Hacker News Reader

I have a funny little habit that probably stems from a certain degree of paranoia when working on a decent-sized change. I like to keep a local patch of it, as follows: "git diff > works.patch"

Later on, you can apply it if necessary with "git apply works.patch".

View on HN

You could use stash for just that use case (even multiple times to make multiple "patches").

For a more robuts method, you could also just commit (without pushing of course). Then, you can do "git reset --hard HEAD~1", and you have a pre-patch env ready for more work. When you need to re-do your commit, you can check reflog to get the ID.

Either way is, to me, an easier way to achieve the same result as patches.

Instead of looking up the commit hash in your reflog, you could also just create a branch on that commit and refer to the commit with the branch name.

Do you find that better than a stash or a branch because it shows up in ls?

Yes, it can be nice as an archival thing. Keep a directory filled with experimental patches you made to certain parts of your code. Maybe branches are better for certain use cases.

I use stash too, but for different purposes. For instance, when I'd like to quickly change branches to fix a bug/issue and then come back to my stuff. Sometimes, I use stash for resolving potentially complicated pre-commit merges as well.

Reply on news.ycombinator.com