Context:
I want to run rubocop in a git pre-commit hook, but I assuming I have a stashed change that would cause it to complain, and an unstashed change that somehow fixes this, the offending change would slip by rubocop.
My solution so far:
Before running rubocop, run git stash --keep-index -u to stash all unstaged changes before running rubocop
The problem:
After running rubocop, git stash pop restores the working tree to its previous state, but stages all those changes (which were previously unstaged)
My question:
Is there an option to pop the stash without automatically staging any of the popped changes?
or
Is there some other way to run a program on the project as it would be committed, instead of the working tree with possible unstaged changes?
To reproduce the situation:
# Set up a dummy repo git init echo foo > file git add foo && git commit -m "Initial" # Make a change and stage it echo bar >> file git add file # Make another change and *don't* stage it echo baz >> file # This is what I later want to restore: git status && git diff && git diff --cached # Stash only unstaged changes (and ignored / untracked files) git stash --keep-index -u # So far everything is as it should be git status && git diff && git diff --cached # Run whatever script on the code rubocop . # This is where things go wrong though: git stash pop # Both changes are now staged to be committed đź’˘ git status && git diff && git diff --cached 
git commit --onlyvsgit commit --include). Usinggit stash --keep-indexdoesn't really get you where you want in all cases. I haven't actually gone and tested this method, but I think the "right" way to test a proposed commit is to usegit checkout-indexto copy the proposed commit to a separate testing area, test it there, and use that result (and never try to update index or work-tree locally).git stash, though, remember to usegit stash pop --indexat the end, so as to keep the distinction between index copy and work-tree copy. Note that you'll have to reset the work-tree first.