Re: [Monotone-devel] [PATCH] lca selector RFC

[Ludovic Brenta]

Emile Snyder writes:
> The attached patch implements an "LCA selector," ie. it lets you do
>
> mtn diff -r p:net.venge.monotone
>
> to see the changes you have in your working copy vs. the ancestor in
> net.venge.monotone which would be selected if you merged the branch
> you're on with n.v.m.

Wow, great! You've made my day if you solved this problem that I
sumitted a while ago:

http://savannah.nongnu.org/bugs/?18302

> Issues remaining with the patch:
>
> * The name.  'p' seems odd to say the least (ostensibly for branch
> divergence Point).  Would people prefer allowing multi letter selector
> naming, ie. just use lca:<branchname>?  Or just some other single
> character choice?

I proposed 'f' for 'fork' but 'p' is just as fine as far as I'm
concerned.

> * Working copy reliance.  At the moment it find your workspace parent
> revision, the head of the specified branch, and then calls
> find_ancestor_for_merge(parent, branch_head).  Does this make sense?
> Should it try to allow for specifying both sides of the
> find_ancestor...() call like
> lca:h:net.venge.monotone.win32,h:net.venge.monotone (or something?)

In my proposal, I envisioned both cases: inside a working copy, it
would find the ancestor of the specified branch and the one in the
working copy; outside a working copy it would require two branches,
just like you suggest.  I picked ':' as the branch separator, but your
choice is as good as mine.

> * Finds merge ancestor, not propagate ancestor.  It doesn't do the smart
> thing about figuring out if either the left or right revision is an
> ancestor of the other.  I just have to fix that.

OK.

> * Needs some more cases in the test (doing it outside workspace, the
> left or right being ancestor of the other thing, etc..)

Would you like me to test for you? I have several databases with
multiple divergence and merge points...  I'll do that when I can find
the time.

> So, thoughts?  Thanks,
> -emile

Thank _you_!

-- 
Ludovic Brenta.