| Commit message (Collapse) | Author | Age | Files | Lines |
... | |
|\| |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Conflicts:
.qmake.conf
src/qml/qml/qqmlengine.cpp
src/quick/items/qquickitemsmodule.cpp
tools/qml/main.cpp
Change-Id: Ida8daf6b4d7e675385f2f5514c446e52dedaf136
|
| |\ \
| | | |
| | | |
| | | | |
Change-Id: I4020a1b3c59dea18faf7cbcbb78b90fcfc3680f0
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Similarly to QOpenGLWidget, functions like QOpenGLFramebufferObject::bindDefault()
should bind the QQuickWidget's FBO, not 0, while rendering the scene graph.
This becomes particularly important on platforms with surfaceless context support.
Here offscreen surfaces are not backed by any surface. Therefore any OpenGL operation
accessing the current draw framebuffer with FBO 0 bound may potentially crash, as there
is no draw framebuffer at all.
The distance field glyph cache exhibits this issue when running with EGL on Mesa:
glViewport crashes when we render via QQuickWidget and the current framebuffer is reset
to 0. The problem goes away when the code changed is to use bindDefault() - as it should have
anyhow - and QQuickWidget is enhanced to communicate the "default" framebuffer to QOpenGLContext,
just like QOpenGLWidget does.
Task-number: QTBUG-46415
Task-number: QTBUG-43269
Change-Id: I35fe375a0870dadecc4a074dfdec122c6a4c92ab
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
Reviewed-by: Paul Olav Tvete <paul.tvete@theqtcompany.com>
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Once we hit removedFromAtlas() and the glCopyTexImage2D() path, bad things
tend to happen with OpenGL ES, both on certain embedded devices (Beaglebone, RPi)
and ANGLE.
ANGLE just rejects GL_BGRA_EXT with INVALID_ENUM. So if it fails, just try
with GL_RGBA. The BGRA extensions do not mention glCopyTexImage2D in any form
and in plain GLES (any version) BGRA does not exist. So rejecting it may be valid,
depending on how one reads the specs.
Same problem on Beaglebone, where the call with BGRA fails as INVALID_OPERATION.
The RPi 1 and 2 fails in a different way: the temporary framebuffer is not complete
because BGRA textures are not supported as color attachments. So our only choice
here is to do what we do for some Android devices already: ignore BGRA support.
Task-number: QTBUG-46806
Change-Id: I89b3b38bf7f8883c39509606ec5ae525f131292b
Reviewed-by: aavit <eirik.aavitsland@theqtcompany.com>
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
We had been accidentally using the 8-bit alpha glyph shader, which
resulted in rendering solid black outlines of the glyphs.
Task-number: QTBUG-37986
Task-number: QTBUG-45514
Change-Id: Ie8728d50068f58d659e24bc5db2d73e01ca8db75
Reviewed-by: Konstantin Ritt <ritt.ks@gmail.com>
Reviewed-by: Tor Arne Vestbø <tor.arne.vestbo@theqtcompany.com>
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Change the gradient position calculations to consider full border, not
just half of it, because the border is fully drawn inside the
rectangle, not half outside, half inside.
Task-number: QTBUG-36059
Change-Id: Ie7dec94d9f6e03ae07c612c6e3b1950781973dc7
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Just like commit fd565ec6, this fixes another vertex shader compilation
issue on certain OpenGL drivers.
Change-Id: Ided3d6082fa0790659e9c242d0b43d7de5a2ae4d
Reviewed-by: Eskil Abrahamsen Blomfeldt <eskil.abrahamsen-blomfeldt@theqtcompany.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
* Put the cheap checks (pointer ones) first
* Don't call currentThread twice as it involves a mutex lock/unlock
Change-Id: I5df92ea99b034301113e9bd34f8a66df1a7e7dfe
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
No reason to iterate through all items to create a temporary
QList we don't need.
Change-Id: I3c5388f026ded0290afb9dd35fc7934011c4998d
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
Reviewed-by: Olivier Goffart (Woboq GmbH) <ogoffart@woboq.com>
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Opaque textures can be a lot faster, so give this option without
forcing the user to reimplement her/his own QSGTexture class.
The old behavior was that createTextureFromImage() disregarded
TextureHasAlphaChannel and looked solely at the image's format. To
keep this behavior intact, we introduce a second opt-in flag to switch
textures from auto-detect to false, namely TextureIsOpaque.
[ChangeLog][QtQuick][QQuickWindow] Add TextureIsOpaque option to
createTextureFromImage()
Change-Id: I248db4bc5f7920864b6aa8d831ce24d23ad370ef
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
No known users of this interface exist, and not having it helps to make things
less complicated.
Change-Id: I3d749cfbde9e84e1c7b8dc5cbd952c5c51d347ee
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | | |
Change-Id: Id4820ac458f48b10f2bf457144767efdef9e2c07
Reviewed-by: Marc Mutz <marc.mutz@kdab.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This adds a few more text items/frame on my macbook, and more or less halves the
amount of transient 608 byte allocations (42mb -> 25mb).
Change-Id: Ib1a67c8203f777a2f3b790ff3597d1af34eed2dd
Done-with: Gunnar Sletta
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Performing a full scan of the pages each time a new node is to be allocated is
wasteful. Instead, store the last known free page, and search from that position
forwards. When an item is freed, reset the last known free page to search
forward from that position. If no free page is found, don't fall back to a full
scan - just allocate a new page.
The creation of lots of new items is a pretty common pattern (PageStack on
mobile and similar), thus it makes sense to optimize for it and avoid scanning
where possible.
This takes the time to instantiate 142k Rectangles in a single
component from 2433ms on the render thread to 495ms on the render thread (down
slightly from 511ms when using no paging at all, and down from 638ms on 5.3.1).
Separately to this, it may make sense to increase the size of pages, but
ideally, this will wait until there is a shared pool between renderers.
Change-Id: I2b335e230bc28fffcbb38b3fb43f3d441789c256
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | | |
Change-Id: Iab7c9949941559b4773e5d4a0406dc58cfc70adb
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
|
|\| |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Conflicts:
src/qml/jsruntime/qv4engine_p.h
src/quick/items/qquickitemsmodule.cpp
src/quick/items/qquicktext.cpp
src/quick/util/qquickpixmapcache.cpp
tests/auto/quick/qquickwindow/tst_qquickwindow.cpp
Change-Id: I90ecaad6a4bfaa4f36149a7463f4d7141f4a516a
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
GLX, and potentially other platforms too, require that compatible
contexts and surfaces are created with the same screen.
This makes Quick functional on X.org configurations where each
physical screen represents a separate X screen on the same X display.
Change-Id: I03c87819e1aee148c880a3caca13d4a0c5409056
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Do the screen check only for toplevel windows and make warning
more verbose.
Change-Id: Ie25271fe34c72c1410022afe8a1a9bdccdf7a12f
Task-number: QTCREATORBUG-10814
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Found with GCC's -Wcast-qual.
Change-Id: Ia0aac2f09e9245339951ffff13c9589afabc7ade
Reviewed-by: Alan Alpert (Personal) <416365416c@gmail.com>
Reviewed-by: Simon Hausmann <simon.hausmann@theqtcompany.com>
|
| | |
| | |
| | |
| | |
| | | |
Change-Id: I91169e719ac97b8e32805e08d6a1f43caa897f73
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Windows render loop already had this capability.
Change-Id: I9391e218568cb6951f9f25d74e732aa5ddf9b5bc
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This hurts interoperability with OpenCL on windows and the other
renderloops don't do it either, so to be consistent, lets not.
Change-Id: Ie06842957561c8933690633469b8b3a026ff399b
Task-number: QTBUG-46273
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
Reviewed-by: Friedemann Kleint <Friedemann.Kleint@theqtcompany.com>
|
|\| |
| | |
| | |
| | |
| | |
| | |
| | | |
Conflicts:
tests/auto/quick/qquicktext/tst_qquicktext.cpp
Change-Id: I075e742da8396a268d97d3ab34bcd9e0c0cf001f
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
There were multiple problems in the implementation of the shader cache:
1) it was not thread-safe
2) nothing was ever removed from the hash
3) since the keys into the hash are the actual shader source code, problem
#2 would lead to serious memory consumption over time
Change-Id: I20d1fb2074932e23f89edddba12e68ab8adcbff0
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | | |
Change-Id: Ic7ffb48c1a148036bdb14a912d3407e9a1b3b4fd
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
|
|\| |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Conflicts:
src/qml/qml/qqmlbinding.cpp
src/qml/jsruntime/qv4arraybuffer.cpp
src/qml/jsruntime/qv4functionobject.cpp
Change-Id: Ic752e9dfd69b282093651c9234c110a49762f06d
|
| | |
| | |
| | |
| | |
| | | |
Change-Id: I5df32609b45047666ea8bde500cf771736c396e5
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
|
| | |
| | |
| | |
| | |
| | |
| | | |
Task-number: QTBUG-45358
Change-Id: I4af0da937253f3ae578333c659dcbb132312a9cc
Reviewed-by: Laszlo Agocs <laszlo.agocs@theqtcompany.com>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When there is no render pending, renderWindow() is not called and so
there is no blocking for vsync. Advancing the animations without any
sleep is wrong in this case and results in animations running at 3x
speed in case there is no change in the scenegraph contents.
To solve this, import the threaded render loop's sleep call.
Task-number: QTBUG-42699
Change-Id: I3658a827af12d4bc2ac05c7b0b29c65e9fc99675
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
NoStage allows scheduling jobs for immediate execution on the
renderer thread.
[ChangeLog][QtQuick][QQuickWindow] Added a render job stage: NoStage
Task-number: QTBUG-44953
Change-Id: I918c79f1d095bc27d911a88d81376d146a04313c
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
Reviewed-by: Venugopal Shivashankar <venugopal.shivashankar@digia.com>
Reviewed-by: Pasi Keränen <pasi.keranen@digia.com>
|
|\| |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Conflicts:
src/quick/util/qquickpixmapcache.cpp
tests/auto/quick/qquickwindow/BLACKLIST
Change-Id: Ie81612f2884f8ea508c48ba2735ec54ea1c2eca5
|
| |\ \ |
|
| | |\|
| | | |
| | | |
| | | | |
Change-Id: I54e66e992f5e5d441b8b5394e7a03ec5352e7bf3
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
This is a follow-up to 657766f19b31d83f361fd3c9ad717b207e1dff8c which
removed it from the destructor itself. This goes one step further and
removes the calls from the windowDestroyed() calls which also sends the
deferred deletes.
Task-number: QTBUG-40920
Change-Id: I491b79bb600914575ba5565d2862d041726217e8
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| |/ /
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
After installing the Android 4.4.2 upgrade on Samsung Galaxy Tab 3 10.1,
one of the vertex shaders in Qt Quick stopped compiling.
The issue seems to be that this particular shader triggers a
bug in the driver, and a simple work-around is to replace the
ternary conditional operator with a simple if-statement.
[ChangeLog][Android] Fixed a vertex shader compilation issue on
certain OpenGL drivers.
Change-Id: I2f878215a753a7e222c46c0f85a84b0bc81d523a
Task-number: QTBUG-43515
Reviewed-by: Andy Nichols <andy.nichols@theqtcompany.com>
|
| | |
| | |
| | |
| | |
| | |
| | | |
Task-number: QTBUG-45076
Change-Id: I03de20da2e3f436339dcf48fc7d7d3d11a509577
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Just like commit 310fd3ed, this fixes another unnecessary detach.
Change-Id: I5108ea5112a4b0e877f86d06868375abb9a80409
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Fixes Coverity CID 10732, 10734, 10735, 10736, 10739, 10740
Change-Id: Idebb0bdaf42ceb06cd86c73592138e758a8a87c9
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
|/ /
| |
| |
| |
| |
| |
| |
| |
| | |
Adding a QSGRenderNode to the scene permanently disabled opaque batches
by disabling depth buffer usage. Reset the depth buffer usage to the
default value once last QSGRenderNode has been removed from the scene.
Change-Id: I760afde83ae9eaaf1b5571c37fd0081eb23b1f20
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| |
| |
| |
| |
| | |
Change-Id: I3f9219dd2fed15094c2f7d670a981406e601959b
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
|\|
| |
| |
| |
| |
| |
| | |
Conflicts:
src/qml/jsruntime/qv4numberobject.cpp
Change-Id: I4e66a03ef4d99cec192c9da30c028fd8c1f4ac0d
|
| |
| |
| |
| |
| |
| |
| |
| | |
It appears to be possible for node->element()->batch to be NULL
Task-number: QTBUG-43129
Change-Id: If6e4e265a02ee305bf3aa9cad387b7a73648367a
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| |
| |
| |
| |
| |
| |
| | |
Add destructor for QSGThreadedRenderLoop to destory QSGContext object.
Change-Id: I479947eb35a5d7888d45655a9b8e2ece5e30bc33
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
This shouldn't be a static decision as the glyph count depends on the
font.
Change-Id: I67ea98595505df4a3a30b16e867e76f89b28c1ef
Task-number: QTBUG-29264
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
Reviewed-by: Robin Burchell <robin.burchell@viroteck.net>
|
| |
| |
| |
| |
| |
| |
| | |
Avoid repeated instantiation of end() in loops, use variable instead.
Change-Id: I6ab1fe2b82406d5ee91710a0333587ffb82c04d4
Reviewed-by: Erik Verbruggen <erik.verbruggen@theqtcompany.com>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Reduce the number of times we construct a painter path from a glyph.
Change-Id: Ic1f32c78ab5dfd3b75d5442da163cda1dedab3ee
Task-number: QTBUG-42853
Reviewed-by: Lars Knoll <lars.knoll@digia.com>
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
On embedded hardware, texture resizes can be quite slow. The glyph
cache currently often needs to resize when new glyphs are added, and
glyph additions are very frequent when using a language with a large
set of glyphs (such as Chinese).
The current glyph cache design minimizes the amount of memory used.
When QSG_PREFER_FULLSIZE_GLYPHCACHE_TEXTURES is set, and we are using
a font with a large number of glyphs, we can instead allocate max-sized
textures. This leads to significantly less time when inserting glyphs
into the cache (often incurred over a span of time), at the cost of
higher initial memory and creation times (often incurred at
application startup).
Change-Id: Id1021b9d213e5f8635c4197b624474f28c6f44ff
Task-number: QTBUG-29264
Reviewed-by: Robin Burchell <robin.burchell@viroteck.net>
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
Reviewed-by: Yoann Lopes <yoann.lopes@theqtcompany.com>
|
|\|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Conflicts:
.qmake.conf
LICENSE.GPLv2
examples/qml/networkaccessmanagerfactory/view.qml
src/qml/jsruntime/qv4runtime.cpp
src/qml/jsruntime/qv4stringobject.cpp
Change-Id: I5d12f436d60995e51d5c2f59d364e9cbc24f8e32
|
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Native text rendering results in very different output on OS X
with QQuickWidget/QOpenGLWidget. When targeting an FBO, enabling
sRGB is futile if the FBO was not created with the correct format.
In this case we need to disable the usage of sRGB.
Task-number: QTBUG-42861
Change-Id: I887482e70be2cbfba40d6758546128d03b36def1
Reviewed-by: Eike Ziller <eike.ziller@theqtcompany.com>
Reviewed-by: Morten Johan Sørvig <morten.sorvig@digia.com>
Reviewed-by: Jørgen Lind <jorgen.lind@theqtcompany.com>
|
| |
| |
| |
| |
| |
| |
| | |
Task-number: QTBUG-43847
Task-number: QTBUG-40789
Change-Id: Iceacaa49bafffb31752a9fb26c896df570153fec
Reviewed-by: Gunnar Sletta <gunnar@sletta.org>
|