Hi Florian,
I'm not sure what resolution in the frontend you're referring to that makes additional configuration unnecessary, but it doesn't sound like the same functionality as the plugin provided in 1.x- IDK how that could be automatic (though that would be nice- maybe the Mercurial clone pooling functionality would work, but I suspect the old redirect is simpler and more flexible).
The use case is basically this- 2 top level C projects, and a subrepo of header files named "include". The canonical layout, both locally and in the wire protocol is thus:
project1
project1/include
project2
project2/include
Normally,
project1/include and project2/include are separate clones on the server, so if a user commits to project1/include locally and pushes, somebody pulling
project2/include doesn't see those new commits. The really nice thing about the 1.x plugin was that we configured the include subrepo to point to the same copy on the server:
project1
project1/include (3xx redirect to subrepos/include)
project2
project2/include
(3xx redirect to subrepos/include)
subrepos/include
With that, there's only one clone of the subrepo and any push to
project1/include or
project2/include would be visible when pulling from the other path.
We could get the same functionality by logging into the server and using the share extension, but it would be nicer using the web interface, and it's probably a bad idea to poke around in the directory tree owned by SCM Manager.
Does that clarify the workflow?
Thanks,
--Matt