Hi Amadeusz, On Sat, Nov 02, 2019 at 10:59:44PM +0100, Amadeusz Sławiński wrote: > When window dies we should wait for child process to die, otherwise > there is no one to reap it. > > So emove WNOHANG in waitpid(), which may cause screen to not wait for > child to die and kill window, leaving zombie process. > > Bug: 25089 > > Signed-off-by: Amadeusz Sławiński <am...@asmblr.net>
Thanks, much appreciated. This issue seems to happen only under very specific circumstances (details are unclear, but VMs or otherwise occassionally slow machines seem to be most prone to run into it), but if it happens on a host, it happens quite often. I'm running this patch now for several weeks in production on a host where I reproducible could provoke this bug; and so far no single zombie has been noticed anymore. (And we monitor that condition with alerts — that's how we noticed it in the beginning. :-) So thanks again for the patch and for applying it in git, too. Any estimate when this will be in part of a release? I'm wondering if I should cherry-pick this commit for the Debian package of GNU Screen. Kind regards, Axel -- PGP: 2FF9CD59612616B5 /~\ Plain Text Ribbon Campaign, http://arc.pasp.de/ Mail: a...@deuxchevaux.org \ / Say No to HTML in E-Mail and Usenet Mail+Jabber: a...@noone.org X https://axel.beckert.ch/ / \ I love long mails: https://email.is-not-s.ms/