kfogel@collab.net writes:
> Ben Collins-Sussman <sussman@collab.net> writes:
> > cmpilato and kfogel: we had a discussion about this issue last week,
> > something deep and theoretical, and I don't remember the outcome. Can
> > you guys reiterate our current feeling about this bug? (i.e. is it a
> > real bug? why or why not?)
>
> Holy cow. I don't remember what we concluded, but on reading it now,
> I'm not sure even that the answer we proposed in the issue is a good
> one...
>
> I kind of think the undernodes *should* have a CR of 2. Why should
> the versioning filesystem's shared-storage optimization -- a mere
> implementation detail -- affect what we report about a path? Those
> files are new, they did not exist before revision 2, so they should
> have a CR of 2 I think.
>
> And the way you can tell they're not the same as the files they were
> copied from is: if you change one, the other does not change, so they
> must be different objects :-). (There's some technical term for this
> kind of identity test, but I don't remember it offhand.)
>
> The fact that our filesystem had trouble distinguishing two different
> objects when they happen to share a storage key is a bug, but one
> that's maybe fixed by cmpilato's created-path work recently?
>
> Thoughts?
The undernodes certainly *should* have a CR of 2. And 'svn update'
*should* be assigning CR's of 2 to those nodes as well.
The problem is that because our filesystem uses the cheap-copy model,
that causes all sorts of discrepancies between the "node created rev"
and the "path created rev". From a user's perspective, the latter is
the interesting one. But the former is the one that doesn't cost an
arm and a leg to calculate, and that's the only one we (currently)
have available.
I'm working on it, I'm working on it.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Tue Jul 1 20:00:47 2003