On 6/1/23 19:39, Dan Carpenter wrote:

It's a false positive because the test is obviously intended for 32-bit
longs.

I'm not an expert in compiler development, but I do not understand
"obviously intended" in this context. An input literally compares
<any unsigned 32-bit> > <max unsigned 64-bit> / 8, which is always
false, and so the compiler complains. If "obviously intended" means
"the compiler should silently optimize away this check for LP64",
I would disagree, and that's why I would like to see the confirmation
from LLVM/clang developers.

Dmitry