Comment # 36 on bug 1068470 from Sebastian Parschauer

(In reply to Matej Cepl from comment #35)
> Hi, this is an upstream maintainer of M2Crypto. So, do I understand
> correctly that you are just working around an issue in kernel, or is this a
> bug in M2Crypto? If the latter, may I ask for filing the issue in
> https://gitlab.com/m2crypto/m2crypto/issues/new, please?

Thank you very much for the upstream link and that offer. I've tried multiple
different kernels. osc is the only tool using M2Crypto on my systems. So it
pretty much points to M2Crypto. Closing TCP sockets too early seems to be a
common problem. At:
> https://stackoverflow.com/questions/8874021/close-socket-directly-after-send-unsafe
it is suggested to read from the socket with a timeout before closing it on the
client. The RST in the tcpdump also indicates that there is still something the
server wanted to tell us.

TODO:
* test with receive before close
* test upstream M2Crypto source
* check how wget or zypper handles this