Date: Wed, 1 Apr 2020 16:31:01 +0200 From: Kamil Rytarowski <n...@gmx.com> Message-ID: <dbc67b9e-b59f-d272-f4a2-ee80b9cfe...@gmx.com>
| Does it look better? | | http://netbsd.org/~kamil/patch-00244-fopsmapper-PAGE_SIZE.txt Apart from it needing to be (expressed in the relevant make syntax, whatever that is) if (WARNS > 3) WARNE=3 rather than simply WARNS=3 then as a hack type fix that looks fine. But this is really a gcc bug, we shouldn't be needing to change our sources at all because of this. The signed vs unsigned comparison warning is because if the signed number happens to be negative, when treated as unsigned it will compare larger (usually) than the unsigned value - whereas most people expect negative numbers to be less than positive ones (whatever the magnitudes). Warning for that potential problem is entirely reasonable. But here we have a known positive (though signed) constant. It can never magically become negative, and so the problem can never exist. There should not (ever) be a warning for this case, it is simply stupid, and we should not need to modify either code, or makefiles, to make it go away. kre