SQlite backend series updated

[Nicolas Sebrecht]

[ Please, don't cull the cc list. ]

On Thu, Mar 31, 2011 at 10:56:11AM -0400, Tim Gray wrote:
> On Mar 30, 2011 at 07:28 PM +0200, Nicolas Sebrecht wrote:
> >Merged into pu.
> 
> How 'experimental' is the pu branch?  Does it mainly consist of
> features whose code has been somewhat vetted?  Or is it more along
> the lines of features that are only roughed out?

The pu branch stands for 'proposed updates'. The purpose is to keep
track on interesting features not ready for merging, yet. They may be
several reasons to not merge a feature: breaking API, highly
experimental stuff, known to cause dammage, need code cleanup, pending
the next release cycle, etc.

The pu branch help developers to easy fetch the branch, extract a topic
and work on it without the need to add a remote repository, fetch a
given branch, checkout, update, etc.

I can't say all topics in pu could not be tested. You could do it if you
know what your are doing. If any feature/topic is not merged as is it's
probably for a good reason.

> I ask because I'm more than willing to run off the pu branch if it
> would help testing things out, but if it's really only for
> development and will break my setup left and right, I'd obviously
> rather skip it.

Yes, I think you should skip it. The way I build pu don't even ensure
the result can pass the python interpreter and if it passes you could
seriously dammage your data. Developers must extract the wanted topic
from other by themselves.

The good branch to work on to _test only_ (with no or no good
understanding on the code) is 'next'. I'd say that an often updated next
checkout make testers very high in the new introduced stuff for testing.
This could be broken in some ways even if it should not be _that_ broken
(I run the next branch on by daily data)

-- 
Nicolas Sebrecht