Nicolas Rybowski
Nicolas Rybowski
Rebase on 91a3454b9755e38e84e4872ae6286539e1f45b36.
I will add some tests in subsequent commits but in the current state of the PR, the BIFT is correctly populated by the [`holo`](https://github.com/holo-routing/holo/pull/32) control-plane and the duplication as well...
For reference, the recommended solution for this issue is cleanly close the Docker Agent and logging an error. See https://github.com/INGInious/INGInious/pull/1051#issuecomment-2980438214 for more details.
Thanks for the report. Do you have more accurate information about the versions you are talking about? It will be simpler to debug with such information.
@phlummox Thanks for the feedback. The documentation is clearly not up-to-date, we will definitively take into account your comments.
**Open question:** How do we handle task migration with common directories? Suppose we have `courseA/task0` and `courseB`. The first course contains a $common directory: `courseA/$common` and `courseA/task0/$common` whose remote `origin`...
With this push, we moved away from `GitPython` to leverage `pygit2` instead. See https://github.com/INGInious/INGInious/pull/1058#discussion_r2243285240.
After discussing with @anthonygego and @AlexandreDoneux, this PR will remain on hold until the course and task factories are correctly refactored.
The previous error is maybe due to my misunderstanding of submodule handling in libgit2. It seems that a repository within a repository is considered as a submodule even when it...
A few additional comments: - Calling `second.index.write()` after `second.index.add('.gitkeep')` solves the issue of 'first' remaining dirty. - Setting the submodule URL triggers the creation of the `.gitmodules` file but it...