Implement a better heuristic for Git worktrees #21
Conversation
The existing heuristic for handling worktrees assumes that one other worktree contains a ".git" subdirectory with the Git database. This heuristic fails when worktrees are attached to a bare repository.
There was a problem hiding this comment.
How do worktrees work in this case? The “main” repo is bare and the other branches are normal directories?
This change seems to break a couple of tests: https://travis-ci.com/barisione/clang-format-hooks/jobs/234482995
In #13 somebody suggested the use of https://git-scm.com/docs/git-config#git-config-corehooksPath. Maybe we could conditionally use that (conditionally as some older git versions don't have it).
| fi | ||
| fi | ||
|
|
||
| if [ -d "$top_dir/hooks" ]; then |
There was a problem hiding this comment.
What if there's a directory called hooks in the repository?
Correct. Just to make sure that random scripts don't go beserk editing the contents of the .git folder, we store that in a separate place as a bare repo, then add worktrees where the real editing is done.
I glanced at the errors, but it looked complicated so I didn't get a chance to try and debug it - I am traveling for the next two weeks.
By default, that does not return anything (tried it both in a worktree and in a cloned repo). I believe it returns a value if it was set (in order to use the hooks from a custom place.) |
For completeness, I have tested it with git 2.23 on Linux. |
The existing heuristic for handling worktrees assumes that one other
worktree contains a ".git" subdirectory with the Git database. This
heuristic fails when worktrees are attached to a bare repository.