The DatabasePager now passes the Terrain pointer into the ReaderWriter's via osgDB::Options object,
rather than pushing a NodePath containing the Terrain onto NodeVisitor. This
change means that the DatabasePager nolonger needs to observer the whole NodePath and
will be lighter and quicker for it.
The change also means that ReadFileCallback can now run custom NodeVisitor's on the scene graph without
having to worry about TerrainTile's constructing scene graphs prior to the Terrain being assigned.
Also changed is the NodeVisitor::DatabaseRequestHandler which now requires a NodePath to the node that you wish
to add to rather than just the pointer to the node you wish to add to. This is more robust when handling scenes
with multiple parental paths, whereas previously errors could have occurred due to the default of picking the first
available parental path. This change means that subclasses of DatabasePager will need to be updated to use this new
function entry point.
http://forum.openscenegraph.org/viewtopic.php?t=7285
This file adds the same string API wrapper to the State object for other older apps that
track Uniforms by string. The original comment about performance is preserved."
are big enough to handle the primitives used in the geometry.
Added usage of verifyArray into osgfilecache so that it reports any problems on reading files in paged database.
* the glsl plugin now supports processing #includes. The file extension sets the shader type.
* the registry releases gl objects of the shared state manager
"
I see that we should expect some performance penalty for using this method. It won’t be painful in my current case because I have only a few animated characters. But I suspect some day I will have to fix osgCal to use int UniformIds natively for larger crowds."
hpux. I have skipped irix this time as irix is too dead to keep osg building
there.
As usual, solaris does not like member templates in stl containers.
Some headers missing and link problems due to missing libraries."
A few things remain to do:
* The binding between a uniform block in a shader program and a buffer indexed target number is fixed, like a vertex attribute binding. This is too restrictive because that binding can be changed without relinking the program. This mapping should be done by name in the same way that uniform values are handled i.e., like a pseudo state attribute;
* There's no direct way yet to query for the offset of uniforms in uniform block, so only the std140 layout is really usable. A helper class that implemented the std140 rules would be quite helpful for setting up uniform blocks without having to link a program first;
* There's no direct support for querying parameters such as the maximum block length, minimum offset alignment, etc. Having that information available outside of the draw thread would make certain instancing techniques easier to implement."
attached you'll find the second part of the IOS-submission. It contains
* GraphicsWindowIOS, which supports external and "retina" displays,
multisample-buffers (for IOS > 4.0) and multi-touch-events
* an ios-specific implementation of the imageio-plugin
* an iphone-viewer example
* cMake support for creating a xcode-project
* an updated ReadMe-file describing the necessary steps to get a
working xcode-project-file from CMake
Please credit Thomas Hogarth and Stephan Huber for these changes.
This brings the ios-support in line with the git-fork on github. It
needs some more testing and some more love, the cmake-process is still a
little complicated.
You'll need a special version of the freetype lib compiled for IOS,
there's one bundled in the OpenFrameworks-distribution, which can be used."
Notes, from Robert Osfield, modified CMakeLists.txt files so that the IOS specific paths are within IF(APPLE) blocks.
* support for NPOT-textures on IOS
* support for FBOs (only renderToTexture for now) on IOS (should work
for other OpenGL ES 1/2 targets, too)
* FileUtils-support for IOS"
2: minor tweak for a DebugHUD drawn improperly case when multiple slave views shared one window. It now uses slave view viewport to correctly position DebugHUD.
3: deactivated ConvexPolyhedron notifications (they were accidentaly activated when you replaced osg::notify calls with OSG_NOTIFY macro). These warnings are useful only for shadow map developer working on shadow volume optimizations. So there is no sense in having them active all the time."
osgGA. My approach is to bundle all touchpoints into one custom data
structure which is attached to an GUIEventAdapter.
The current approach simulates a moving mouse for the first touch-point,
so basic manipulators do work, sort of.
I created a MultiTouchTrackballManipulator-class, one touch-point does
rotate the view, two touch-points pan and zoom the view as known from
the iphone or other similar multi-touch-devices. A double-tap (similar
to a double-click) resets the manipulator to its home-position.
The multi-touch-trackball-implementation is not the best, see it as a
first starting point. (there's a demo-video at http://vimeo.com/15017377 )"
serialization libraries. My submission mainly includes:
1. Add two new macros USE_DOTOSGWRAPPER_LIBRARY and
USE_SERIALIZER_WRAPPER_LIBRARY. Applications using static OSG must
include corresponding static-link libraries and use these two macros
to predefine native format wrappers. Please see osgstaticviewer and
present3D in the attachment for details.
2. Add a LibraryWrapper.cpp file in each
osgWrappers/deprecated-dotosg/... and osgWrappers/serializers/...
subfolder, which calls all USE_...WRAPPERS macros inside. The
LibraryWrapper file is automatically generated by the
wrapper_includer.cpp (with some slight fixes), which is also attached
for your reference. The deprecated-dotosg/osgAnimation is not included
because it doesn't us REGISTER_DOTOSGWRAPPER to define its wrappers.
3. Modify the ReaderWriterOSG.cpp to prevent calling loadWrappers()
when static build.
4. An uncorrelated fix to Serializer and ObjectWrapper.cpp, which
ensures version variables of serialziers are initialized, and
serializers out-of-version are not written to model files.
"
Initial email from Tim : "I've implemented using a timestamp, available with ARB_timer_query and OpenGL 3.3, to gather GPU stats. This is nice because it can accurately fix the GPU draw time with respect to the other times on the stats graph, rather than having to estimate the wall time of the end of GPU drawing. This also prevents anomalies like the GPU phase starting before the draw phase..."
Changes to Tim's submission by Robert: Removal of need for swap buffer callback in ViewerBase.cpp, by
integrating a osg::State::frameCompleted() method that does the stats timing collection. Introduction of a
GraphicsContext::swapBuffersCallbackOrImplementation() method that calls the State::frameCompleted() and
the swap buffers callback or the swapImplementation as required.
functionalities. It includes two main parts: a version checking macro
for handling backward-compatiblity since 3.0, and enhencement of
current schema mechanism. I also change the option handling process to
use getPluginStringData(), and add new USE_SERIALIZER_WRAPPER macro in
the Registry header to allow for static-link usage as well.
The enhencement of schema machanism just tells the type of each
serializer while outputting them, such as:
osg::Group = Children:1
The meaning of the number can be found in the osgDB/Serializer header,
BaseSerializer::Type enum. It may help 3rdparty utilities understand
the structure of the wrapper and do some reflection work in the
future.
The new macro UPDATE_TO_VERSION can help indicate the InputStream (no
affect on the writer) that a serializer is added/removed since certain
OSG version. An example wrapper file is also attached. The
Geode_modified.cpp is based on the serializers/osg/Geode.cpp file
(hey, don't merge it :-), but assumes that a new user serializer
'Test' is added since version 65 (that is, the OSG_SOVERSION):
REGISTER_OBJECT_WRAPPER( Geode, ... )
{
ADD_USER_SERIALIZER( Drawables ); // origin ones
UPDATE_TO_VERSION( 65 )
{
ADD_USER_SERIALIZER( Test ); // a serializer added from version 65
}
}
All kinds of ADD_... macros following UPDATE_TO_VERSION will
automatically apply the updated version. The braces here are only for
typesetting!
While reading an osgt/osgb/osgx file, OSG will now check if the file
version (recorded as the writer's soversion, instead of previous
meaningless "#Version 2") is equal or greater than Test's version, and
try reading it, or just ignore it if file version is lesser.
And we also have the REMOVE_SERIALIZER macro will mark a named
serializer as removed in some version, with which all files generated
by further versions will just ignore it:
UPDATE_TO_VERSION( 70 )
{
REMOVE_SERIALIZER( Test );
}
This means that from version 70, the serializer Test is removed (but
not actually erased from the list) and should not be read anymore. If
the read file version is less than 70 (and equal or greater than 65),
Test will still be handled when reading; otherwise it will be ignored
to keep compatiblity on different OSG versions.
"
Also I've done the osguserstats example. I've kept the "toy example" that was in the modified osgviewer.cpp I had sent you, because they show different uses of custom stats lines (a value displayed directly, a value without bars and a value with bars and graph). I also added a function and a thread that will sleep for a given number of milliseconds and record this time in the stats. I think it clearly shows how to record the time some processing takes and add that to the stats graph, whether the processing takes place on the same thread as the viewer or on another thread.
BTW, feel free to modify the colors I've given to each user stats line... I'm not very artistic. :-)
I've also added more doc comments to the addUserStats() method in ViewerEventHandlers, so hopefully the arguments are clear and the way to get the results you want is also clear. Maybe I went overboard, but the function makes some assumptions that may not be obvious and has many arguments, so I preferred to be explicit."
A new AutoRotateMode was added. I named it ROTATE_TO_AXIS to be
consistent with the other AutoRotateModes, even though it changes from
how is called in Billboard (AXIAL_ROT).
Setters and getters for rotation axis and normal were also added to the
AutoTransform class interface.
The implementation is mainly a copy-paste from Billboard code.
"
- Algorithm doesn't try to merge double and single precision arrays together
- Algorithm doesn't try to merge incompatible geometries (ex: one with "vertices + texoords", and another with only vertices)
2. Better TextureAtlasBuilder
Algorithm is still sub-optimal, but it now tries to fill more blanks, using "unused space in the current line".
(Don't know if I already submitted it, but I guess not)
One day, someone should try to find a good solution to this NP-problem... For instance : http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1.140.200&rep=rep1&type=pdf
"
The user calls statsHandler->addUserStatsLine() providing:
- the label they want for that line in the graph
- the text and bar colors they want in the graph
- the stats names they want queried (one for time taken, one for begin and one for end time) and a few settings for how these will be displayed.
Then all they have to do is call viewer->getViewerStats()->setAttribute(framenumber, name, value) for their three attributes each frame and they'll have their stats in the graph.
They can also give only a time taken attribute (or some other numerical value they want printed, which can be averaged or not), or only begin+end attributes, and the graph will accordingly display only the (average or not) numerical value or only the bars.
Along the way I cleaned up the existing code a bit:
* Each time the setUpScene() or createCameraTimeStats() methods added a line to the graph, they did pretty much the same thing, so I moved that into a separate method called createTimeStatsLine() which is called by setUpScene() and createCameraTimeStats().
* I moved the font, characterSize, startBlocks and leftPos variables to member variables, since they were being passed around everywhere but were set only once at the beginning.
* The geode on which stats lines are added is also kept in a member variable, and createCameraTimeStats() adds the per-camera lines to this geode instead of returning a new Group with a new Geode. This further reduces the number of variables the createCameraTimeStats() method needs as input.
"
to download the texture object to graphics card.
Calling IncrementalCompileOperation::assignForceTextureDownloadGeometry() assigns a geometry
to the job.
changed extensions from .c to .cpp and got compiling as C files as part of the osg core library.
Updated and cleaned up the rest of the OSG to use the new internal GLU.
change was to use doubles for reading and writing matrices regardless of type of Matrix
being serialized.
Change does break backwards compatibility though, so code
path supporting original format has been left in for the
time being. However, this code is not reliable enough and
is over complicated compared to the simplified handling. Once
the new code has been bedded down for a while I'll remove this code block.
I also fixed a possible bug in osgDB::XmlParser that doesn't handle control characters (like " to ") when reading node attributes, because the writeWrappedString() and readWrappedString() now depend heavily on control characters. An additional improvement is that osgx now supports comments."
osg/Camera. The ::glName issue masked the fact that the "(..)" isn't handled
well in doxygen and leads to "(." in the generated files.
I'm also submitting a minor documentation fix to osgGA/DriveManipulator to get
it out of my patch queue ;)"
performance option, I added a flag to control whether the depth writing
pass is performed.
Since text is alpha-blended when rendering, it is placed in the
transparent bin and rendered back to front. Writing to the depth buffer
should therefore be unnecessary. Indeed, rendering something behind text
(or any blended object) after that object is drawn will give incorrect
results whether the depth buffer is written or not. I therefore think it
is safe to keep this option off by default. Users can turn it on for any
special needs they have.
I did not fix the existing backdrop implementations to work with the new
code since this new method of rendering intrinsically handles backdrops
correctly. Its results are more accurate than all of the existing
backdrop implementations. Its only downside is that it requires two
passes if depth buffer updates are desired, whereas DEPTH_RANGE and
POLYGON_OFFSET achieve their (less accurate) results in one pass. The
NO_DEPTH_BUFFER method also only uses one pass, but it disables depth
tests and not depth writes so will have serious problems if anything is
drawn in front of the text before OR after the text is drawn.
Given the better all-around behavior of the new method, I believe the
other backdrop implementations can be safely removed. Code that adjusts
the backdrop implementation will of course be broken if the member
functions are removed. For this reason I left them in, but set the new
rendering method as the default backdrop implementation. At the very
least I think the old backdrop implementations should be deprecated and
removed at a later date.
"
Note from Robert Osfield, testing this submission with osgtext I found that the
text would not render correctly when different text labels were overlapping
in deth and screen space. I change _enableDepthWrites to default to true and
found the that which artifacts still occurred around the alpha blended edges
the artifacts where better than issue with occlusion of nearer pixels that was
happening with _enableDepthWrites set to false.I therefore set the
_enableDepthWrites to true as I feel it's the lesser of the two artefacts.
I move declaration of classes TestResult, QueryGeometry from cpp to header file and made a void createSupportNodes() a virtual method.
Now is possible to inherit from class OcclusionQueryNode."
Goals:
- to handle INCR_WRAP values nicely if not supported by OpenGL (old hardware)
- to support two side stenciling of OpenGL 2.0. Current implementation does not work on ATI as it uses Nvidia extension.
Ready for commit:
- Stencil and Stencil.cpp - please, review them
Ready with "hack":
- StencilTwoSided.cpp: please, see the line 113 in apply():
glEnable(GL_STENCIL_TEST_TWO_SIDE);
This line used to be in getModeUsage() as
usage.usesMode(GL_STENCIL_TEST_TWO_SIDE);
but it produces OpenGL errors on ATI as it is unknown value there (it is Nvidia extension).
Problems with my "glEnable" solution:
- it enables two side stenciling forever, and it will disturb any other single-side stenciling in the scene graph.
"
All size() methods are now renamed to volume(). At present only the CompositePlacer will use it for randomly choose a place according to the volumes of all children.
"
type is supported at present. The attached osgparticleshader.cpp will
show how it works. It can also be placed in the examples folder. But I
just wonder how this example co-exists with another two (osgparticle
and osgparticleeffect)?
Member variables in Particle, including _alive, _current_size and
_current_alpha, are now merged into one Vec3 variable. Then we can
make use of the set...Pointer() methods to treat them as vertex
attribtues in GLSL. User interfaces are not changed.
Additional methods of ParticleSystem are introduced, including
setDefaultAttributesUsingShaders(), setSortMode() and
setVisibilityDistance(). You can see how they work in
osgparticleshader.cpp.
Additional user-defined particle type is introduced. Set the particle
type to USER and attach a drawable to the template. Be careful because
of possible huge memory consumption. It is highly suggested to use
display lists here.
The ParticleSystemUpdater can accepts ParticleSystem objects as child
drawables now. I myself think it is a little simpler in structure,
than creating a new geode for each particle system. Of course, the
latter is still compatible, and can be used to transform entire
particles in the world.
New particle operators: bounce, sink, damping, orbit and explosion.
The bounce and sink opeartors both use a concept of domains, and can
simulate a very basic collision of particles and objects.
New composite placer. It contains a set of placers and emit particles
from them randomly. The added virtual method size() of each placer
will help determine the probability of generating.
New virtual method operateParticles() for the Operator class. It
actually calls operate() for each particle, but can be overrode to use
speedup techniques like SSE, or even shaders in the future.
Partly fix a floating error of 'delta time' in emitter, program and
updaters. Previously they keep the _t0 variable seperately and compute
different copies of dt by themseleves, which makes some operators,
especially the BounceOperator, work incorrectly (because the dt in
operators and updaters are slightly different). Now a getDeltaTime()
method is maintained in ParticleSystem, and will return the unique dt
value (passing by reference) for use. This makes thing better, but
still very few unexpected behavours at present...
All dotosg and serialzier wrappers for functionalities above are provided.
...
According to some simple tests, the new shader support is slightly
efficient than ordinary glBegin()/end(). That means, I haven't got a
big improvement at present. I think the bottlenack here seems to be
the cull traversal time. Because operators go through the particle
list again and again (for example, the fountain in the shader example
requires 4 operators working all the time).
A really ideal solution here is to implement the particle operators in
shaders, too, and copy the results back to particle attributes. The
concept of GPGPU is good for implementing this. But in my opinion, the
Camera class seems to be too heavy for realizing such functionality in
a particle system. Myabe a light-weight ComputeDrawable class is
enough for receiving data as textures and outputting the results to
the FBO render buffer. What do you think then?
The floating error of emitters
(http://lists.openscenegraph.org/pipermail/osg-users-openscenegraph.org/2009-May/028435.html)
is not solved this time. But what I think is worth testing is that we
could directly compute the node path from the emitter to the particle
system rather than multiplying the worldToLocal and LocalToWorld
matrices. I'll try this idea later.
"
Texture.cpp:applyTexImage2D_subload:
<code>
unsigned char* data = = (unsigned char*)image->data();
if (needImageRescale) {
// allocates rescale buffer
data = new unsigned char[newTotalSize];
// calls gluScaleImage into the data buffer
}
const unsigned char* dataPtr = image->data();
// subloads 'dataPtr'
// deletes 'data'
</code>
In effect, the scaled data would never be used.
I've also replaced bits of duplicate code in Texture1D/2D/2DArray/3D/Cubemap/Rectangle
that checks if the texture image can/should be unref'd with common functionality in
Texture.cpp.
"
osgWidget::Input:
[Functional changes]
- Previously, the field would be filled with spaces up to its max length, and typing would just replace the spaces. Also, there was a _textLength variable that kept track of the real length of text in the field, since the osgText::Text's length just reflected the length of spaces+text entered. This was not great, as you could still select the spaces with the mouse and it just feels hacky. So I changed it to only contain the text entered, no spaces, and _textLength was removed since it's now redundant (the osgText::Text's length is used instead).
- Fixed the selection size which (visually only) showed one more character selected than what was really selected.
- Fixed selection by dragging the mouse, it would sometimes not select the last character of the string.
- Cursor will now accurately reflect whether insert mode is activated (block cursor) or we're in normal mode (line cursor) like in most editors.
- Implemented Ctrl-X (cut)
- Added a new clear() method that allows the field to be emptied correctly. Useful for a command line interface, for example (hint, hint).
- Mouse and keyboard event handler methods would always return false, which meant selecting with the mouse would also rotate the trackball, and typing an 's' would turn on stats.
[Code cleanup]
- Renamed the (local) _selectionMin and _selectionMax variables which are used in a lot of places, as the underscores would lead to think they were members. Either I called them selection{Min|Max} or delete{Min|Max} where it made more sense.
- Fixed some indenting which was at 3 spaces (inconsistently), I'm sure I didn't catch all the lines where this was the case though.
- Put spaces between variable, operator and value where missing, especially in for()s. Again I only did this where I made changes, there are probably others left.
The result is that delete, backspace, Ctrl-X, Ctrl-C, Ctrl-V, and typing behaviour should now be consistent with text editor conventions, whether insert mode is enabled or not. I hope. :-)
Note, there's a nasty const_cast in there. Why isn't osgText::Font::getGlyph() declared const?
Also, as a note, the current implementation of cut, copy and paste (in addition to being Windows only, yuck) gets and puts the data into an std::string, thus if the osgText::String in the field contains unicode characters I think it won't work correctly. Perhaps someone could implement a proper clipboard class that would be cross-platform and support osgText::String (more precisely other languages like Chinese) correctly? Cut, copy and paste are not critical to what I'm doing so I won't invest the time to do that, but I just thought I'd mention it.
"
"enable thread locking in libavcodec
This is required for a multithreaded application using ffmpeg from
another thread."
"Prevent the audio from videos from hanging on exit if they are paused.
The video decoder already has similar logic."
"Add a way to retrieve the creation time for MPEG-4 files."
"fmpeg, improve wait for close logic
Both audio and video destructors have been succesfully using the logic,
if(isRunning())
{
m_exit = true;
join();
}
since it was introduced,
but the close routines are using,
m_exit = true;
if(isRunning() && waitForThreadToExit)
{
while(isRunning()) { OpenThreads::Thread::YieldCurrentThread(); }
}
which not only is it doing an unnecessary busy wait, but it doesn't
guaranteed that the other thread has terminated, just that it has
progressed far enough that OpenThreads has set the thread status as
not running. Like the destructor set the m_exit after checking
isRunning() to avoid the race condition of not getting to join()
because the thread was running, but isRunning() returns false.
Now that FFmpeg*close is fixed, call it from the destructor as well
to have that code in only one location."
about the Observer header:
/home/cubicool/local/include/osg/Observer: In copy constructor
?osg::ObserverSet::ObserverSet(const osg::ObserverSet&)?:
/home/cubicool/local/include/osg/Observer:66: warning: base class ?class
osg::Referenced? should be explicitly initialized in the copy
constructor
I've been fixing this by hand by using the attached Observer header; it
does exactly what the warning requests. Purely cosmetic, I believe, but
other than that OSG seems to compile w/ -W -Wall just fine."
* Added support for floating windows and context menus in QMDIAreas.
* Protected the size (_width and _height) by a mutex to prevent threading problems.
Then my own:
* Made sure the embedded widget's size follows the graphicsView's size at all times so that window resizes will resize the widget as expected in fullscreen mode."
creation of main shader to ShaderComposer and
collection of ShaderComponent to osg::State.
Also added very basic shader set up in osgshadecomposition example.
