Thomas Haller

hm. I am not familiar with this topic. And from your question, I can tell that you already know more than me. A new API is certainly possible, if somebody...

in general, changing behavior (and more importantly: breaking users) would be a problem. I do think that it's unlikely that there are users who depend on this behavior, so we...

Note that since commit 55ea6e6b6cd805f441b410971c9dd7575e783ef4, we use recvmsg() with MSG_PEEK by default, which means that for each message two syscalls are required.

Not tested, because correctness trumps performance. To get the previous behavior, call `nl_socket_disable_msg_peek()` (this also works against older versions of the library). Then it's up to you to ensure that...

I took part of the patch, about the fix for the description. the remainder ``` - err = nl_cache_pickup(sk, cache); + err = nl_cache_pickup_checkdup(sk, cache); if (err == -NLE_DUMP_INTR) {...

The route handling (when putting routes into the cache) has many issues in libnl3. For example, a cache requires some sort of identity for its objects. Specifically, if you can...

I agree this mapping is a problem. I think it's not great, that the integer range of libnl errors overlaps with errno, and that some errno are coerced to libnl...

maybe. I don't know. I guess, if we get `MSG_CTRUNC`, we are already in a bad situation, at least, we `flags` is not `MSG_PEEK`, because then we lost a message....

I agree that this was wrong. We should have kept the errno values separate from our internal error codes. And we should not map different errno to the same nlerror....

I think this is the same issue as https://github.com/thom311/libnl/issues/224#issuecomment-526910349