Ben Collins-Sussman wrote:
> On Mar 18, 2005, at 12:44 AM, André Pönitz wrote:
> >
> > So is the fear of client-out-of-dateness the only possible problem
> > or is there something else?
>
> It's much much worse than "out of dateness". Suppose a client commits
> a change to foo.c, and the server silently edits the contents on the
> way into the repository. Then the server says, "committed revision 5".
> The working copy then marks its copy of foo.c as being at revision 5.
>
> Then a different user comes along, and changes foo.c again, creating
> revision 6.
>
> Now the first user runs 'svn update'. His working copy says to the
> server, "I have revision 5 of foo.c." But the working copy and the
> repository now have *different definitions* of foo.c@5, without even
> realizing it. The server sends a binary diff to convert foo.c@5 to
> foo.c@6, but the diff doesn't even apply -- the working copy's file has
> different contents.
I see.
> So the official party line is: pre-commit hooks may not change data
> going into the repository. They may, however, enforce certain
> standards. (That is, if the code doesn't conform in a certain way,
> reject the commit.)
Understood now. Thanks for the explanation.
Andre'
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Mar 22 12:28:34 2005