databasePager->setUpThreads(16, 1);
We experienced problems with multiple databasepagers loading files in parallel, when two threads start to load the same file (usually a texture referenced by multiple models). The second thread to add the file to the cache (sometimes) manages to do so while the refcount from the cached object still is zero, causing the object loaded to be destroyed.
Sometimes the second thread manages to ref() the object before Referenced::signalObserversAndDelete does the final recount check, causing a warning:
"Warning Referenced::signalObserversAndDelete(,,) doing delete with _refCount=1"
With a deleted object added to the scenegraph we get some undesired results, I think the program only crashes if the object was a Node, and just has some untextured surfaces if it was a texture, but I'm not completely sure.
Attached is a modified version of the Registry.cpp, returning the object in cache and let the duplicate loaded object to be destroyed.
A more efficient option would be to add some sort of blocking entry to the objectcache to stop the second thread from reading the file, and just wait until the first thread added it to the cache. If you think that's worthwile we would be happy to implement that version. A bit tricky to implement and test, that's why I submit a simple version that stops my program from crashing."
git-svn-id: http://svn.openscenegraph.org/osg/OpenSceneGraph/trunk@14300 16af8721-9629-0410-8352-f15c8da7e697
"I noticed that Text3D objects would change there z alignment depending on the alignment mode. I'm not sure if this was intentional or just a simple mistake. My expectation was that the front of the object would always stay aligned to the 0 z-plane, regardless of the alignment mode. I've attached an updated version that retains a consistent z-alignment."
"I just now noticed another issue with Text3D objects. It was not properly computing the bounding box when non-axis aligned rotations were being applied. In this case all corners of the bounding box need to be transformed in order to get the correct containing box. I've attached the updated file."
"The incorrect bounding box problem also applies to regular Text objects. I've attached the fix for that as well as the original Text3D fix."
git-svn-id: http://svn.openscenegraph.org/osg/OpenSceneGraph/trunk@14296 16af8721-9629-0410-8352-f15c8da7e697
optimizations to reduce cpu overhead:
1) Avoid a load-hit-store in UpdateBone. b->getMatrixInBoneSpace()
returns the same matrix that was just stored with b->setMatrix()
2) Avoid calling element->isIdentity() for the whole transform stack
(can be expensive is element is a matrix)
3) Make the key frame interpolator use binary search instead of a
linear one. This is very noticeable in scenes where some geometry has
long repeating animations that start at the same time, you will see
the update time grow then reset and grow again."
git-svn-id: http://svn.openscenegraph.org/osg/OpenSceneGraph/trunk@14294 16af8721-9629-0410-8352-f15c8da7e697
debug message it prints out on the console.
Around line 1040 of Registry.cpp (see code below) the method returns
"simpleFileName" but prints about returning "filename".
In attachment the modified file, based on osg 3.2.0
ricky
<code>
if(fileExists(simpleFileName))
{
OSG_DEBUG << "FindFileInPath(" << filename << "): returning " <<
filename << std::endl;
return simpleFileName;
}
</code>
"
git-svn-id: http://svn.openscenegraph.org/osg/OpenSceneGraph/trunk@14285 16af8721-9629-0410-8352-f15c8da7e697
The osg 3ds plugin modify the exported materials name in the same way it modifies the node names.
I've added an option to preserve originals materials names, with the assurance of unique material names are preserved."
git-svn-id: http://svn.openscenegraph.org/osg/OpenSceneGraph/trunk@14283 16af8721-9629-0410-8352-f15c8da7e697
Initially I described issue in message:
http://forum.openscenegraph.org/viewtopic.php?t=13820
It solves issue with compiling texture using ico from image with mipmaps
I added enviroment variable OSG_GL_TEXTURE_STORAGE_ENABLE to control usage of glTexStorage2d. Initially it is disabled.
It used only if image have mipmaps.
Another issue is converting from internalFormat + type to sized internal format. I created sizedInternalFormats[] struct where sized internal formats are ordered from worse->best.
also this struct have commented lines. Commented formats are listed in
http://www.opengl.org/wiki/GLAPI/glTexStorage2D
but looks like not using in osg."
Note from Robert Osfield. Changed the env var control to OSG_GL_TEXTURE_STORAGE and made it's value true by default when the feature is supported by the OpenGL driver. To disable to
use of glTexStorage2D use OSG_GL_TEXTURE_STORAGE="OFF" or "DISABLE"
git-svn-id: http://svn.openscenegraph.org/osg/OpenSceneGraph/trunk@14275 16af8721-9629-0410-8352-f15c8da7e697
- Added apply(Drawable) and apply(Geometry) to NodeVisitor
- Added accept(NodeVisitor) method to Drawable/Geometry
- Added traverse(NodeVisitor) to Geode which calls accept(NodeVisitor) on all Drawables
- Updated CullVisitor to use new apply(Drawable) to handle drawables. The apply(Billboard) method still manually handles the drawables since it is depends on the billboard settings. I needed to disable the traverse within billboard to prevent duplicate traversal of drawables.
- Update other osgUtil node visitors (GLObjectsVisitor, IncrementalCompileOperation, ..) to use new apply(Drawable) method.
"
pd->tid.set( (void*)_beginthreadex(NULL,static_cast<unsigned>(pd->stackSize),ThreadPrivateActions::StartThread,static_cast<void *>(this),0,&ID));
the method "pd->tid.set" sets the thread id, however via the startup function "ThreadPrivateActions::StartThread" that thread id is used (see further down the call hierarchy the line "int status = SetThreadPriority( pd->tid.get(), prio);".
Until now I never ran into any problem in debug or release builds, though. It seems that furtunately the tid.set method was executed always before the tid.get method in the startup code. However, this may make trouble in the furture. A simple solution is the following: just replace the line above with following two lines:
pd->tid.set( (void*)_beginthreadex(NULL,static_cast<unsigned>(pd->stackSize),ThreadPrivateActions::StartThread,static_cast<void *>(this),CREATE_SUSPENDED,&ID));
ResumeThread(pd->tid.get());
The trick is just starting the thread in suspended mode so the StartThread function does not get executed and we can safely store the tid by pd->tid.set. Then start the Thread by calling ResumeThread."
ran into two issues.
At first you get a bunch of warnings that osg::ComputeBoundCallback
and osg::UpdateCallback were unsupported wrapper classes when
converting fbx models with skeletal animation to osg(t/b).
The second issue was that when reading, the readers fail to read the
ComputeBoundCallback and UpdateCallback and set them to NULL which
messes up the RigGeometry.
Because a RigGeometry makes his own classes in the constructor it
might be preferable to not write them at all, because now those
classes are being made two times when reading a RigGeometry. But after
thinking about this that would place too much limits on them (you
won't be able to share or name them and save that information or make
a new inherited class from them and write that one) So I ended up
thinking the best way was to just write the files.
"
We were running into issues occasionally in osgEarth where multiple threads were writing out files like /1/2/3.jpg and /1/3/4.jpg. Both threads would try to create the /1 directory and only one of them would succeed. So the first thread would write out the full /1/2/3.jpg while the second thread wouldn't create the /1/3 directory b/c /1 was already created and the writing of /1/3/4.jpg would fail.
"
There was code in the osgViewer/Viewer.cpp and osgViewer/CompositeViewer.cpp that transformed the Y-coordinates of an event. The code in the composite viewer did however miss the touch-data of the event. I thought that it should really be the GUIEventAdapter that should know about this, and hence I added the
GUIEventAdapter::setMouseYOrientationAndUpdateCoords which is re-computing the coordinates. First I simply added a boolean to the setMouseYOrientation function:
setMouseYOrientation( MouseYOrientation, bool updatecooreds=false );
but then the serializer complained.
This function is called from both the Viewer and the CompositeViewer. We have not tested from the viewer, but I cannot see it would not work from visual inspection.
The other change is in MultiTouchTrackballManipulator::handleMultiTouchDrag. I have removed the normalisation. The reason for that is that it normalised into screen coordinates from 0,0 to 1,1. The problem with that is that if you have a pinch event and you keep the distance say 300 pixels between your fingers, these 300 pixels represent 0.20 of the screen in the horizontal domain, but 0.3 of the screen in the vertical domain. A rotation of the pinch-fingers will hence result in a zoom in, as the normalised distance is changing between them.
A consequence of this is that I have changed the pan-code to use the same algorithm as the middle-mouse-pan.
The rest of it is very similar from previous revision, and there has been some fine-tuning here and there.
"
To select standard OpenGL 1/2 build with full backwards and forwards comtability use:
./configure
make
OR
./configure -DOPENGL_PROFILE=GL2
To select OpenGL 3 core profile build using GL3/gl3.h header:
./configure -DOPENGL_PROFILE=GL3
To select OpenGL Arb core profile build using GL/glcorearb.h header:
./configure -DOPENGL_PROFILE=GLCORE
To select OpenGL ES 1.1 profile use:
./configure -DOPENGL_PROFILE=GLES1
To select OpenGL ES 2 profile use:
./configure -DOPENGL_PROFILE=GLES2
Using OPENGL_PROFILE will select all the appropriate features required so no other settings in cmake will need to be adjusted.
The new configuration options are stored in the include/osg/OpenGL header that deprecates the old include/osg/GL header.
- ReaderWriterFBX.cpp: add "z up scene axis" support: FBX provides facility to convert model scene axis during conversion. Currently fbx plugin convert axis to fbx:opengl axis system (which is arbitrarily at Y up, as opengl is in reality axis agnostic) and sometimes what is needed is Z up so added an option for Z up conversion
- FindFBX.cmake: add support for latest fbx sdk ( 2014.2 )"
format. Turned out that the serializer didn't handle bone names with
spaces very well (the 3ds studio max biped for instance has spaces by
default). Here is a small fix for the problem."
- materialName used to be not stripped of whitespace, making number of models
fail to load materials; now fixed
- stripping was considering spaces only, thus models using tabs had problems
to load correctly; fixed
- fixed references to textures; they did not performed conversion to native
directory separators
- make d (dissolve) takes precedence over Tr (transparency); there seems to be
a confusion about the Tr item - some claiming 1 to be opaque and 0
transparent, while number of models uses exactly the opposite. d (dissolve),
if present in the model, does not suffer from this confusion, thus using it
instead fixes the problem for many many models.
I put many comments to the file concerning d and Tr item as others may further
investigate. Let me know in the case of any problems."
This behavior is also described in the pthreads man page (http://man7.org/linux/man-pages/man3/pthread_create.3.html):
>
> Linux-specific details
> The new thread inherits copies of the calling thread's capability
> sets (see capabilities(7)) and CPU affinity mask (see
> sched_setaffinity(2)).
>
To prevent this behaviour I wrote a patch that explicitly sets the affinity mask to all cores of the system, if no specific affinity was defined with PThread::setProcessorAffinity(unsigned int) .
Thank you!
"
* Changes are made against trunk
* Reason: crashes when using specific constructor from RayIntersector
* Info: Line 42: added in constructor
RayIntersector::RayIntersector(const Vec3d& start, const Vec3d&
direction) missing initialisation of _parent
"
My fix was to rename the standard request handler to a specialized user-event-handler which handles only requests for "/user-event“
So fonts should work on iOS when loaded remotely, even when a local file is available and with the resthttp-plugin serving the presentation.
"
Added the Cmake option OSG_USE_LOCAL_LUA_SOURCE to control whether to build and use the Lua source code in the lua plugin, or look for lua as an external dependency.
To the Lua plugin added support for assigned lua functions to C++ osg::Objects via the new osg::CallbackObject mechanism. To invoke the scripts function from C++ one must get the CallbackObject and call run on it.
Renamed ScriptCallback to ScriptNodeCallback to avoid possibly confusion between osg::CallbackObject and the ScriptNodeCallback.
From Robert Osfield, changed the example so that the vertical and horizon scalar bars are rotated to the XZ plane so you can see them with the default viewer's camera orientation.
Tweaked the positioning of title text of vertic scalar bar to avoid overlap of text.
There are two problems:
1> for DrawElementsUShortPrimitiveType (and UInt) the source_pindex still equals -1 and causes a crash
in DrawElementsUBytePrimitiveType source_pindex is incremented, and in DrawElementsU(Short/Int)PrimitiveType primitiveNum is incremented, but never used
2> The drawelements need to be rewritten as the vertices are reordered.
created a patch for osg stable branch(r14038): attached as Geometry-osg-3.2.zip
and for svn brach(r14044): attached as Geometry_osg_svn.zip"
This can lead to inconsistency if you bind a framebuffer with multiple attachments in DRAW mode and then a framebuffer with different attachment count in READ mode (for example to manually "blit" from a FBo to another).
On some ATI cards (at least RADEON HD) this also leads to an "incomplete " FBO status
I've added a test to enable drawbuffers only if target is "DRAW" or "READ_DRAW", this solves my problems on ATI cards."
* forwarded touch-events do have a correct input-range from 0 .. 1
* I refactored sending touch-events per osc so the receiver can detect a TOUCH_ENDED better"
One solution is naturally to create a new class that would inherit the osg::ComputeBoundVisitor, and use that. I don't like that idea as the ComputeBoundVisitor does actually have what I need - it is only hidden in a protected function.
I am therefor suggesting a slight generalization of the ComputeBoundVisitor with the attached patch, which is tested.
The patch has two parts:
we add applyBBox() so that one can use that in a customized traverse-function and add a bbox to the visitor. I considered calling this function expandByBBox(), but I though applyBBox was better.
The MatrixStack is made available to the outside world. That enables a traverse-function to do whatever it wishes.
I do actually only need one of the two, as I can implement what I wish either way, but adding getMatrixStack() will make more generic expansions possible.
"
From Robert Osfield, changed the name of the new applyBBox(..) method to applyBoundingBox(..) to keep it's naming more consistent with the rest of the OSG.
I added a new tag to p3d called forward_touch_event_to_device and renamed the existing forward_event_to_device to forward_mouse_event_to_device. This new tag will transmit touches to the virtual trackpad as touch events. I added the MultitouchTrackball to the p3d-app so zooming and moving a model remotely should now work, if you use forward_touch_event_to_device. I kept (and fixed) forward_mouse_event_to_device for background compatibility, so old presentations works as in previous versions, without the ability to zoom + scale. of course.
forward_touch_event_to_device needs some more testing, (e.g. with image-streams and keystone, afaik there’s no support for touch-events...) but for a first version it works nice.
"
Examples of supported matches are:
<Slide> and <slide> will be treated the same
<bgcolor>WHITE</bgcolor> and <bgcolor>White</bgcolor> will be treated the same
<text alignment="TOP_LEFT"</text> and <text alignment="top left"</text>, <text alignment="TopLeft"</text> will all be treated the same
From Robert Osfield, changed Roni's code to use a #define GETDEVICEPIXELRATIO to access the versioned Qt devicePixelRatio() method to avoid duplication of the Qt version checking.
I have tested this on:
mac/qt5.2
linux/qt5.2
windows/qt5.2, and
mac/qt5.1
All platforms perform as expected.
The previous fix removed the -f flag to the moc-pre-processor, but on windows, it turned out that -f "osgQt/QGraphicsViewer" was needed.
This becomes an include-statement in the file generated by moc which is needed for compiling it. I ask you consider this patch for the trunk and the 3.2 branch.
Secondly, I wonder if it would be possible to apply my patch for FindRSVG.cmake from 22nd November in the 3.2 branch.
In short, the version of librsvg must be equal or higher to 2.35:
PKG_CHECK_MODULES(RSVG librsvg-2.0>=2.35)
"
Updated lua plugin to use new osgDB::PropertyInterface to run methods.
Added addChild/removeChild() etc to Group.cpp, and addDrawable/removeDrawable() etc. to Geode.cpp serializers.
Hence, moc will complain when osg throws in a -f without anything after it. Hence I propose removing the -f on Qt5 builds. I have tested building without -f on both qt520 and qt511, and that works well.
The attached src/osgQt/CMakeLists.txt that can be patched into 3.2 safely. For the trunk, I would consider dropping the check on the version, and simply remove the option on qt5. I have tested that on qt5.1.1, and that worked fine. Question is however if it works on qt5.0. Probably it does, so the question is simplicity of CMakeList.txt vs safety."
Undefined symbols for architecture x86_64:
"lua::LuaScriptEngine::pushValue(osg::Quat const&) const", referenced from:
PushStackValueVisitor::apply(osg::Quat const&) in LuaScriptEngine.o
"lua::LuaScriptEngine::pushValue(osg::Plane const&) const", referenced from:
PushStackValueVisitor::apply(osg::Plane const&) in LuaScriptEngine.o
ld: symbol(s) not found for architecture x86_64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
Looks like LuaScriptEngine was missing implementation of those 2 member functions. Fixed src/osgPlugins/lua/LuaScriptEngine.cpp file in attachment.
"
* a new command-line-option to present3d and a new option to the p3d-plugin to suppress any found <env> tags
* a new command-line-option to present3d to forward mouse-events via osgGA::Device (defaults to off) so we can test the interface-files with present3d better
* I added a new attribute forward_to_devices for click_to_event to forward the event to all attached devices instead of handling the event locally. This will fix the annoyance with the new interface-files when toggling polygon-mode or switching light on/off.
Here’s an example:
<click_to_event forward_to_devices="true">0x72</click_to_event>
"
* osgDB::FileUtils uses now the Cocoa-API to determine the paths of the application-support-folder
* DarwinUtils uses now modern functions of the quartz-api to get and set screen-resolutions. Removed some of the osg-deprecated stuff.
"
I'm not sure if the IVE was simply generated incorrectly, or if the Image::getTotalSizeInBytesIncludingMipmaps() was modified since the file was generated. Either way, I added a simple check in the IVE loader so that it clears the mipmap offsets if the actual data size does not match the computed data size. This seems like a safe fallback since the mipmap data can be automatically generated, and it fixes the problem in my case.
Also, while looking into this issue, I noticed that the osgDB::InputStream class applies the serialized image allocation mode. However, since the serializer is allocating the image data itself, it seems like it should force the allocation mode to USE_NEW_DELETE.
"
* force _OPENTHREADS_ATOMIC_USE_BSD_ATOMIC for IOS device + simulator as the test does not pick the right implementation
* fixed a small compile-bug for iphone-example
* added a check to prevent multiple realization of a GraphicsWindowIOS-object
"
* MultiTouchTrackball: some code cleanup and support for normalized touch-points
* oscdevice: receiving and sending multi-touch-events via the Cursor2D-profile from TUIO
* added some documentation
Added new WindowSystemInterface::setDisplaySettings() method to provide a mechanism for passing settings onto the WindowSystemInterface so it can then set up the system appropriately.
Added assignment of the DisplaySettings to the WindowSystemInterface in Viewer/ComppsiteViewer::realize().
set to "SCENE" in the file header. But if the file is stored and
then extracted again from an osga archive this header info is lost,
and the resulting file is just an "OBJECT". Possibly other plugin
operations would have the same effect. The osgt/osgb plugin won't
then return the scenegraph contents.
I have updated the osgt/osgb plugin to return a node from an "OBJECT"
file."
But the nested stream implementation for osga archive files doesn't
support seeking. So osgb files can't currently be used in an osga
archive e.g. if osgdem is used to output a osgb format database it
can't be packaged in an archive file, in the same manner that ive
files could.
I've added seek support to the osga nested stream implementation."
* fixed a bug with multi-touch and touch-id-generation on iOS and OS X. (will fix a bug reported by Colin Cochran, without ditching the existing logic)
* removed unnecessary warning-flagss when generating xcode-projects via cmake, will enable the usage of OSG_AGGRESSIVE_WARNING_FLAGS
* added support for 10.9 (OS X)
* new cmake-variable: IPHONE_VERSION_MIN, this will set the deployment-target (previously hard-coded) If you set the IPHONE_VERSION_MIN to something like 7.0 osg gets compiled also for 64bit (amd64)
* cmake defaults now to the clang compiler if IPHONE_VERSION_MIN > 4.2
* cmake now sets some xcode-settings so the compiler uses the c++98-standard (clang defaults to c++11, w/o this I got a lot of linking errors)
* removed include-dir for avfoundation-plugin as not needed on OSX/IOS.
* enhanced the ios-example, will now show multitouch-information on a hud (similar to the osgmultitouch-example), and more importantly, will compile + link out of the box
* small enhancements for the osc-device-plugin (send only one msg for MOVE/DRAG, even if multiple msgs/event is enabled)
* better memory-handling for the zeroconf-plugin
* fixed a possible bug in the rest-http-plugin when receiving mouse-events.
* incorporated a fix from Colin Cochran "forwarded touch events are not transformed into the GL UIView“
"
the wrappers it contains. The registration creates a prototype
object for all of the wrapped classes. For some of the higher-level
classes this can be a bit heavy.
I noticed a problem with a model which required a single class from
osgSim. When osgdb_serializers_osgsim.so was loaded it registered
wrappers and created prototype objects for all of the osgSim classes,
including osgSim::ScalarBar. The constructor for that class creates
several drawables, and loads arial.ttf using the freetype plugin. I
don't need that, and don't even ship the font or plugin with my
application, resulting in an unexplained warning message loading
the model.
I've modified the ObjectWrapper class to defer the prototype object
creation until if & when actually required."
required to build newever OSG on this OS. Also corrects file name
in the error message - I was confused not to find OSCHostEndianness.h
after I've got this error.
Tested by successfully building OSG 3.2.0 with this patch on FreeBSD
9.1."
loading multiple VRML files in parallel. Christopher R. Baker has
detected this bug and crafted a patch. In addition, libcoin has to be
also built with the "--enable-threadsafe" option.
I copy here his report, extracted from
(https://bugs.launchpad.net/ubuntu/+source/openscenegraph/+bug/1211993)
and attach his fix. All credit is due to him:
«
There are three instances of a classical method-local-static
multithreaded initialization bug in the Inventor plugin for OSG that
trigger various memory faults when reading multiple VRML files in
parallel via osgDB::readNodeFile. These bugs are of the form:
static std::map<Stuff,OtherStuff> myHandyMap;
static bool once = true;
if(once) { ...fill myHandyMap; once = false }
... use myHandyMap;
To repeat: try loading multiple VRML files from multiple threads. The
liklihood of the bug depends on many factors, but my application, which
parallel-loads some dozens of small (<100K) VRML files on startup,
triggers this problem 25% of the time or more.
The attached patch (inventor-plugin-multithread.patch) rectifies this
problem by:
1 - Inheriting MyHandyMap from std::map, then
2 - Moving the map initialization into the derived constructor, which
3 - Is intrinsically protected from multithread issues by g++ (and is
part of the C++ standard), unless you pass -fno-threadsafe-statics,
which is strongly discouraged by the man page.
»
"
The delay between the 2 types of messages is more noticeable on Windows 8 and leads to serious disruptions in our application.
Mouse messages generated by touch input are only present for legacy support. I think they should be filtered out by OSG (real click events originating from a physical mouse will of course still go through).
This is what this patch does, according to this suggestion: http://msdn.microsoft.com/en-us/library/dd693088%28v=VS.85%29.aspx (third issue in this page)."
In CMakeList.txt there is a case sensitivity issue which a fix was posted by Robert Osfield in the users forum.
I also had to rename the files OSXAvFoundationCoreVideoTexture.h and OSXAvFoundationCoreVideoTexture.cpp to OSXAVFoundationCoreVideoTexture.h and OSXAVFoundationCoreVideoTexture.cpp
Finally in OSXAvFoundationCoreVideoTexture.cpp the include OSXAVFoundationVideo.H was updated to OSXAVFoundationVideo.h"
