On Mon 05 Mar 2018 05:59:47 PM CET, Max Reitz wrote: > Secondly, I've reverted 3d5d319e1221082 to test this, and I could > reproduce failure exactly once. Since then, no luck (in like 20 > attempts, I think)...
Oh, I see. I was bisecting this and it seems that 1a63a907507fbbcf made this problem more difficult to reproduce. If you revert that one too (in addition to 3d5d319e1221082, that is) then it crashes very easily. I don't think it makes sense to tweak the test ever further, I would simply mention that "this triggers the bug that was fixed by 3d5d319e122 and 1a63a907507fbbcf" or something like that. Berto
