When opening a branch in a worktree and running gh stack init, it does not keep track of the stack locally. gh-stack keeps stack membership in local state tied to the checkout where init ran. If I create a worktree & use gh-stack, the stack doesn't persist on my machine, even if I open the branch back in my main worktree. Inside a worktree, $GIT_DIR is .git/worktrees// — so worktree remove deletes the stack state with it
This is confusing & not obvious. Additionally, a modern LLM workflow may involve different sub-agents in different worktrees. This makes gh-stack worse than, say git-spice for this purpose, since I can't easily update my work if I checkout the branch on my main worktree.
The stack is stored remotely on Github, but THAT doesn't always work - seems like GitHub only records a stack for a chain of two or more PRs.
When opening a branch in a worktree and running
gh stack init, it does not keep track of the stack locally.gh-stackkeeps stack membership in local state tied to the checkout whereinitran. If I create a worktree & usegh-stack, the stack doesn't persist on my machine, even if I open the branch back in my main worktree. Inside a worktree, $GIT_DIR is .git/worktrees// — soworktree removedeletes the stack state with itThis is confusing & not obvious. Additionally, a modern LLM workflow may involve different sub-agents in different worktrees. This makes gh-stack worse than, say git-spice for this purpose, since I can't easily update my work if I checkout the branch on my main worktree.
The stack is stored remotely on Github, but THAT doesn't always work - seems like GitHub only records a stack for a chain of two or more PRs.