On Mon, Mar 17, 2014 at 10:18 PM, John Meinel <[email protected]> wrote: > Thanks for clarifying the process. It does still seem strange that we copy > to mirrors before the official location, but I understand the build process. > I'm not sure if we want to slow down dev releases to wait a day to get > streams updated or not. I like the layout, but if we are doing more than one > release per week, that slows velocity a lot. > So *definitely* stable releases should always wait for > streams.canonical.com. And maybe dev releases should move to that (but > maybe not this week).
Per the release log[1], I do list publishing to streams.canonical.com before the CPCs. Since the process was manual and required builds outside of my control, I deferred the step. 1. Ben has updated the release-juju-public-tools process to watch the control branch juju-qa uses to configure which tools to release. The process will be automatic. The process can be run once when the PPA tools are built, and then again after the new archs are published 2. The process to extract jujuds from the deb *was not* idempotent. Repeated runs produced different check sums. I was concerned that repeated publications of a version would produce tools that contradicts the tools that are cached in each juju environment. I concluded the problem was caused by the random date in the tgz; the files in debs don't have dates. The process now constructs a sane date that is repeatable. I believe we can republish several times a day. [1] https://docs.google.com/a/canonical.com/document/d/1qxNRRw3jblX2IGWjlbFvoyCG8zWK2FxPZlj-u3nHpXM/edit# -- Curtis Hovey Canonical Cloud Development and Operations http://launchpad.net/~sinzui -- Juju-dev mailing list [email protected] Modify settings or unsubscribe at: https://lists.ubuntu.com/mailman/listinfo/juju-dev
