Re: [PATCH v3 11/20] path: construct correct path to a worktree's index

Jonathan Nieder [off-list ref] writes:

Brandon Williams wrote:

quoted

So your suggestion is to completely avoid doing any location when asking
for a worktree_git_path, I guess those code paths which request those
paths should be aware enough that if they need something in commondir to
use git_common_path instead.  My only worry is that it may be difficult
to catch misuse of worktree_git_path during code review, at least that
was one of the motivating factors for originally respecting
GIT_INDEX_FILE and the like.

Correct: I'm saying that when someone calls worktree_git_path, the
intent is to resolve a path within the worktree git directory.  File
relocation just gets in the way of that.

I am not too worried about misuse because the only reason to call
worktree_git_path is to access a worktree-specific file like HEAD or
index.

Until somebody has a brilliant idea "git_path() can be implemented
in terms of worktree_git_path()---give it the current worktree!" ;-)

Just joking.  I agree with the general direction you've shown in the
thread.

Thanks.