Thanks for the review. I looked into this further. The issue could not be reproduced in the current GCC and I have not found an existing evidence that the affected MIPS code can recieve a non-canonical stack-pointer RTX.
I agree that the patch is not sufficiently justified without such a reproducer, so I will drop it from the series in v4. If the issue resurfaces, we can investigate the actual source of the non-canonical RTX and provide a targeted fix with a testcase. Link to v4: https://sourceware.org/pipermail/gcc-patches/2026-July/724124.html Thanks again, Eldar.
