AW: externals definition: surprising behaviour

Hi,
I consider that behaviour as dangerous.

But a fix could be:

When a wc with externals def. is checked out, the information that it belongs to an externals def. can be stored in the controlfiles under '.svn'. When one is in the wc of some stuff that belongs to an externals def and will do an update or an commit on an externals def. with a fixed rev. it can be blocked.

Regards
Hartmut

-----Ursprüngliche Nachricht-----
Von: Ryan Schmidt [mailto:subversion-2009a_at_ryandesign.com]
Gesendet: Donnerstag, 19. Februar 2009 10:34
An: Schroeder, Hartmut
Cc: users_at_subversion.tigris.org
Betreff: Re: externals definition: surprising behaviour

On Feb 19, 2009, at 03:12, Schroeder, Hartmut wrote:

> I detected a surprising behaviour of externals definition (svn 1.5.5):
>
> - the repository path http://ABC:9090/def/ghi
> has an svn:externals with a value "-r 100 /uvw/xyz xyz"
>
> - checkout of http://ABC:9090/def/ghi
>
> - One can make changes/commits on the working copy of
> http://ABC:9090/def/ghi/xyz,
> and it will take effect on trunk of /uvw/xyz

If there have been no other changes in the place the external points
to, that's true. Otherwise, you would get a message that the
(externals) working copy is out of date, and you would have to "svn
up" the (externals) working copy before being able to commit the
changes.

> , even an update will not
> get
> revision 100 but trunk!

An update in the externals working copy will update the external to
HEAD.

An update in the directory containing the externals definition will
restore the external to its pinned revision.

> But: If I use an explicit revision in my externals I don't want access
> to trunk.
> I think that is a bug cause that may result in instable software.

I believe this behaves as designed. It could probably be improved,
but I'm not sure what specifically should be changed.

------------------------------------------------------
http://subversion.tigris.org/ds/viewMessage.do?dsForumId=1065&dsMessageId=1191167

To unsubscribe from this discussion, e-mail: [users-unsubscribe_at_subversion.tigris.org].
Received on 2009-02-19 11:15:34 CET