On Thu, May 14, 2020 at 06:20:17PM -0700, Alexander Vasarab wrote: > The plot thickens. > > In a nutshell, the problem is "solved" by changing the virtual(8) hosts
[ That's virtual(5) by the way, but not important. ] > to use UNIX domain sockets (rather than e.g. 127.0.0.1). In both cases, > the connection settings are for a local PostgreSQL database. > > While debugging in gdb, I noticed that when it would break on > SSL_shutdown, the backtrace showed a bunch of pgsql stuff. Twice in > a row, always. That was enough of a clue for me. Please build and test the latest commit from the tls-debug branch in my git repo, you should find it works regardless of how you access Postgres... -- Viktor.