Working out outdated migrations (PR #11)

Several migrations were failing when running db:migrate at the initial release version from an empty database. These changes took out statements that were causing migrations to fail, usually because of references to models that no longer exist.

Understood that the vagrant vm is kinda unofficially the only thing recommended at this stage, but it was trivial to get db:migrate to move with no problems, and all tests green from a working rvm/ruby and postgres.app on OS X 10.7.

Added CREATE EXTENSION hstore to the meta data migration, as that was missing when running postgres via postgres.app.

GitHub

I really hate to be a stickler but we really need a CLA https://github.com/discourse/core/blob/master/CONTRIBUTING.md we can’t even look at PRs without it

If the CLA process is too much of a pain can you raise on http://meta.discourse.org

Totally plan on it; no worries. No access to proper equipment until later today, will get it sorted in a few hrs.

We just updated our CLA to the much more sane https://docs.google.com/a/discourse.org/spreadsheet/viewform?formkey=dGUwejFfbDhDYXR4bVFMRG1TUENqLWc6MQ

If you fill it I can review this.

Yeah, anyone is going to be making these same trial & error changes to get it running from scratch.

A further solution is to collapse as much of the older migrations into a starting schema that can be loaded. Then we can migrate from there.

Yeah I hate the fact we have migrations people can not run, it defeats the purpose. we should collapse all but 1 week ago old ones, and we should never call models from our migrations, just pure wonderful SQL

@SamSaffron posted my OK through the form mentioned above. Thanks for your patience.

sorry, pr still is in unmergable state, mind fixing it ?

Sure; lemme get up to speed with things.

Rebased; should be in a merge-able state now (I hope)

K that looks fine, we lose a bit of a historical record, but to be honest we should not have been moving around data that way