[osg-users] Unwanted culling in 3.6.4 vs 3.5.1

[Robert Osfield]

Hi Anders,

On Thu, 9 Jan 2020 at 14:07, Anders Backman <andersb at cs.umu.se> wrote:

> Problem solved:
>
>     obj->setAutoScaleTransitionWidthRatio(0.001f); // Was 0 earlier.
>
> This seems to be something that changed between the two versions!
> Now it works.
>

Good to hear that you've found a workaround.

There were a number of fixes made to osg::AutoTransform so it looks like
one of the changes has created a regression for your usage case.  This is
why I make so many calls for testing before release go out so we can catch
these cases where the changes are still relatively fresh in out minds.

This commit may have made the code sensitive to a zero
setAutoScaleTransitionWidthRatio :


https://github.com/openscenegraph/OpenSceneGraph/commit/92092a56ae920b41b25b984592d69a7aaba28480#diff-02ae8731c81cbf820759403a17780405

I think this PR addresses a bug associated with using AutoTransforms in
multiple views at one time.  Looking at the code I wonder if the commented
out line (line 153 of src/osg/AutoTransform.cpp):

    //if (_autoScaleTransitionWidthRatio>0.0)

Is what has introduced this sensitivity to a zero value
of_autoScaleTransitionWidthRatio.
The following code block looks like it would provoke a divide by zero with
a _autoScaleTransitionWidthRatio of zero when the i and j values ended
becoming the same value.

I will need to think about what should be happening in the code in your
usage case.  Do you know what was intended with the original settings?  I'm
a bit cold on this code as it's nearly three years since I last worked on
it.

For the upcomming 3.6.5 release I'd like to get a fix checked in to handle
this case.

Cheers,
Robert.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openscenegraph.org/pipermail/osg-users-openscenegraph.org/attachments/20200110/660e8658/attachment.html>