Reconfiguring a job to use 'all_nodes' doesn't work

[rhettg]

I had to clear the state file and start over.

It seems that all_nodes works by assuming build_runs() will be called once, then the timer for each job will reschedule the next one. But this doesn't work for reconfigs

[mtytel]

was all_nodes the only thing you changed? If so, then it may have been a job comparison/replacement bug. It thought the new job was identical so nothing was changed

[rhettg]

That's likely all I changed. But also manual start on a job appears to call build_run rather than build_runs... which seemed suspicious.

[mtytel]

Well i was assuming when you manually start it, you only want one. But in retrospect maybe it's better if it runs on all nodes

[rhettg]

I could see picking a node on the command line, but by default I would assume it starts them all.

[rhettg]

Looks like matthewtytel got this fix in