seems like coverity scanning tools have changed which is why new defects have been found rather than us introducing new bugs recently.
one of them is a false positive, so i marked it as such -- coverity doesn't see we guarantee a value is within a [0,64] range (get_last_valid_cap) before converting unsigned int to signed int (calling cap_get_flag with `i`).
i sent out a cleanup for parse_size. coverity is correct there's an underflow, but it happens after we no longer use the variable (since it's a `i--` operation), so it doesn't really matter.
i sent out a cleanup for --seccomp-bpf-binary where we checked for NULL that could never happen, but coverity made it think that phantom NULL would be passed to APIs that assume non-NULL.
the other "new" failure is related to libminijailpreload.c and the internal marshal/unmarshal code. coverity says it detected it back in 2021, so not sure why it's saying it's "new". at any rate, it thinks the fd/buffer is untrusted between minijail_to_fd->minijail_from_fd, but that's internal to our library, so i don't see how it could be exploited. so i'll mark it ignored.
-mike