On Tuesday 30 August 2005 10:41, Dustin Kirkland wrote:
When I originally asked David about this patch, he asked me to work
against the latest -mm upstream release. I'll port to another tree, but
I agree with Amy: we should agree on a branch to work from.
I think the patches should be developed to the -mm tree, but we will need to
test them in something more stable. When we did the audit development, we
picked an "in house" kernel so we could make changes and get everyone to test
it. The pure -mm tree may be too unstable to get people to test with it and
use on a daily basis.
My guess is that there is going to be about 30 people working and testing
various pieces. While some people don't mind being at the cutting edge and
can rescue themselves if something goes wrong...not everyone can. We need
something that is a little more stable for a group that big to be working
with.
Also think about bug reporting. When someone finds a bug, we'll need to be on
the same kernel to reproduce it and make fixes.
Andrew also sent a note to lkml saying that too many of the patches he was
getting were being immediately followed up with a correction to it. This
means the patches did not get enough testing before being sent. I really hope
that we test the patches for a little while and make corrections before
sending it upstream.
Speaking of which, you guys probably have no idea about the amount of problems
userspace has had trying to support broken and buggy old kernels. I want to
see whole units of work get sent upstream as a unit as much as possible. For
example, the netlink comm had 2 major revisions while we worked on it. I now
have 3 different communication techniques to support. What's worse, I have no
way of telling which method is being used.
-Steve