Reinhard Brandstädter <r.brandstaedter@gmx.at> wrote on 01/18/2005
12:22:20 PM:
> On Tuesday 18 January 2005 15:34, Mark Phippard wrote:
>
> > Why do you consider this a bug? When I originally tested this, I
thought
> > it was a nice feature. Normally the complaint with svn:externals in
> > Subversion is that commit does not work well with it. The reason that
> > TSVN explicitly prevents the commit is that it does not work due to
the
> > way they use the Subversion API's. The way Subclipse does commits it
does
> > work.
> >
> > I am not clear why we would want to prevent people from committing to
> > externals if we can do so correctly.
>
> Well considering what's written in the subversion book it seems the
Subversion
> design is that you need an explicit commit for external references.
The issue is that a single svn commit cannot commit them both. Subclipse
simply issues multiple commits on your behalf, just as it does if you
select multiple projects in your workspace.
> I see it that way:
> I'm using externals to refer to libraries and other projects (tags).
> Now some of the developers are also responsible for maintaining the
libraries
> (write access to that repository) and also may have write access to
other
> projects' tag directory.
>
> So what happens if the use a other repositories tag in an externals
reference
> and by mistake commit a change they made in the external part of the
working
> copy? (They are using the same user account) Then you have a tag that
was
> modified after tag creation (which I really dislike).
>
> I've seen Tortoise warning about changing a referenced tag...but
subclipse
> doesn't? Maybe it would be enough that eclipse can commit "straight
through"
> but would warn about the tag-externals?
Subclipse issues the same warning as Tortoise. Like Tortoise, it is only
looking at the URL of the item you selected to commit. The difference is
that Tortoise ignores anything that is external.
I do not see any reason we should not allow people to commit to external
folders when we are capable of doing it. A command line user could make
the same mistake you mentioned. If you are truly concerned about it, then
do the recommended "best practice" of adding a pre-commit hook to disallow
commits to the tag.
Mark
_____________________________________________________________________________
Scanned for SoftLanding Systems, Inc. by IBM Email Security Management Services powered by MessageLabs.
_____________________________________________________________________________
Received on Wed Jan 19 05:29:52 2005