Have you merged debian/patches/* ?

[Ian Jackson <ijackson@chiark.greenend.org.uk>]

If you have done a nontrivial merge of debian/patches/*, I'd like to
hear from you.  This is because I have a theory about how this can be
done in a way that does not involve editing, or worse, resolving
conficts in, diffs.

I would like some test cases for experimenting with various algorithms
to see what works well.  I also want to gain an understanding of what
kinds of patch queue merge situations occur in the real world, so I
can focus on algorithm(s) which handle those cases properly (perhaps
at the expense of giving up on more unusual cases).

For each case I will need:
  * left, right: The two diverged patch series that were merged
  * result: The merged version that you created
  * ancestor: The common ancestor patch series that each of the two
    patch series versions were derived from

Forms I can easily consume are:
  * Debian `3.0 (quilt)' source packages (eg refs to snapshot.d.o)
  * git-buildpackage patches unapplied trees with the queue
    in debian/patches/*, assuming the patch queue is up to date

Less easily [1] I can consume:
  * git-dpm git branches (presumably, naming one commit for each
    of the merge-base, left, right, and result queues)
  * bare git branches if you tell me which commit ranges to look at
  * bzr or svn branches containing patches-unapplied
    `3.0 (quilt)' source package trees, again with up to date
    patch queue in debian/patches/*
  * bare-packaging vcs branches containing quilt patches,
    if you can point me to the corresponding upstream (eg tell me
    for each patch queue version which orig tarball the series
    was against).

I'm interested both in situations where you did not have to manually
resolve conflicts, and ones where you did.  If you had to manually
resolve conflicts then feel free to say something about what were the
worst aspects of the experience.  In general I'd like to know what
tools you used for this merge and what yuu thought ofo them.

I'm only interested in merges which did reconciled two patch *queues*
- that is, where two similar stacks of patches, each derived from a
common ancestor stack of patches, were merged into a single stack of
patches very like the original but with both sets of patch queue
changes.  (Sorry if this is too opaque.)

I _am_ interested in situtions where the upstream (the patch series
baseline) changed at the same time as the merge (if you were unlucky
enough to have to deal with this).

I actually don't mind if the patches are of Debian packages.  So if
you have done a nontrivial merge of patch queues in some other setting
please also let me know.

Please send references to your examples by private email to
    Ian Jackson <igs6xj@fyvzl.net>

When I have enough examples I'll follow up here to tell you to stop,
or maybe to say "I have enough ones like this".

[1] FYI would convert everything to gbp pq patch-queue format.

Ian.

-- 
Ian Jackson <ijackson@chiark.greenend.org.uk>   These opinions are my own.

If I emailed you from an address @fyvzl.net or @evade.org.uk, that is
a private address which bypasses my fierce spamfilter.