I spent some time investigating this bug, and my conclusion is that the problem lies elsewhere (i.e., it's not related to http_proxy being set).
In a nutshell, what I found is that the delay used in the "stop_server" procedure is too low for armhf machines. Here's the code in question: 200 def stop_server 201 @server.stop! 202 @thread.kill 203 204 100.times do 205 break unless EM.reactor_running? 206 sleep 0.01 207 end 208 209 raise "Reactor still running, wtf?" if EventMachine.reactor_running? 210 end As can be seen, there's a 1 second delay to let the server actually stop. This seems to be enough in all architectures but armhf. I don't know why the same problem doesn't manifest on Debian. Maybe because they have bare metal armhf servers, whereas we use lxd instances inside arm64 machines? Anyway, I was able to verify that if we extend this delay to, say, 10 seconds, the tests pass on armhf. I also think that's something OK to do, especially because the loop is actively checking if the server has stopped and will break the loop if that's the case. I'll put an MP up soon. -- You received this bug notification because you are a member of Ubuntu Bugs, which is subscribed to Ubuntu. https://bugs.launchpad.net/bugs/1939086 Title: autopkgtest failures due to squid "(111) Connection refused" on armhf To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/thin/+bug/1939086/+subscriptions -- ubuntu-bugs mailing list ubuntu-bugs@lists.ubuntu.com https://lists.ubuntu.com/mailman/listinfo/ubuntu-bugs
