On Tue, Mar 31, 2015 at 2:32 PM, Matthew Jordan mjor...@digium.com wrote:
Hey everyone:
Hi!
Just as a brief status update on the Asterisk Git Migration project,
here is where we stand today, and what the next steps are:
Awesome work so far. :-)
There's a large amount of intervening work that has to be done, and
some questions that still need to be answered. In no particular order:
(1) In test runs of building the Asterisk repo, I've successfully
managed to pull in the team repos that are still 'active'. Since
Gerrit acts as the canonical Git repo, we can set it up so that direct
pushes to team branches are allowed and don't require a code review -
although code reviews can be done if someone desires it. For those who
use team branches a lot, does this sound acceptable?
What's the benefit of trying to keep team branches at all vs. just letting
people use their own git trees hosted on one of the many personal git
hosting options (github or whatever) ?
If it's about several people collaborating on a branch, that makes sense to
me to do as a feature branch in gerrit, but it seems like the exception,
not the rule.
(2) Today, we merge up the branches - 11 = 13, 13 = trunk, etc.
After the move, it looks like the best way to handle the merge process
is going to be to make patches against trunk, then cherry-pick them
back to the currently maintained branches. For long time users of
Gerrit, does this seem appropriate?
Yes.
(3) In Asterisk 13, we pulled in menuselect (yay). On the downside,
Asterisk 11 doesn't have menuselect. What's more, Asterisk 1.8 and 12
are still in security fix mode. We really have two options here:
(a) Update the build scripts to pull in menuselect as they do now
from SVN, and more or less keep the existing process. My fear is that
this may turn into a bit of hackery to keep the Git/SVN lines from
getting confused.
(b) Bite the bullet and just backport menuselect into all the
branches. Moving to Git is going to be a breaking change for people
using Asterisk 11 anyway, and this way things end up in a reasonably
pristine state.
Thoughts?
(b) sounds the least painful overall to me. People consuming releases
shouldn't notice a difference, right?
(4) Asterisk records the SVN revision in each file using the special
keyword $Revision:. This is then registered in a linked list for
retrieval by the CLI/AMI. Unfortunately, Git doesn't support this
concept, as adding data into a file after commit would change the
checksum . It does allow doing this on checkout via the $Id$ keyword,
which may be an acceptable workaround.
That whole thing seems questionably useful, anyway. Just removing it is
another option.
Anyway, the goal is to kick off the move of Asterisk to Git
immediately after we get the next batch of releases out. I'll send out
an e-mail once we know exactly when that is.
\o/
--
Russell Bryant
--
_
-- Bandwidth and Colocation Provided by http://www.api-digital.com --
asterisk-dev mailing list
To UNSUBSCRIBE or update options visit:
http://lists.digium.com/mailman/listinfo/asterisk-dev
