trunk reliability

Now that the trunk is on its way to rails 2.0, is it still a good idea to run apps off the trunk?

Right now the app generator is broken (and I couldn't fix some of it).

In the past I've run trunk apps with little to no breakage. Will the road to 2.0 be a bit more rocky?

I'm running on edge. There aren't any major breaking changes that I
can think of right now. But as always, you need good unit tests :slight_smile:

However, 2.0 will see some more drastic changes, so you should keep
one eye on the changelogs (or subscribe to the various blogs reporting
on edge features, like http://ryandaigle.com/).

I fixed the app generator: http://dev.rubyonrails.org/changeset/6224

I'm running on edge. There aren't any major breaking changes that I
can think of right now. But as always, you need good unit tests :slight_smile:

However, 2.0 will see some more drastic changes, so you should keep
one eye on the changelogs (or subscribe to the various blogs reporting
on edge features, like http://ryandaigle.com/).

Generally speaking I track stable, only moving to edge if there's a
big feature I'm after. While we'll always do our best not to bust up
edge too badly, you're kinda crazy if you're using it 'just because'
;).

Generally speaking I track stable, only moving to edge if there's a
big feature I'm after. While we'll always do our best not to bust up
edge too badly, you're kinda crazy if you're using it 'just because'
;).

Zing! I actually use edge because of the refactoring to
ActionController::Rescue :slight_smile:

I was using trunk before 1.2 because of the long gap since 1.1. Lots and lots of new useful features.

I've just switched the app I'm working on back to 1.2.2, for now.

I'll keep an eye on ryan's blog and switch to trunk if something cool pops up.

Do you think maybe as work on 2.0 progresses, it might be a good idea to maintain a branch to merge stable new features in, along side with a not-so-stable trunk?

The purpose would be to provide sort a reliable selection of the more stable features, without actually putting out a release. Or just trunk snapshots (tags) at the most stable revisions. Maybe it's too much work, maybe it'll be useful as a way to garner wider testing for 2.0 features before an RC is out.

Do you think maybe as work on 2.0 progresses, it might be a good idea
to maintain a branch to merge stable new features in, along side with
a not-so-stable trunk?

We definitely don't want to have such a long gap between releases
again, to this end we've already done a few things differently. Most
importantly point releases have included low-impact new features like
:conditions=>{:foo=>1..5} and map.resource.

We'll keep that up.

Once 2.0 is feature complete, we should also push more beta gems, to
make it easier to test. However maintaining yet-another-branch is
probably just as much work as stablising for a release. If a feature
can be easily back ported to 1.2.x, then we should probably just do
that.