On Thu, 13 Mar 2014, Kevin Brott wrote: > Running test-exec.sh with the set -x in it shows it trying to run this > command line: > /usr/src/openssh/regress/sshd -f /usr/src/openssh/regress/sshd_config > -E/usr/src/openssh/regress/ssh.log > Yet, watching the process table during that part of make tests shows that > command never actually executed, so the pidfile being tested for never > materializes. > > Simply manually running that command in another window while the test is > waiting for the pidfile, gets past the failure and gets 'make tests' to > cleanly finish to 'all tests passed'. > > So I'm not sure exactly what's going south here, it's almost as if it's not > really running the command, or can't - but it's not throwing any errors as > to why. I installed CentOS 4.8 from some godforsaken mirror site and was able to reproduce this - it turns out to be a race condition in the test itself: adding a "sleep 1" after each "kill" statement makes the test pass. We need a better way of handling this in the tests, perhaps polling for sshd's pidfile being deleted. -d _______________________________________________ openssh-unix-dev mailing list openssh-unix-dev@xxxxxxxxxxx https://lists.mindrot.org/mailman/listinfo/openssh-unix-dev
