[Gnu-arch-users] Smart Merging with Multiple Branches From a Common Ancestor

[John Goerzen]

Hi,

I'm going to try to explain this question that's a little fuzzy in my
head as clearly as possible.  Please ask if I fail :-)

Let's say I have a branch out there called mainline:

base-0
patch-1
patch-2

Now, let's say my branch foo works like this:

base-0:     tag of mainline patch-1
patch-1
patch-2
patch-3:    contains mainline patch-2 [not tagged, but copied
            with a trick like tla set-tree-version -- that is,
            it does preserve the patch history]
patch-4
patch-5

Now, there's also a branch bar that contains unrelated
(non-conflicting) changes to the mainline:

base-0: tag of mainline patch-2
patch-1
patch-2
patch-3

Now, let's say I yet another developer.  I want to create a branch baz
that contains the original plus the changes on foo and bar.  So, after
setting up my archive, I might do this:

tla tag whatever--mainline--1 whatever--baz--1
tla get whatever--baz--1 wc
cd wc
tla star-merge whatever--foo--1

What I want to do with that star-merge is merge in the changes on the
foo branch into mine.  But it only complains that the two trees are
unrelated (despite the fact that they do have a common ancestor).  I
could use repay --exact to pull in certain patches, but that's a
hassle since I'd have to do patch-4 and patch-5 manually.  What should
I be doing?

Now I'd like to also merge in bar.  What should I do here?

What happens when mainline reves but nobody else has?  Could I
star-merge it?

Thanks,
John