Wietse Venema <wie...@porcupine.org> wrote: > Wietse Venema: >> This is a site-specific problem. I ran "openssl s_client" and >> "posttls-finger -w" against one of the affected servers, and reliably >> crashed their postscreen daemon. I've been doing similar tests >> against my own servers without any problems. > > That was with FreeBSD 13.0, as well as Fedora Linux. > > I was able to reproduce the problem with FreeBSD 13.1 RC4 (a release > candidate) from ftp.freebsd.org, with the command > > openssl s_client 127.0.0.1:25 > > This was a Postfix problem that was exposed by a more aggressive > compiler. See patch below. > > The real fix, already available with the Postfix 3.8 unstable > release, is to roll back the change from 20211023 that resulted in > this confusing code, so that the mistake cannot be made again. > > The code confused both me (so that I made the mistake) and the > compiler (so that it did not report a missing initialization).
Wow! I want to thank you very, very much for your efforts in identifying and solving this issue so quickly. Your dedication and help is very much appreciated! Thanks and regards, Michael P.S. Now I do know why recompiling base, kernel, and port with clean ccache still throws 'signal 11' ;-) As I am following postfix-current, I will soon run a fixed postscreen.
