|
Mark, Indeed, the original test (SSH with keys switched to http(s) with passwords) failed. But I'm not saying it has to do with the code of the Jenkins plugins. Situation 1 uses an SSH key. Hence, the ssh call to expand the submodule only disposes of the userid/password combination that worked fine on the main http(s) call. But the ssh engine does not share the same user an psw db with the BitBucket http(s) server on my system, which explains the failure. I would even say more: the other way arround would certainly never work: if the main repo uses an SSH key, but the submodule is referenced as a http(s) link, then it can never work (unless the http(s) server has no security) simply because an SSH key would not work on a http(s) link AFAIK. Given that the git protocol nor the local file protocol require any kind of authentication, these 2 are unaffected by this ticket.
Thus if keys are used, all repo access MUST use the SSH protocol otherwise the git client has no credentials at all to use. This gives the below overview:
Auth type: SSH KEY main repo=ssh, submodule=ssh : tested OK (my first test report)
Auth type: User & PSW main repo=http(s), submodule=http(s) : tested OK (my second test report) main repo=ssh, submodule=ssh : to test main repo=http(s), submodule=ssh : to test main repo=ssh, submodule=http(s) : to test
Has anybody validated any such combination yet?
|