On Sun, 13 Sep 2009, Alok Aggarwal wrote:
> Since the code changes proposed are significant, I've
> scheduled a code walk through to facilitate the review
> process.
Thanks to everyone who showed up for the code walk through.
For those of you that were present (and that were not), who
would be able to commit to reviewing the code?
A few things to keep in mind while reviewing the code -
a) You'll notice that the DC manifest for building the AI image,
enables only one instance of filesystem/root and then the
live-fs-root-minimal method enables/disables the appropriate
instances. This was done to satisfy the SMF requirement of it
needing to have a clear path to reaching the console-login
service. That is, SMF requires *a* way to get to console-login
where all the intermediate services must be enabled otherwise
it execs sulogin and drops you into the console.
The legacy installer gets away with this because startd doesn't
perform this check if some of the legacy installer services are
the terminal end points in the dependancy graph.
b) You'll also notice the specification of the "dependent" tags in
the live-root-fs-minimal.xml manifest as opposed to the
"dependancy" tags in live-root-fs.xml. This was done to work around
6517953 multiple manifests delivering the same service/instance can stomp
all
over each other
http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6517953
c) The changes for adding a timeout to AI and the transfer module are
not finished. I will be posting an updated webrev with the finished
changes later today or early tomorrow and will send out an email
when that's done.
Alok
