widget weak refered the logical parent slot but never unreffed
the weak ref - ever. this should fix that. in fact it does. one crash
less with:
elementary_test -to "icon standard"
@fix
I was reading it to understand this new focus system. So I also
made some cosmetic changes here and there in the file: wrap, fix
a few typos, add missing docs. I'm not pretending that the doc
is perfect now. This really was just a personal review.
In a few classes, this requires some manual expansion. This should
not break anything but it's also fairly ugly; a better solution
would be appreciated, for now we do this.
Similar changes will be done to a few other Efl.Object APIs as
well at later point.
Summary: This introduces a new focus system and the migration to this system.
Test Plan: run elementary_test
Reviewers: cedric
Differential Revision: https://phab.enlightenment.org/D4511
this just means the menu is registering a new manager, it is not really
doing anything, its just there so the children which are searching for
one are finiding one. Elm menu has not handled anything with keys in the
past and will later learn to do so.
The Efl.Ui.Focus.Manager abstracts the creation of a localization graph
and a logical tree. The localization graph is used to find a object
right left up or down of a given object. The logical tree is used to
iterate throuw the containers which are used to build a ui.
Those managers can be used bound to some layer in the ui, so for
example the window is a layer, the content of a scroller is a layer.
With those layers, we can make sure that movements of a scroller for
example just means that this graph of objects in the scroller needs to
be recalculated, and not the complete ui.
The advantage of having this to layer bound datastructures is that you
can easily debug those graphs, since the complete layer of this
managerobject can be calculated completly.
Currently, script block is removed when an edje group inherits from other group
after defining its own script block.
group { "somegroup";
script {
...
}
parts {
...
}
inherit: "othergroup"; // <= previous script block is removed here.
}
If parent group doesn't have script block, it doesn't need to overwrite previous
one. This will keep script block and print warning when script block is overwritten.
If we disable preload, then the second file set on an elm_image
object would not trigger a deletion of the first image. As a
consequence, both images would be visible... really bad if there's
alpha or different dimensions!
Thanks Anand Kumar for the report!
@fix
This fixes all warnings for "make examples" for:
-Wunused-parameter
-Wshadow
-Wformat-security
-Wenum-conversion
Some remaining warnings include:
-Wdeprecated-delcarations
Summary:
Applies the correction purely mechanically using the following shell
command-line:
src/examples/evas$ grep -sr 'fprintf(stdout' . | cut -d: -f1 | uniq | \
xargs sed -i "s/fprintf(stdout/printf(/"
This fixes a few warnings about lack of a format string:
warning: format not a string literal and no format arguments [-Wformat-security]
fprintf(stdout, commands);
Reviewers: cedric
Subscribers: cedric
Differential Revision: https://phab.enlightenment.org/D4691
This patch addresses an issue where plane formats were not being
properly copied into our Plane State structure and causing any usage
of our atomic code paths to crash and burn
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Since efl-jenkins does not have libinput >= 1.7.0, this commit broke
the jenkins build due to missing
LIBINPUT_POINTER_AXIS_SOURCE_WHEEL_TILT value. We should be able to
get by with letting 'default' case handle it....
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Summary:
When auto update is enabled, the label of the hoversel will be that of selected
item. This feature is usually used when changing state of something.
Highlighting item previous selected will show what is current state more
explicitly especailly hoversel has many items.
Test Plan: elementary_test -to hoversel
Reviewers: cedric, jpeg
Differential Revision: https://phab.enlightenment.org/D4799
Signed-off-by: Cedric BAIL <cedric@osg.samsung.com>
libinput >= 1.7.0 added support for
LIBINPUT_POINTER_AXIS_SOURCE_WHEEL_TILT, however we do not yet support
that in elput, thus gcc was spilling a warning about an unhandled
enumeration value here. We'll add a case for the above, plus a default
to cover any future additions.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Gcc complains that 'flags' here may be used uninitialized. In looking
at the code, 'flags' does not seem to be needed in the debug prints
here. If we keep and initialize the variable to 0 during declaration,
it would only ever print out 0 anyway as 'flags' is never changed in the
code.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Summary:
When table items are left aligned and mirrored is set, items should be placed
from the right side, but align is not changed. (still left-aligned)
@fix
Test Plan: make and run attached example
Reviewers: cedric, jpeg
Subscribers: thiepha, woohyun
Differential Revision: https://phab.enlightenment.org/D4758
This is similar to efl_super but the specified class is the one
we want to call the function on. This is similar to dynamic_cast<>
in C++.
Note: both efl_super() and efl_cast() need documentation!
This is an experimental feature.
Fixes T5311
@feature
Maniphest Tasks: T5311
Differential Revision: https://phab.enlightenment.org/D4797
I've always really disliked this lock. If someone calls a non-eo
function by accident with efl_super() then you'll most likely end
up in a deadlock.
This adds the cur_klass pointer to the object itself, exploiting
the fact that we have 8 bytes of padding (on 64 bits, at least).
Also, this opens the door to efl_cast() which would be similar to
efl_super() except that only a dynamic cast is done, not a call
to the parent function.
make benchmark shows a performance improvement, surprisingly.
This is a bit experimental. See also the following commit (efl_cast)
See a76ebea2d8 and the following
commits on this file.
The following test scenario let to easily reproducible
application hangs:
elementary_test -to "Icon Desktops"
# then scroll vigorously with the mouse wheel up/down
This patch was applied as a new revision on the below diff:
Differential Revision: https://phab.enlightenment.org/D4754
This patch adds 2 new API functions, one which we can use to test atomic
commits before actually applying them, and another which does the
actual Atomic commit.
@feature
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Small commit to symlink to drmModeAtomicMerge function so we can use
that for atomic commit tests.
@feature
Signed-off-by: Chris Michael <cp.michael@samsung.com>
When we do an atomic commit, we need to know where to place a given
plane (in the case of overlays) in relation to the CRTC, so provide an
API function that can be used for that purpose.
@feature
Signed-off-by: Chris Michael <cp.michael@samsung.com>
As we will need the plane state source values when we do an atomic
commit, we can store them when plane_assign is called as we already
have the FBO available.
Signed-off-by: Chris Michael <cp.michael@samsung.com>
This patch adds a new file where we can store any additional functions
we may need to work with hardware planes. Currently the file contains
a public function that can be used to assign a given Ecore_Drm2_Fb to
a hardware plane
@feature
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Small patch to make sure we free memory previously allocated for
hardware planes when we destroy an output
Signed-off-by: Chris Michael <cp.michael@samsung.com>
As we may need these defines in other files, move them to the private
header so there is access to them.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
As we will need these values later to determine if an FBO can go onto
the cursor plane, we should store this in the device structure to
avoid having to refetch them later.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Various hardware can support multiple planes on a given output. As
such, we need to be able to store multiple plane states per-output.
This small patch adds support for that.
@feature
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Small patch to store supported formats on a given plane state. This
will be used for assigning dmabuf clients to a hardware plane based on
size and supported format.
@feature
Signed-off-by: Chris Michael <cp.michael@samsung.com>
As we are refactoring the usage of hardware planes and atomic commits,
we need to remove the old usage of atomic flipping for ecore_drm2_fb
because atomic flipping will be handled differently.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
While having the ability to test for specific driver and kernel
versions is nice to ensure that Atomic is supported, it quickly can
get out of hand trying to maintain this whitelist so (for now) disable
it and rely on the kernel results from drmSetCap.
@fix
Signed-off-by: Chris Michael <cp.michael@samsung.com>
Summary:
This uses Eina_Cow to implement support for rarely used features
in EO objects. This covers:
- composite objects (eg. UI widgets: combobox, text, video, win)
- vtable for efl_object_override
- del_intercept
All of these features are quite tricky to get right and while
very useful, should still be used with great care. With this patch,
the size of an _Eo_Object struct comes down from 80 bytes (rounded
up from 72b) to 64 bytes (rounded up from 56b) on 64 bits.
Also I haven't measured precisely but I don't expect any performance
impact since the COW data is more likely to remain in L1/L2 cache,
as the default one will be used most often. Unfortunately, the
results of "make benchmark" have been quite inconsistent over
multiple runs.
This saves ~64kb in elementary_test (>4k objects) at the cost of
~100 calls to COW write (del intercept on some events).
@optimization
Reviewers: raster, cedric
Differential Revision: https://phab.enlightenment.org/D4796
Summary:
Modern, standards-compliant compilers already test p, so as per the C
spec it is superfluous to do so before the call.
Reviewers: jpeg
Reviewed By: jpeg
Subscribers: cedric, jpeg
Differential Revision: https://phab.enlightenment.org/D4791
it seems coverity didn't like our checks like if end - start > 0xffff
then dont do anything. this should effectively stop any issues but
seemingly not, so try another way to keep coverity happy.
CID 1361219
it seems coverity didn't like our checks like if end - start > 0xffff
then dont do anything. this should effectively stop any issues but
seemingly not, so try another way to keep coverity happy.
CID 1361220
Summary:
part.clip_to statement has short handler named "part.clip"
while part.description.clip_to does not.
Its short handler name should be "part.description.clip"
in order to have same experience with "part.clip".
Reviewers: cedric, Hermet, jpeg, conr2d
Reviewed By: conr2d
Subscribers: conr2d
Differential Revision: https://phab.enlightenment.org/D4789
Summary:
Applications want to know the current keyboard mode state to handle application's size manually.
So added a new ecore_imf_context_keyboard_mode_get API and input_panel_event_callback type for
keyboard mode.
Test Plan: Tested in Tizen device
Reviewers: jihoon, woohyun, id213sin
Reviewed By: jihoon
Subscribers: cedric, jsuya, z-wony, jpeg
Differential Revision: https://phab.enlightenment.org/D4790
Elm_fileselector has the hardcoded value for calculate the item size
with thumbail, this size was 16, it's looks like thumbnail size plus
labal text height, but hardcoded value haven't effect to scale. Other
problem with items without labels. As a result we have correct
thumbnail size only with default theme and with scale 1.0.
This commit made the item size accordingly to size what user set. It's
made more clearly this API behavior.
@fix
By some reason style does not applyed to genlist/gengrid in
fileselector. Also fixed issue with applyed style for files view on
change mode. Now this problem is fixed.
@fix
After talking with @eunue I realised that the way I'd first
implemented the box/grid "pack" API was simply too complicated.
I had tried to make it possible to change the layout function
at runtime, like good old evas box, but since there are no function
pointers in EO the final design was really convoluted.
If someone really needs to change the layout of a box at runtime,
just create your own subclass, or unpack all items and repack them
in a new box.
Note: there are still some issues with the layout params & flow
If efl_object_override() is called with a function that does
not exist in the original class, it may lead to a crash on
indexing an non-existing array in the vtable.
This is really just a safety check, as the usage was wrong:
* You are only allowed to override functions that are defined in the
* class or any of its interfaces (that is, efl_isa returning true).
Honestly I can't see why gfx & gfx.path "changed" need a manual
definition, instead of relying on EO. If the API needs to be
internal only, then EO needs to handle internal APIs. In this
case, the event was exposed as a C API but not a EO... why?
Summary:
The values(259200, 86400) are hard to know the meaning.
And we don't have to call gmtime() in for loop.
Test Plan: elementary_test -> calendar
Reviewers: jpeg, Hermet, shilpasingh, cedric
Reviewed By: cedric
Subscribers: cedric, jpeg
Differential Revision: https://phab.enlightenment.org/D4717
ecore_wl2 contains a bunch of code for compositor side seat handling.
There's really no need for a client to do this, and E does the
compositor side seat stuff internally, so this code will never
be used.
This removes the unused code.
Summary:
When the keyboard mode state is changed, the keyboard_mode_event_cb will be called, too.
But there is no way to get keyboard mode manually.
Test Plan: Tested in Tizen device
Reviewers: jihoon, woohyun, id213sin
Reviewed By: jihoon
Subscribers: jpeg, z-wony, jsuya, cedric
Differential Revision: https://phab.enlightenment.org/D4786
so our yuv import funcs for gstreamer 1.x engine were ignoring the
plane offsets and strides provided by gstreamer. though this nicely
shows that these numbers provided are actually wrong - at least in the
testing with vaapi back-ends with gst.
so this fixes emotions' badness but there is still badness in gst
apparently. the numbers provided if used are just simply wrong for teh
image data. commented code in the src to show how to "Fix it up" by
forcing some alignment of content to get it to work.
@fix
This might not be used as over two consecutive runs all the
same buffers should be used. But it could happen if some
parameters in the filter change (eg. blur radius).
Fixes major (GPU) memory leaks. Reuse mode is still leaking.
An odd-sized image scaled down by 2 was losing 1 pixel during the
downscale, and it was not restored after scaling up. The same
happened with downscaling by 4 except the effect was even more
visible.
This meant that a moving snapshot with a large blur would trigger
some really ugly sampling issues if the content below was precise
(such a text).
If the obscured area in a snapshot object changes a lot, do
not try to keep track of it forever. Instead, redraw the filter
over the entire object region, without obscure.
This fixes a performance issue when an opaque window is moved
above a fixed transparent window (the latter has a snapshot with
blur filter).
This dramatically improves the performance and now seems
to give acceptable results. Eventually we need a quality flag
in order to enable this or not. Alternatively, "gaussian" blur
mode would skip this optimization, while "default" would trigger
it.
When the filtered object is an image, without borders, map,
fill info or anything of this sort, then the filter input
buffer is really just a copy of the original image. We can
skip that to save on memory usage and pixel fetches.
This can help with performance when a large region of the
filtered image (eg. snapshot) is fully hidden by an opaque
object. For instance the window border is hidden by the
opaque window content.
This improves over the previous code for handling
snapshot objects and cutouts. Basically any opaque object
above a snapshot should be obscuring it. That is true
unless a crazy filter is applied, or the snapshot object
is itself the source of a map or proxy.
This also uses eina_tiler instead of a custom (and really
bad) algorithm to compute the obscure regions.
This make save() work on snapshot objects, provided the call
is done from inside render_post.
Also, this saves the filtered output of an image, rather than
its source pixels. Any call to save() on a filtered image must
be done from post-render as well.
Fixes T2102
@feature
If we delete the image that was the target surface for gl
rendering, a crash would occur on the next render cycle.
Unlikely but not impossible to trigger from app side.
@fix
This was a poor attempt at improving the performance but
obviously the root cause isn't fixed (too many texel fetches).
Uniform should (theoretically) work better than an attribute
the for loop. Just a guess here.
This also makes GL blur use a float value as radius, allowing
future extension to non-integer blur radii, as well as using
linear scaling as a fast blur approximation.
This factorizes some of the common code for image render
and resolving is_inside (verifying alpha value of a pixel).
This should also be used by save(), as well as buffer_map().
This patch introduces lots of whitespace changes by using return
instead of long if() {} or else {} blocks.
The situation is clearly visible in the Snapshot test case:
increase the radius and a red glow would appear. This is because
the snapshot object was not marked as needing redraw and so had
no pixels under the opaque rectangle.
This will reuse existing buffers by resetting only the minimum
required in the filter context (also reused). Work in progress,
as the actual reuse is disabled for now.
This avoids creating one more FBO and doing one more draw,
by rendering the image input data directly into the input
buffer. This also makes the code common between SW and GL.
This optimizes the GL blur algorithm by reducing the number of
texel fetches (roughly half the number of before this patch). This
works by exploiting GL's interpolation capabilities.
This skips extra tests with image objects that have the
is_opaque() function. That way, if an object is marked as COPY,
rendering of all objects below it will be skipped.
This can dramatically help with performance when flagging a
snapshot object as COPY. This should not be done if a filter is
applied and is meant to blend with the underlying UI.
When using a snapshot object we have access to exactly all
the pixels below it inside the snapshot surface. So, in order
to produce a nice blur, it is necessary to expand this snapshot
and then clip it otherwise the edges will look a bit ugly.
Unfortunately, there are still places where blurs didn't look
so good, as objects below an opaque region would never get
rendered into the snapshot object. So the edges, inside a
snapshot object, around an opaque region would have blur
artifacts.
This fixes that by shrinking the cutout regions by the radius
of the filter. Eg for blur this is the blur radius.
The test case in elm_test can exhibit this fix very clearly:
a red glow would be visible around the opaque rectangle, but with
these changes we instead see the blurry edges of the objects
below the rectangle.
This will be most useful in a special case, where a filter is
used in a window decoration, applied to a snapshot object.
Another optimization that might be wanted is passing a list
of update regions (from the proxy or snapshot).
The filters don't support the obscuring region yet, only some
of the high-level logic is implemented.
If anything in the canvas needs redraw and a snapshot object
happens to intersect with the update region then it was redrawn,
even if all objects below it hadn't changed. This has an insane
performance impact when you apply a blur filter on the snapshot
object. Walking the object list will always be cheaper than
rendering the snapshot!
Note: Added a FIXME comment and forced clean_them to be true
because some odd behaviour happens when breaking with GDB and
the array snapshot_objects keeps growing at each frame (I guess
only if we miss a frame or something like that).
By simply splitting X and Y blurs in two passes we can improve
the performance of the blur filter a lot.
There is still much to be done to make it really fast and nice
looking:
- implement true gaussian blur (not sine-based approximation,
right now the actual blurs look different in SW and GL)
- exploit linear interpolation for R tap instead of R*2+1 taps
(a tap being a texel fetch)
- downscale & upscale large images with large blur radii
Wait a second though, this implementation is not only incomplete
(no support for box vs. gaussian blur), it's also insanely bad in
terms of performance. Small radii may work fine, but at least blurs
render properly in GL with this patch (no more glReadPixels!).
The shader needs a lot of love, including in particular:
- support for 1D box blur single pass
- support for 1D gaussian (or sine) blur
- use linear interpolation and N-tap filters
- separation of 2D blur in two passes (high-level logic)
- potentially separation of large 1D blurs in 2 or more passes
knowing that 2sigma == sigma + sigma when it comes to the gaussian
bell curve.
This one was a bit more... "fun". I had to add a new vertex
attribute and obviously using a VertexAttribPointer led to
incomprehensible crashes. But a simple glVertexAttrib2fv makes
it work like a charm!
A rare option is not handled yet.
This reuses the existing mask infrastructure, but adds a color
flag to use the whole RGBA range, rather than just the Alpha
channel.
Filters are still very slow (glReadPixels and non-optimized use of
GL buffers...), but this is progress :)
This corrects two things:
- the blur filter high-level logic, that lead to reusing some
temporary buffers which contained garbage;
- the versatile gl buffer implementation so that it now properly
switches between the RGBA_Image and the FBO content (yes, this
is insanely slow and inefficient... but it works and that was
the only point).
Alright, so this is a massive patch that is the result of
trying to get rid of unused or poorly implemented classes in
ector. Originally ector was meant to support VG but extend to
things like filters as well. At the moment, ector's design
makes it quite hard to plug in the filters.
For now I think it's easier to implement the GL support for
the filters directly in the engine, where I hope to interfere
as little as possible.
This massive patch keeps only the required minimum to support
a versatile gl buffer that can be mapped, drawn or rendered to (FBO).
It's extremely inefficient as it relies on glReadPixels and lots
of texture uploads, as well as conversions between ARGB and Alpha.
Another type of GL buffer is a wrap around an existing GL image,
but that one is read-only (map or draw: no write map, no FBO).
No, all the filters run fine, and the high-level implementation
(evas_filters.c) does not need to know whether the underlying engine
is SW or GL. One problem though appears with the blending or blurring
of some Alpha buffers, the colors are wrong.
This patch removes more lines than it adds so it must be good ;)