Thanks a lot to all who've replied, this mailing list rocks! I'll try to summarize:
Still using subversion, so the hint about a too old git version on stock CentOS7 doesn't matter here.
My apologies, Mark Waite. I didn't know that the maintainer of the changelog url changed, and i guess all of us jenkins users respect your effort to make it happen as good and fast as possible. Since you've asked for help to get an automatic changelog page, i would like to give it a try when you give me more infos about how it should go, and which kind of environment is available. Shell and/or Tcl scripting would be no problem, but don't ask me about groovy.
About personal opinions about M$ file systems: agreed. scp or sshfs would help, but isn't much better. The key point here is, regression test machines in the local environment report results and logfiles to a network file system. Not a real performance issue. Other regression tests do some ftp transfers, and scripts need to check what and how much does get there (via CIFS share). So, instead of starting a JNLP session on each Windows agent manually (with access to a network share), it would be nice to see a solution which runs "jenkins as a service" (currently without error-less solution that has access to a network share). Thanks Simon for your hints, tried again, but still without success. Gives something like 'no access to Windows Event log' and exited. Thanks Simon Richter for your tips. Need to investigate and will report when having success.
About the "Check job prerequisites" jenkins plugin (aka "Slave Prerequisites plugin"... shouldn't this be named "Agent Prerequisites plugin" soon?): I have checked this with a fresh jenkins LTS and latest weekly distribution, and issue2 is reproducible: Select "windows batch command" at "Check job prerequisites" in any job, it is saved ok, reading the job's
xml file. When loading job configuration again, it is displayed as
"shell script" and saved as such (not as "windows batch command"). Ok, i guess i have to create an JIRA account&issue...
Thanks a lot, Ivan Fernandez Calvo, for opening J
ENKINS-60289. It's not a real showstopper, but it is time consuming to manually login to each (of many) ssh clients witthout detecting any issues, just (not) to get rid of a jenkins warning (which possibly sits in some kind of not-updated cache?). BTW, a 'grep sshHostKeyVerificationStrategy nodes/*/config.xml' in CWD=~jenkins reports the expected class="hudson.plugins.sshslaves.verifiers.KnownHostsFileKeyVerificationStrategy" for all nodes.
Thaaanks & Cheers, Peter