I have a weird situation. I'm running puppet enterprise, using code deploy with ssh transport to a local git repo. I upgraded my RHEL 7 server to puppet enterprise 2021.7.0.16 a while ago and everything was great except code deploy failed with a core dump due to memory corruption in libssh2. While I have a valid license, it is through our central IT group, and I have to go through them to open a ticket, and long story short I could never get a ticket working for various reasons. I found I could simply copy the system libssh2 libraries into /opt/puppetlabs/server/apps/r10k/lib/ and everything was working fine, and I've not had any problems with that server since. So now I've tried to upgrade my Ubuntu 20.04 server to puppet enterprise 2021.7.1.10 and I've run into the same problem. Copying the system libssh2 libraries doesn't fix it this time (no change). Copying the system libssh2 and libgit2 gets further, but gives another error due to the version of libgcrypt they use. I'm at a standstill on how to fix this, other than I can get it to run (very slowly) by running it in valgrind, but this isn't practical solution. I'm beginning to think since this has happened on two disparate systems now, that it maybe isn't puppet or libssh2 so much as something else specific to my environment (the ssh key used, or the git server's ssh configuration, the git-shell I've implemented there, etc). So, my questions are: Has anyone seen anything like this before? Any ideas on why this is happening? Any solution? Any debugging tips?