On Tue, 2007-06-12 at 16:13 +0100, Robert Osfield wrote:
> Hi Jeremy,
> 
> Bug 1.  Impressive... Impressively daft keyboard/mouse wizzardary :-)
> 
> My guess is that the bug isn't actually related to
> TrackballManipulator, it just happens to be the thing in the way when
> things start falling apart.  In theory the bulk event handling should

Well, I couldn't get it to happen no matter how hard I tried if I used
any of the other manipulators (Drive or Flight, although it does happen
with the NodeTracker--but it uses some Trackball code). I do agree
however that it could still be a red herring.

> all be done single threaded, there shouldn't be any race condition,
> for instance the call to handle() should certainly be single threaded.
> 
> So.. this leaves me wondering if the GraphicsWindow event handling on
> your platform is the source of the problem and something is being
> tripped up due to the frantic event activity.
>
> BTW what platform are you working on?

The only platform I know anything about is Linux. I used Wieblen's Win32
installer this weekend on my XP partition, but I felt dirty inside for a
few minutes after. :) How can _anyone_ get anything done on Windows? :)

> Another note, you won't find me chasing this bug, I have RSI so no
> frantic activity on the mouse.  I'll get Martin to go mad at the
> keyboard mouse though to see if we recreate the problem.
> 
> Item 2) The className() and libraryName() should be public, so I've
> changed this in the GraphicsContext, GraphicsWindow and its
> subclasses.   These changes shouldn't have too much of effect on the
> world though, as you don't generally serialize GraphicsWindow objects.
>  This changes are now checked in.

The only reason I spotted it was from using my ref_ptr<> wrapper, which
attempts to find out what class/lib a failed operator->() comes from.

> Robert.
> 
> On 6/12/07, Jeremy L. Moles <[EMAIL PROTECTED]> wrote:
> > I did a lot of OSG fiddling over the weekend and came up with a few bugs
> > (1 of which is definitely a bug and 1 of which "might be"); I wanted to
> > try and get these in before 2.0 is released, but it may be too late.
> >
> > 1) I can crash any application that uses a Trackball Manipulator (as of
> > svn update about an hour ago) by pressing and releasing the left mouse
> > button _immediately_ followed by a spacebar press; this seems to be a
> > timing/race issue, as it only happens when you do the sequence quickly.
> > Sometimes it takes 5-10 seconds of "doing it", but it will eventually
> > happen. If all else fails, just go crazy pressing both and it'll
> > happen. :) I'll build a debug release of OSG sometime later today and
> > run it under valgrind to see exactly where it happens, but release
> > builds in valgrind say the crash occurs in handle(), which isn't very
> > helpful, as you can imagine...
> >
> > 2) GraphicsContext puts className(), libraryName(), etc. into a
> > protected namespace, but GraphicsWindowEmbedded doesn't override them.
> > Perhaps this is encouragement not to use this class? :) (I know the
> > general idea is to avoid it's use, so this might be on purpose...)
> >
> > It's unlike me to report bugs without patches, but I've been trying to
> > learn about and get my head around picking this weekend, and haven't had
> > a chance to really investigate these things.
> >
> > _______________________________________________
> > osg-users mailing list
> > [email protected]
> > http://openscenegraph.net/mailman/listinfo/osg-users
> > http://www.openscenegraph.org/
> >
> _______________________________________________
> osg-users mailing list
> [email protected]
> http://openscenegraph.net/mailman/listinfo/osg-users
> http://www.openscenegraph.org/
> 

_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/

Reply via email to