No subject

(more) reasonable if I were employed by Canonical or had a large
amount of time to intimately get to know the inner workings of Bazaar
with the ultimate aim of becoming a core committer or something.
That's not the case here.

Someone in my position needs to be able to rely on the tests to prove
whether they've gone wrong or not, and they will probably do things in
other ways, maybe other "abstraction layers", than what you as an
experienced person will do. The question is whether you then accept
that and make a note to refactor when you have time, or stonewall them
out. In the latter case, you obviously reinforce the impression that
this is a Canonical project, not a FOSS project.

> I don't see your fix as limited at all. =A0I see it as a sweeping change.

> If you had tried to fix just add, I think your change would have landed
> already.

Well, wait a minute. I wrote a test for the specific usage of add in
the bug I wished to fix, made it work, and ran 17000 tests. All clear,
let's go, I thought.

Since then you've raised a spurious benchmark, much discussion of
abstraction layers, discussion of a completely different bug involving
links between different working trees, a pathological edge case
involving symbolic link cycles (which does work today) and another
pathological case above - adding via an outside link which points
directly at a repository file - and this last case doesn't work with
today's code anyway. Your latest comment now indicates you won't
support my change even when you can't think of any more pathological
cases, because there still might be some unforeseen ones somewhere...
and then you tell me it's me that wants to fix everything at once!

/Geoff