On Jul 26, 2006, at 07:38, Gavin Lambert wrote:
> Quoth Tom Vaughan:
>
>> why? why would ci'ing and rm'ing until the revision number is correct
>> fail me?
>
> It might work if *everybody* is guaranteed to be on that exact
> revision
> number, and you manage to get the final revision looking *exactly*
> like
> it did when they last updated. If anyone at all is on a lower or
> higher
> revision, though, or if the revision ends up even slightly different,
> then their working copies are toast.
>
> I could even be wrong about it working in that (very narrow) case --
> there might be something subtle that I've overlooked.
The above was my line of thinking as well. The working copy contains
pristine copies of all files, but those files won't necessarily all
be at the same revision; it is usual for a working copy to be in a
mixed-revision state -- it will be unless an svn update was done on
the root of every working copy before the crash, which I imagine is
unlikely to have occurred.
In order to submit changes to the repository, Subversion constructs a
diff between the pristine local files and the changed local files. If
the pristine local files at those revisions don't match what the
server thinks they look like in those revisions, then the diff that's
sent to the server won't make sense, and who knows what will happen.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Wed Jul 26 12:10:25 2006