Git worktrees are broken, slow and generally not good

So far I have not had a good experience with the new merging processes versus the old checkpoint method, I am using the IDE interface and frequently get that it just cant do things correctly:

Failed to apply worktree to current branch: Unable to read file ‘/Users/dan/.cursor/worktrees/mindlattice-rag/KpKbd/chatbox-v2/next.config.ts’ (Error: Unable to resolve nonexistent file ‘/Users/dan/.cursor/worktrees/mindlattice-rag/KpKbd/chatbox-v2/next.config.ts’)

I then have to manually go and find the file and copy and paste it - this isnt a good experience

5 Likes

Yeah this is actually unusable now - its just lost all of the work in the worktree - wont apply and now I cant do anything with it. Back to Claude

Same thing here. I get:

Failed to apply worktree to current branch: ENOPRO: No file system provider found for resource ‘vscode-notebook-cell:/Users/[username]/.cursor/worktrees/[folder]/EexbJ/[file].ipynb#W2sZmlsZQ%3D%3D’

Yes same issue. Feature not usable then

Same on this site.

Failed to apply worktree to current branch: Unable to read file ‘/Users/juanpablorti/.cursor/worktrees/[repo_folder]/qdfzU/[repo_folder]/[repo_path]/get_endpoint_documentation_tool.py’ (Error: Unable to resolve nonexistent file '/Users/juanpablorti/.cursor/worktrees/[repo_folder]/qdfzU/[repo_folder]/[repo_path]

Facing the same issue

the same here

Same issue. This is very frustrating.

yes same issue here

Failed to apply worktree to current branch: Unable to read file

I’m considering abandoning Cursor all together because of this…

This is huge issue that is more than month old with no fix or simple official response.

At least have feature flag to disable the feature while you make it work correctly.

I just lost few hours of work because of this issue. And now I afraid to ask it anything of fear that will loose my work / context at anytime and waste my time.

I’m not going to talk about running multiple parallel queries (especially if running expensive models On-Demand..) without user consent and increasing token usage without good reason..

Today I finally tried the worktree feature. Like a 5yo kid who just got his new bike, I thought this is def the coolest thing out there. Launched 3 instances, 2 worktree and 1 on the main branch. The main branch agent had the most difficult task to complete.

Post 3 hours, the worktree are still struggling with basic cleanup requests while I’ve shipped a complete new feature from the main branch.

Not sure what’s the underlying problem. There were a lot of explanations around bloating code and branches, etc. But couldn’t find any good explanations.