Hi Robert,
On 30/05/11 11:36, Robert Osfield wrote:
Hi J.P, et. al,
I have tracked down and fixed the setCursor() crash issue ;-)
The problem occurred because of subtle timing issue with the
GraphicsContext::_threadOfLastMakeCurrent variable in the
GraphicsContext::makeCurrent() and how this variable is used in the
GraphicsWindowX11::getDisplayToUse().
What was happening is that viewer.realize() would create the windows
and start the graphics thread on each window created and this thread
would then do a makeCurrent() which onces complete would set the
_threadOfLastMakeCurrent and then any getDisplayToUse() would then
select the appropriate GraphicsWindow11::_eventDisplay or
GraphicsWindow11::_display connection to the X server. The problem
was that while the GrapohicsWindowX11::makeCurrentImplementation() was
happening in the graphics thread, the main thread was doing a
GraphicsWindow::setCursor() which in turn was calling
getDisplayToUse() which returned _display rather than _eventDisplay as
the makeCurrent() hadn't yet reset the _threadOfLastMakeCurrent, but
using _display while another thread was doing a glxMakeCurrent isn't
safe - and the crash occurred.
To test it out I added a micro sleep into the main thread before the
setCursor and this solved the issue if the sleep was long enough, the
sleep didn't have to be too long - just long enough for the first
makeCurrent() to do it's stuff, but it was easy to see how this would
really timing sensitive.
Once I confirmed it was a timing issue on the setting of
_threadOfLastMakeCurrent it just became a simply tweak of
GraphicsContext::makeCurrent() so it set this member varaible prior to
calling makeCurrentImplementation(). With this change I could remove
the micro sleep from the main thread and everything runs smoothly with
the setCursor. I run 100 iterations of your test app and every time
it came up correctly without any warnings :-)
An svn update will get you this fix. Please let me know how you get on.
thanks, just updated and the multithreaded version now works without the
stop/start threading.
Thanks for the detailed story, I love reading bug resolution quests :)
Someone should make a book out of a collection of them one day.
rgds
jp
Cheers,
Robert.
_______________________________________________
osg-users mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
--
This message is subject to the CSIR's copyright terms and conditions, e-mail legal notice, and implemented Open Document Format (ODF) standard.
The full disclaimer details can be found at http://www.csir.co.za/disclaimer.html.
This message has been scanned for viruses and dangerous content by MailScanner,
and is believed to be clean.
_______________________________________________
osg-users mailing list
[email protected]
http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
