Re: [egit-dev] Issue with ProjectSetCapability for EGit team provider

[Robin Stocker <robin@xxxxxxxxx>]

Robin Stocker wrote:
> "Tim Webb" <tim@xxxxxxxxxxxx> wrote:
> > I was hoping to get a technical assist/guidance on a problem I've
> > been
> > running into after migrating from CVS to GIT. Specifically, when
> > importing projects using team APIs, subsequent projects are failing
> > to
> > import. I believe the main difference in the tooling is that in our
> > usage, we do one project at a time using the
> > ProjectSetCapability.addToWorkspace(...) with one reference at a
> > time.
> > The first project is successfully imported once the repository is
> > mirrored locally however subsequent projects regardless of
> > restarting
> > Eclipse fail with the following error:
> >
> >
> > Destination directory <path>/core exists. Don't clone
> > [<project-name>]
> > from ssh://<host>:<port><path-to-repo> to prevent data loss.
> >
> >
> > I do not see anything in the JavaDoc for ProjectSetCapability that
> > indicates it should not be re-entrant/reusable, so I am looking for
> > guidance as to whether this is a BZ that I should log against EGit
> > or
> > for suggestions in getting further details on this particular error.
> 
> The way it's implemented right now, it can only be called multiple
> times
> if the repo doesn't already exist. As a work-around you could call
> it just once with all the references at (if that's possible for you).
> 
> > I should also indicate that we have multiple projects in a single
> > repository and that once I let the errors occur, I can correctly
> > "Import Projects" manually from the Git Repository view for the repo
> > that was brought down locally.
> 
> We could add additional logic that checks if a Git repository
> already exists with the given URI and branch, and in that case
> skip the clone step and directly import and connect the project.
> 
> Could you file a new bug report against EGit for that?

For future reference, Tim has filed the bug here:

https://bugs.eclipse.org/bugs/show_bug.cgi?id=366207

Pushed two changes to implement this (see bug).

Regards,
  Robin Stocker