With Azure-Thebes wrapper enabled: Visit Wolfram Alpha and input search terms. Any will do. Then, scroll the page. For example: http://www.wolframalpha.com/input/?i=x^2%2By^2%3D1 I used autoscroll, but should be reproducable without autoscroll. Breakpad IDs: https://crash-stats.mozilla.com/report/index/bp-1216c1c5-048e-4688-86d7-cf65f2120118 https://crash-stats.mozilla.com/report/index/bp-a88e1f30-ee32-4f49-a7fd-240ba2120118 https://crash-stats.mozilla.com/report/index/bp-1611c9de-3d6e-4960-8b01-fdd042120118

This is what I was able to reproduce, it has a more useful stack: https://crash-stats.mozilla.com/report/index/bp-9986771d-34ea-4532-9bf6-af9ef2120118

This brings to attention the fact we actually need another bug for getting the right upload region for an edge/corner for the extend modes case. For now I think we should get this fix in, we don't generally ever need the other functionality for content drawing, just canvas.

Test case?

(In reply to Jeff Muizelaar [:jrmuizel] from comment #3) > Test case? I'm pretty sure this is some scrolling based edge-case. Normally we don't try to draw when something is outside the surface. I had to scroll a bunch to get this to occur. Maybe there's a way to reproduce it in a simple environment, but I don't know what it is.

Would this be easier to test if we had an azure test suite?

Yes it would.

Similar Crash (with nVidia in my case: nvwgf2um.dll@0x281db0)scrolling Wolfram and this one: http://www.overclock.net/t/149879/howto-nvidia-bios-flashing/490 Azure-Thebes wrapper enabled, (Crash only with Direct3D 10 layers, Direct3D 9 layers dont crash) https://crash-stats.mozilla.com/report/index/bp-9511e8be-8d80-4ae3-834e-9e9ab2120124 https://crash-stats.mozilla.com/report/index/bp-9ba17c6a-3757-4546-8dea-f740d2120125

(In reply to Bas Schouten (:bas) from comment #6) > Yes it would. Would you be willing to write the Azure test even if we don't have a framework for running it. I'd like us to start having an Azure test suite for testing the lower level semantics of the backends and for being able to run tests without having to rebuild the browser.

(In reply to Jeff Muizelaar [:jrmuizel] from comment #8) > (In reply to Bas Schouten (:bas) from comment #6) > > Yes it would. > > Would you be willing to write the Azure test even if we don't have a > framework for running it. I'd like us to start having an Azure test suite > for testing the lower level semantics of the backends and for being able to > run tests without having to rebuild the browser. I want to do that, I actually want to do the framework very, very soon. I'll make this the first test in there.