[yocto] Best way to "freeze" Yocto for production

[Cliff Brake]

On Tue, Jun 10, 2014 at 1:31 PM, Marlon Smith <marlon.smith10 at gmail.com> wrote:
> Hi everyone,
>
> We've decided to use Yocto in our company's product!  And I have a question:
>
> Since Yocto is updated pretty frequently, at some point we'll need to freeze
> everything and keep a permanent copy so that we can make small changes/bug
> fixes without having to worry about anything being changed outside of our
> control.  My current theory is that we'll do a bitbake -c fetchall, then zip
> up the entire Yocto directory and save it somewhere so we'll always have a
> static copy.  We'll place our application and custom bsp layer under a
> separate Git repository, and then to do a build we'll just combine
> everything together and run bitbake on the whole thing.

Lately, I've been implementing projects based on yocto releases (dora,
daisy, etc).  Most layers have corresponding branches.  You can use
git submodules or repo to lock everything down.  Philip provided a
repo example.  Here are a couple more:

https://github.com/cbrake/oe-build (uses submodules)
https://github.com/kraj/angstrom-manifest (uses repo)

Custom bits for a project goes in a new layer.  If anything needs to
change in an upstream layer, I typically put a bbappend in the custom
layer, rather than modifying an upstream layer, and then needing to
maintain my own copies of the upstream repos.

I'm sure there are many other ways ...

Cliff