Re: [GIT PULL afs: Development for 5.4

[Linus Torvalds]

On Thu, Sep 19, 2019 at 2:49 AM David Howells <dhowells@xxxxxxxxxx> wrote:
>
> Actually, waiting for all outstanding fixes to get merged and then rebasing
> might not be the right thing here.  The problem is that there are fixes in
> both trees: afs fixes go directly into yours whereas rxrpc fixes go via
> networking and I would prefer to base my patches on both of them for testing
> purposes.  What's the preferred method for dealing with that?  Base on a merge
> of the lastest of those fixes in each tree?

If you absolutely *have* to have something from another development
tree, that's generally a sign that something is screwed up with the
model in the first place, but when it happens,  you should make sure
that you have a stable point in that development tree.

You might ask the upstream developer (ie Davem, in the case of the
network tree) what would be a good point, for example. Don't just pick
a random "tree of the day".

The same very much goes for my tree, btw. You should simply never just
pick a random tree of the day as your base for work if you start with
my tree. That's true whether you do a merge or just start new
development on top of some point, or anything else, for that matter.

Generally, you should never merge other peoples code without having
them _tell_ you that some particular point is a good thing to merge.
Releases are obviously implicitly such points, but generally
cross-tree merges need communication (a pull request to upstream is
the obvious such communication, but not necessarily the only one:
we've had cross-tree development that has involved separate branches
and just various synchronization emails between the two groups).

Looking at rxrpc in particular - if that is what you were waiting for
- it looks more like you should just had an rxrpc branch, and asked
David to pull it for the 5.4 merge window. Then you could have used
that branch itself, as a starting point, perhaps. Or - better yet,
perhaps - merged it into your development tree based on a good AFS
starting point, with a *big* merge message explaining what you are
merging and why.

Right now there is a merge with absolutely no explanation for why the
merge exists at all, and with some very non-obvious bases that really
look like they are just random points of development for both me and
for Davem.

              Linus