Difference between 'git push -f' and 'deleting remote branch and push again'

Question

When I found my wrong mistakes at remote branch, I always delete this, squash locally and push again. Because I heard 'git push -f' was dangerous. However I noticed push result showing after git fetch is some thing like below.

new BRANCH_NAME "(force)"

My workflow deleting and pushing again is correct? What is different between the way and 'push -f'?

Answer 1

Functionally, deleting and pushing again is at least as dangerous as git push -f.

However, deleting and pushing again is actually worse than git -f push for a very specific reason:

At least with git -f push, the replacement is atomic.

Albeit, this isn't a big deal when you're working solo.

1. When you use git -f push, you're told exactly what replacement took place in the resulting message: X...Y branch -> branch (forced update). From there, you can make sure X is really what were meaning to blow away.
2. When you do git push origin :branch && git push origin branch, you can't really tell what you deleted if you aren't absolutely the only person possibly pushing to branch.

Answer 2

No, deleting the remote branch first is not going to make it safer or anything.

The reason forced push is dangerous is due to workflow reason, not technical reasons. If someone pulled your erroneous branch, then you force push the correct change set on the branch, that may cause havoc on their history and on any changes they made on top of the branch because their pulls will no longer be fast forward (unless they fix their erroneous branch with force pull or if they rebase their changes), which could mean that they might reintroduce the erroneous changes again. In a team with just one committers, this usually won't be much of an issue, with larger projects, with lots of committers and lots of watchers, then the likelihood of problem much larger because it may not be just one or two person who've downloaded the erroneous branch.