I suspect this is a feature of the VM SSL implementation to close the inbound socket when the application closes the outbound connection, discarding any data that is in transit outside or in the SSL path. That happened initially also in the other direction, but I was able to change some minds because you have little control over when the other side closes the socket. For the outbound socket, we can avoid closing it early with a FANIN and ELASTIC and waiting for the inbound connection to close. Like this: ... | i: fanin | tcpclient ... | .. hole | elastic | i:
A working example is here: https://gist.githubusercontent.com/rvdheij/fb03574799200c14f636ef7913c753b9 I will have a look why LINGER doesn't postpone closing the socket for the SECURE case. Maybe I left a piece of pipe on the floor. If you remove the feedback to FANIN, you'll notice it does not get anything. Sir Rob the Plumber On Tue, 15 Aug 2023 at 02:16, Donald Russell <[email protected]> wrote: > I also use the linger 10 and timeout 15 options on tcpclient. > > Typo in the original message: should be I CANNOT explain why one works and > the other doesn’t. > > > > On Mon, Aug 14, 2023 at 16:48 Donald Russell <[email protected]> > wrote: > > > I wrote a CMS exec to send an http post request using the tcpclient pipe > > stage. > > > > When I use port 80 everything works as expected. I get the http response > > and so on. > > > > If I change to use port 443 and specify the secure getsecinfo options, > the > > request goes to the web server, the secinfo data shows tls1.2 was > > negotiated ok, the web server log shows the post request complete, but > > tcpclient doesn’t output any of the https response. I get “0 socket > closed”. > > > > Any suggestions for what to look at? I can explain why one works and the > > other not. 🙁 > > > > > > > > >
