Well it was a borderline case that surprised me too because it worked.
Basically i was able to determine in the screen was rotated (in cores that doesn’t prerotate content) by comparing the original size with the viewport one.
Since this query works only in a pass that scales to viewport, but i needed that info in the early pipeline stage, I stored that info in the alpha channel of “isrotated_pass” (if memory serves me), and then i queried that pass feedback from several passes before.
Since GLcore works as well, I don’t consider this an ‘hack’ or the exploit of a glitch, and is probably as you say, dx11 implementation is not 100% working in this regard.
Granted, this is not a crucial feature, but I’not considering the hypotesis of stripping it from the shader to overcome implementation issues.
And btw, thank you very much to have taken the time to test, I really appreciate it.