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.
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."
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 get QWidgetImage to a point where it can fill a need we have: to be
able to use Qt to make HUDs and to display widgets over / inside an OSG
scene.
---------------
Current results
---------------
I've attached what I have at this point. The modified QWidgetImage +
QGraphicsViewAdapter classes can be rendered fullscreen (i.e. the Qt
QGraphicsView's size follows the size of the OSG window) or on a quad in
the scene as before. It will let events go through to OSG if no widget
is under the mouse when they happen (useful when used as a HUD with
transparent parts - a click-focus scheme could be added later too). It
also supercedes Martin Scheffler's submission because it adds a
getter/setter for the QGraphicsViewAdapter's background color (and the
user can set their widget to be transparent using
widget->setAttribute(Qt::WA_TranslucentBackground) themselves).
The included osgQtBrowser example has been modified to serve as a test
bed for these changes. It has lots more command line arguments than
before, some of which can be removed eventually (once things are
tested). Note that it may be interesting to change its name or split it
into two examples. Though if things go well, the specific QWebViewImage
class can be removed completely and we can consolidate to using
QWidgetImage everywhere, and then a single example to demonstrate it
would make more sense, albeit not named osgQtBrowser... You can try this
path by using the --useWidgetImage --useBrowser command line arguments -
this results in an equivalent setup to QWebViewImage, but using
QWidgetImage, and doesn't work completely yet for some unknown reason,
see below.
----------------
Remaining issues
----------------
There are a few issues left to fix, and for these I request the
community's assistance. They are not blockers for me, and with my
limited Qt experience I don't feel like I'm getting any closer to fixing
them, so if someone else could pitch in and see what they can find, it
would be appreciated. It would be really nice to get them fixed, that
way we'd really have a first-class integration of Qt widgets in an OSG
scene. The issues are noted in the osgQtBrowser.cpp source file, but
here they are too:
-------------------------------------------------------------------
QWidgetImage still has some issues, some examples are:
1. Editing in the QTextEdit doesn't work. Also when started with
--useBrowser, editing in the search field on YouTube doesn't
work. But that same search field when using QWebViewImage
works... And editing in the text field in the pop-up getInteger
dialog works too. All these cases use QGraphicsViewAdapter
under the hood, so why do some work and others don't?
a) osgQtBrowser --useWidgetImage [--fullscreen] (optional)
b) Try to click in the QTextEdit and type, or to select text
and drag-and-drop it somewhere else in the QTextEdit. These
don't work.
c) osgQtBrowser --useWidgetImage --sanityCheck
d) Try the operations in b), they all work.
e) osgQtBrowser --useWidgetImage --useBrowser [--fullscreen]
f) Try to click in the search field and type, it doesn't work.
g) osgQtBrowser
h) Try the operation in f), it works.
2. Operations on floating windows (--numFloatingWindows 1 or more).
Moving by dragging the titlebar, clicking the close button,
resizing them, none of these work. I wonder if it's because the
OS manages those functions (they're functions of the window
decorations) so we need to do something special for that? But
in --sanityCheck mode they work.
a) osgQtBrowser --useWidgetImage --numFloatingWindows 1
[--fullscreen]
b) Try to drag the floating window, click the close button, or
drag its sides to resize it. None of these work.
c) osgQtBrowser --useWidgetImage --numFloatingWindows 1
--sanityCheck
d) Try the operations in b), all they work.
e) osgQtBrowser --useWidgetImage [--fullscreen]
f) Click the button so that the getInteger() dialog is
displayed, then try to move that dialog or close it with the
close button, these don't work.
g) osgQtBrowser --useWidgetImage --sanityCheck
h) Try the operation in f), it works.
3. (Minor) The QGraphicsView's scrollbars don't appear when
using QWidgetImage or QWebViewImage. QGraphicsView is a
QAbstractScrollArea and it should display scrollbars as soon as
the scene is too large to fit the view.
a) osgQtBrowser --useWidgetImage --fullscreen
b) Resize the OSG window so it's smaller than the QTextEdit.
Scrollbars should appear but don't.
c) osgQtBrowser --useWidgetImage --sanityCheck
d) Try the operation in b), scrollbars appear. Even if you have
floating windows (by clicking the button or by adding
--numFloatingWindows 1) and move them outside the view,
scrollbars appear too. You can't test that case in OSG for
now because of problem 2 above, but that's pretty cool.
4. (Minor) In sanity check mode, the widget added to the
QGraphicsView is centered. With QGraphicsViewAdapter, it is not.
a) osgQtBrowser --useWidgetImage [--fullscreen]
b) The QTextEdit and button are not in the center of the image
generated by the QGraphicsViewAdapter.
c) osgQtBrowser --useWidgetImage --sanityCheck
d) The QTextEdit and button are in the center of the
QGraphicsView.
-------------------------------------------------------------------
As you can see I've put specific repro steps there too, so it's clear
what I mean by a given problem. The --sanityCheck mode is useful to see
what should happen in a "normal" Qt app that demonstrates the same
situation, so hopefully we can get to a point where it behaves the same
with --sanityCheck and without."
I found very useful to have a control whether osgView::setCameraManipulator does or does not reset camera to home position.
I extended method signature as follows:
void setCameraManipulator(osgGA::MatrixManipulator* manipulator, bool resetPosition = true);
keeping the current usage intact (default parameter), while enabling user to disable the position reset. That can be useful in the situation when manipulator position was already loaded, for example from a file (user specification), or defined any other way, while we do not want to be reset to home position. Other usability is usage of two manipulators in a modeling program (orbiting around the model, walking on the model) and changing between them while we want to preserve the position of a camera in the change. Games may benefit from it as well when we change from user-defined helicopter manipulator to soldier manipulator because the user escaped the helicopter. The camera will change manipulator but the position is expected to be kept in the transition (provided that user makes the state transition between the two manipulators himself).
"
osg::GraphicsContext, in order to give good integration with the
application's GUI toolkit. This works really well.
However, I need to share OpenGL texture resources with the standard
osgViewer GraphicsContext implementations, in particular the
PixelBuffers. This is essential for my application to conserve graphics
memory on low-end hardware. Currently the standard osg implementations
will not share resources with another derived osg::GraphicsContext,
other than the pre-defined osgViewer classes e.g. PixelBufferX11 is
hardcoded to only share resources with GraphicsWindowX11 and
PixelBufferX11 objects, and no other osg::GraphicsContext object.
To address this in the cleanest way I could think of, I have moved the
OpenGL handle variables for each platform into a small utility class,
e.g. GraphicsHandleX11 for unix. Then GraphicsWindowX11, PixelBufferX11
and any other derived osg::GraphicsContext class can inherit from
GraphicsHandleX11 to share OpenGL resources.
I have updated the X11, Win32 and Carbon implementations to use this.
The changes are minor. I haven't touched the Cocoa implmentation as
I'm not familiar with it at all and couldn't test it - it will work
unchanged.
Without this I had some horrible hacks in my application, this greatly
simplifies things for me. It also simplifies the osgViewer
implementations slightly. Perhaps it may help with other users'
desires to share resources with external graphics contexts, as was
discussed on the user list recently."
Notes from Robert Osfield, adapted Colin's submission to work with the new EGL related changes.
I've needed to run a recorded simulation offscreen and save it to a sequence of images, and the ScreenCaptureHandler seemed to be the simplest way to do that, and with this change it's possible.
Another change: I've also added the ability to specify continuous capture of all frames, or a certain number of frames. ScreenCaptureHandler now has a setFramesToCapture(int) method. The argument will be interpreted as:
0 : don't capture
<0 : capture continuously
>0 : capture that number of frames then stop
I also added startCapture() and stopCapture() methods so that user code can start capturing (either continuously or the given number of frames) at a given point in their program. setFramesToCapture() won't start capturing, you have to call startCapture() afterwards. The handler also now has another key to toggle continuous capture (defaults to 'C').
Note that continuous capture will of course only work if the CaptureOperation writes to different files (for example, a WriteToFile with SEQUENTIAL_NUMBER mode) or does something different each time... Otherwise it will just overwrite of course. :-)
I've also taken the chance to refactor the addCallbackToViewer() method a bit too, since finding the right camera is needed in two places now.
I've tested all cases (I think). If you want to try, in osgviewer.cpp and replace the line
// add the screen capture handler
viewer.addEventHandler(new osgViewer::ScreenCaptureHandler);
with
// add the screen capture handler
osgViewer::ScreenCaptureHandler* captureHandler = new
osgViewer::ScreenCaptureHandler(
new osgViewer::ScreenCaptureHandler::WriteToFile(
"screenshot", "jpg",
osgViewer::ScreenCaptureHandler::WriteToFile::SEQUENTIAL_NUMBER),
-1);
viewer.addEventHandler(captureHandler);
captureHandler->startCapture();
And vary the "-1" (put 0, 10, 50) and then use the 'c' and 'C' keys and see how it reacts.
"
GraphicsWindowCocoa-implementation, which enhances multithreaded
stability, it ensures that modifications to the size of an openglcontext
is done only from one thread.
"
implementation of GraoicsWindowCocoa:
Enhancements/Bugfixes:
+ now it's possible to integrate osgViewer better into existing
cocoa-applications:
* create one or more NSOpenGLView(s) and add these to your window(s)
* create one or more NSWindows
* disable the integrated event-polling of osgViewer, and let the work be
done by Cocoa / NSApplicationRun. You'll have to run the osgViewer's
runloop in a separate thread
+ missing menu-event-handling implemented
+ added NSAutoReleasePools where necessary, this fixes some memory-leaks
+ fixed some crashes and thread-issues"
Original email from Frederic at start of thread:
"he patch attached, made from r10068, fix two things, in other of importance :
- the selected cursor is never shown ( second change in file ). Only the left arrow is always displayed.
- remove the arbitrary ( in my sense ) limitation that the user cannot choose a cursor with the same shape that one used when resizing the window. This limitation doesn't exist for X11, and we have a diverging behaviour there ( first change in file ). Flightgear use the LeftRightCursor in look around mode."
Follow up email from Frederic (with changes that finally made it into this check in:
"I've just tested Mark's suggestion and it works perfectly, even when the
cursor goes to the border then come back inside the window.
But his patch doesn't seem to be based on the last revision of the
files, or at least not on the trunk, and there are more changes than
expected in them, including some loss from the previous patches.
The patch attached is based on r10068 of
src/osgViewer/GraphicsWindowWin32.cpp and r10067 of
include/osgViewer/api/Win32/GraphicsWindowWin32"
http://www.mail-archive.com/osg-users@lists.openscenegraph.org/msg23734.html
The change is source compatible with current osg code and will not affect current users, it simply adds an additional parameter to the GraphicsWindowWin32::WindowData struct constructor and defaults to the current behavior.
Attached are the files "include/osgViewer/api/Win32/GraphicsWindowWin32" and "src/osgViewer/GraphicsWindowWin32.cpp" with my changes, based on svn revision 10045. In addition, I have provided an svn patch file with the same changes for your convenience.
I have discussed the matter with my supervisor, and agreed that my company makes no copyright claim over this extremely trivial change (or to put it another way, we assign copyright to the open scene graph community.)"
It's a minimal change, it just calls an already existing protected method. It was trivial to subclass the handler to do it in our code, but pushing the change into OSG makes sense as it's generally useful to have it in the handler itself.
I also noticed that the handle() method was overridden from osgGA::GUIEventHandler but wasn't marked virtual. It wasn't intended that subclasses not be able to override it in turn, so I've added the keyword.""
set to off. But could be activated/decativated via CMake as well as system
environment variable. I also modified src\osgViewer\CMakeLists.txt to turn
off this workaround by default as suggested."
facilitate the subclassing of Image providing interactive behaviours so as
used in the vnc interactive VncImage class.
osgViewer::InteractiveImageHandler provides an event handler that convertes osgGA
mouse and keyboard events into the coordinate frame of an image based on ray intersection with geometry in
the associated subgraph.
Changed the ordering of events processing in Viewer and CompositeViewer to allow
scene graph event handlers to take precidence over viewer event handlers and camera manipulators
This code will add two extra statistics options:
-Camera scene statistics, stats for the scene after culling (updated at 10 Hz)
-View scene statistics, stats for the complete scene (updated at 5 Hz)
Each camera and each view will expand the statistics to the right.
I also added the requests and objects to compile of the databasepager to the databasepager statistics.""
viewer.addEventHandler(new osgViewer::ScreenCaptureHandler(
new osgViewer::WriteToFileCaptureOperation("filename", "jpg")));
and the filename will be what you want. The WriteToFileCaptureOperation will add the context ID and the file number (if in SEQUENTIAL_NUMBER mode) to the file name.
(The attached also clarifies some notify messages, and corrects the comment when adding the handler in osgviewer.cpp)
I also remembered, the current architecture could allow a different CaptureOperation for each context, but currently the API only allows setting one CaptureOperation for all contexts. This could be improved if need be.
"
avoid threading issues associated with compile running in a parallel with
update/cull on the first frame.
Also added automatic recompile when a new SceneData is applied to a View.
get an excess Tab key report when switching back to an OSG
application (usually FlightGear :-). Although KDE has consumed
the Tab, it's sometimes still in the XKeymapEvent's key_vector,
and followed by a Tab KeyRelease event.
Avoid this artifact by
- asking for a "fresh" keymap (via XQueryKeymap()), rather than
using the unreliable(?) XKeymapEvent's key_vector, and by
- flushing all key events on focus-in (to avoid the KeyRelease)
After Super-press, Tab-press, Super-release, Tab-release (note
the wrong release order!) I still get an extra Tab event. But
this is not surprising and not exactly wrong either. Also it's
hard to avoid, as we can't see what happened to the keyboard
before we regained focus.
Files changed:
src/osgViewer/GraphicsWindowX11.cpp
include/osgViewer/api/X11/GraphicsWindowX11"
It's implemented in the same way that 3D Spherical Display and Panoramic Spherical Display.
You can test it running:
osgviewer --wowvx-20 cow.osg
osgviewer --wowvx-42 cow.osg
depending on the size of your Philips WOWvx display (20" or 42")
Other arguments you can use to control the 3D effect are:
--wow-content <value>
This value defines the kind of content that can be:
0: No depth
1: Signage
2: Movie
3: CGI
4: Still
--wow-factor <value>
Percentage of the display recommended depth value. Default 64, Range [0-255]
--wow-offset <value>
Amount of range behind the screen. Default 128, Range [0-255]
0: Range is shifted in the direction of the viewer.
128: Range is equally divided in front and behind the screen.
255: Range is shifted away from the viewer.
"
key, but it didn't pick up the initial state. So, if NumLock was on for
the OS at startup (LED on), it was still off for OSG. And the first
keypress turned the LED off, and NumLock on for OSG. The attached fix
picks up the state on every FocusIn, just like it was done in the last
commits for CapsLock. The difference is, that the NumLock mask isn't
standardized (e.g. 0x10 for Linux, and 0x80 for AIX), so we have to do
a reverse lookup (::rescanModifierMapping()).
Note that I could not reproduce the problem on my system, but someone
else confirmed it twice on his, and the patch fixed it for him.
Changed files:
./include/osgViewer/api/X11/GraphicsWindowX11
./src/osgViewer/GraphicsWindowX11.cpp
"
It sets osgGA's keymask when restoring keys on FocusIn, according
to the state values of XKeyEvent and XCrossingEvent. (These are
the only source for X11's current capslock state that avoids
pulling in the XKB extension.)
"
The win32 pbuffer implementation returned an error unless both the
WGL_ARB_pbuffer and the WGL_ARB_render_texture functions were present.
This was too restrictive, as a pbuffer can usefully be created without
render-to-texture, e.g. for use with glReadPixels. The osg 1.2/Producer
pbuffers worked without RTT, and osgUtil::RenderStage has all the code to
handle both RTT and non-RTT pbuffers, doing a read and copy in the
latter case.
With these changes I have successfully tested the osgprerender example
on a graphics card which supports RTT, and one which doesn't. Plus
tested in my own application.
In order to aid diagnostics I have also added more function status
return checks, and associated error messages. I have included the win32
error text in all error messages output. And there were some errors
with multi-threaded handling of "bind to texture" and a temporary window
context which I have corrected.
These is one (pre-existing) problem with multi-threaded use of pbuffers
in osgViewer & osgprerender, which I have not been able to fix. A win32
device context (HDC) can only be destroyed from the thread that created
it. The pbuffers for pre-render cameras are created in
osgUtil::RenderStage::runCameraSetUp, from the draw thread. But
closeImplementation is normally invoked from the destructor in the main
application thread. With the additional error messages I have added,
osgprerender will now output a couple of warnings from
osgViewer::PixelBufferWin32::closeImplementation() at exit, after
running multi-threaded on windows. I think that is a good thing, to
highlight the problem. I looked into fixing it in osgViewer::Renderer &
osgUtil::RenderStage, but it was too involved for me. My own
application requirements are only single-threaded.
Unrelated fix - an uninitialised variable in
osg::GraphicsThread::FlushDeletedGLObjectsOperation().
"
remain in pressed state after revealing, even if they are no
longer pressed on the keyboard. This can have bad effects,
especially if the stuck keys are modifier keys. One has to
press and release the stuck keys again to reset the wrong state.
The fix keeps track of all key presses and releases. On FocusOut
and UnmapNotify it releases all keys that are in pressed state,
and on KeymapNotify (following a FocusIn), it sets the currently
pressed keys again. To avoid confusion in the OSG-using application
normal keys are always reported released /before/ and pressed
/after/ modifier keys.
As current key states are returned as char[32] keymap by
XQueryKeymap and XKeymapEvent, this format is also used to
recognize modifier keys and for maintaining the current
internal key state. Functions to set/clear/query bits in
such a keymap are added.
The patch was extensively tested with osgkeyboard and
FlightGear under KDE and fvwm2. It was not tested on a
Xinerama setup or with multiple windows, but as _eventDisplay
is used throughout, there should be no problems. The patch also
makes the following changes:
- removes old and obsolete handling of modifier keys in ::adaptKey().
This wasn't only unused, but also wrong (and for that reason commented
out in revision 7066). The modifier states are actually handled
in ./src/osgGA/EventQueue.cpp (EventQueue::keyPress/keyRelease).
- fixes some spelling"
pbuffer functions or exactly ask for the extensions we need to call the
apropriate glx extension functions for and around pbuffers extensions.
The glx 1.3 version of this functios are prefered. If this is not pressent we
are looking for the glx extensions and check for them.
Prevously we just used some mix of the glx 1.3 functions or the extension
functions without making sure that this extension is present.
"
carbon-implementation of GraphicsWindow. Now you can use an AGLDrawable
in conjunction with osgViewer/osgCompositeViewer."
Changes from Robert Osfield, changed std::cout to osg::notify(osg::INFO)
