PVS-Studio delved into the FreeBSD kernel

Spotted some mistakes even in that article:

In acpi_package.c 83 they write that the fixed version is "bzero(dst, sizeof(*dst));", but dst is a void pointer. That won't work either.

And the "This is a nonsensical comparison: pointer >= 0. geom_vinum_plex.c 173" explanation looks suspicious. Even if sdno is not accessed later, it was just calculated - having an assert on it may be a very good idea. The issue is that the pointer is being checked, not the calculated value.

It's also cool to see that clang already warns about a lot of those situations. Especially most of the logic / signedness issues are caught and reported.