Re: [Gnu-arch-users] pure-merge is impossible (was Re: Star merging with 3 branches)

[Tom Lord]

    > From: Andrew Suffield <address@hidden>

You're a very clever guy.

    > [....] Just a little background on how we got to this point. (I
    > don't really have time to get into that discussion right now,
    > unfortunately)

Ok, but this helped:

    > That's what they're for - we can easily spot C2/C3 pairs that are
    > different from the ones already applied for the same C1 changeset. We
    > couldn't do that if they were always mixed in (unless we encode the
    > extra data in the mixed changeset, but that seems an unnecessarily
    > complicated way to achieve the same thing, and requires modifying
    > tla - as described, pure-merge can be implemented as a wrapper).

So the answer is that you _do_ plan to use C2 and C3 mostly by
comparing (somehow) to the analogous C2' and C3' when you merge into
another branch that also merged C1.

We wouldn't need _any_ of these various C2s and C3s around at all but
for the fact that we can't count on having the orignal archive that
contained C1 around (or, on the other hand, if the C2s and C3s aren't
computed automatically from R2 but rather with some additional "hints"
from the user but nevermind that for now).

The nature of that "comparison" and what you do with the results are
left as an exercise for the interested reader, as they say.

So, on that exercise:

It seems like you want to _not_ use S2 (the original tree resulting
from the orignal C1) as Rmid but rather want to absolutely minimize C2
and C3 (whereas using S2 literally would maximize those changesets).

Hence -- you _sort_of_ use S2 but then you (in a straightforward way)
subtract out all the inessential parts of the resulting C3_candidate,
leaving a smaller C3_actual.  After which the computation of C2 is
deterministic and C2 and C3 are (in some sense) minimal.

It's that particular subtracting out that you meant when you talked
about diff munging --- and now it makes perfect sense to me.

That's damned interesting.

-t