Big git disk usage in development machines

[yajo]

Copied from Tecnativa/doodba#54:

---

When one deals with many projects based on this one in a development computer, git tends to use too much disk, as you can see in this graph:

This happens because of the way it is getting cloned, downloading code in the setup-devel.yaml environment each time from github or other repos.

It would be great to support having a local mirror of those repositories, so you don't have to download them each time. Besides, when cloning locally, git uses hard links, which leave the same files but uses the same disk space. I think the requirement would be that both must reside in the same partition. Not sure how would that affect a docker-based environment as it is right now.

Otherwise we could provide instructions on how to re-shallow your repositories to remove additional remotes, branches, pointers, and other git garbage, so we can run this command on these repos and get back some disk space.

[yelizariev]

It should be easier to implement on git-aggregator side: acsone/git-aggregator#45

[yajo]

You have to keep into account that, for git to use hardlinks, both folders must be in the same volume. Although it's possibly the case, docker abstracts that away in a manner that might confuse git.

That being said, I agree that git-aggregator is a better place to fix this, but a git module installed in the host won't necessarily be fix inside the container, so I'm pretty sure that we will need some patches here too to benefit from whatever comes into git-aggregator.

[yelizariev]

Do you reference to hardlinks used on git clone --local? WIth git worktree commands there are no hardlinks and .git folder becomes just a one-line file.

[yajo]

Do you reference to hardlinks used on git clone --local?

I never tried that option, but git automatically creates hardlinks when you're cloning from another folder in the same volume, AFAIK. With a normal clone command: git clone ./src ./dst