Re: svn commit: rev 525 - trunk/subversion/libsvn_client

cmpilato@collab.net writes:

> Karl Fogel <kfogel@newton.ch.collab.net> writes:
>
> > Mike, pardon the possibly dumb question:
> >
> > Are we protecting against these on the client side merely for faster
> > response time in case an invalid operation is attempted? That is, the
> > filesystem itself also fully protects against these things, right,
> > even if some client tries to do them?
>
> Actually, I had this code working just fine, allowing this:
>
> svn mv -r 5 foo bar
>
> to mean this:
>
> copy revision 5 of foo to bar, delete foo (from head revision, not
> revision 5)
>
> Some chatter on IRC convinced me last night to not allow folks to do
> this, though, since it reads as though we are removing something from
> revision 5.

Yeah. Basically, in the example above ('svn mv -r 5 SRC DST'), the
only valid interpretation is:

  - for the 'cp' part, the revision refers to SRC
  - for the 'rm' part, the revision refres to DST

But ugh. That's so confusing. It's easier to just say that moves
only operate on head trees.

> As for whether or not the filesystem fully protects -- your guess is
> as good as mine. I feel pretty confident that it does, though.

Doesn't the DAG system prevent cycles? It's the 2nd letter in the
acronym, after all. :-)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:36:49 2006