Re: [Monotone-devel] Approval revisited...

[Bruce Stephens]

Richard Levitte - VMS Whacker <address@hidden> writes:

[...]

> I did a little bit of experiment, and found out that I had
> misunderstood what the heads of a branch with a disconnected graph
> would be.  That was basically my worry with this scheme.
>
> My mantra for the night: experiment a little first, ask questions
> later.

Well, I did a quick experiment, setting up a branch "a" with a few
revisions (merged).  Then an entirely separate branch "b" with one
revision.  Then I approved h:b onto "a", and "a" now has two heads.

So I suspect if someone wants to add approval to something in
net.venge.monotone.contrib.monotree and their cat steps on enter at
the wrong moment:

      monotone approve ... --branch=net.venge.monotone

then net.venge.monotone gets an unwanted head.

That's probably not an enormous issue: nothing will update to it, for
example.  Doesn't it make net.venge.monotone unmerged, though, and
mess up propagate and stuff?

And if you try and recover by updating to a proper revision in
net.venge.monotone, then "monotone automate selector
h:net.venge.monotone.contrib.monotree > MT/revision; monotone commit;
monotone merge", or something, then anyone wanting to pull just nvm
gets bits of monotree, don't they (because those are now ancestors of
nvm, even though they have no actual input to it)?

Shouldn't we have a way to kill individual certs locally?  (I mean a
proper one, not involving db execute!)

[...]