BGLView /OpenGL resize bug when direct mode is enabled

[jackburton]

Launch GLTeapot, grab the resize knob, and start resizing the window very fast lots of times. After a while the application will hang. Reproducible on VirtualBox, didn't try yet on real hardware.

[the ringmaster]

happens for me as well. When I grab and flail the handle, the teapot window goes black and the window is unresponsive. Rev 40042

[jscipione]

I've noticed the same thing happening if you move the window around wildly as well. Grab the window and move it around vigorously. Eventually the teapot will stop rendering inside the view and the app will crash and need to be killed.

[stargatefan]

heh, does the teapot image still detach from the windo if you resize after you move it ? I was gonna file a bug report for this months ago but opengl is pretty low on the totem pole as I gather.

[jackburton]

I think what happens is that BGLView::FrameResized() and BGLView::DirectConnected() are interacting in bad ways: FrameResized() is triggered asynchronously, while DirectConnected() is called synchronously. And in fact, the bug only shows up when direct mode is enabled. In both these functions we change some member variables which are closely connected and depends one on the other (fBounds and the direct_buffer_info stuff).

[Disreali]

Replying to stargatefan:

heh, does the teapot image still detach from the windo if you resize after you move it ? I was gonna file a bug report for this months ago but opengl is pretty low on the totem pole as I gather.

That has already been reported on #6614.

jackburton, Is it possible that these two issues are related?

[jackburton]

Replying to Disreali:

That has already been reported on #6614.

jackburton, Is it possible that these two issues are related?

Yes, they are definitely related, both are caused by the non-defined order in which DirectConnected() and FrameResized() can be called.

[jackburton]

Reopening, since the symptoms are different (but the cause is the same)

[kallisti5]

Is this still an issue as of R1A4?