git clone --single-branch includes loose objects, how to avoid?

Question

I have a repo with two branches dist and master. When I do

git clone --no-hardlinks -b dist --single-branch test test-dist

then the resulting clone has the same (large) size as the original repo. Once I run

git gc --prune=all

the single-branch clone "test-dist" becomes much smaller, i.e. the size that I expected in the first place. When doing

git pull

in the single-branch repo afterwards the size isn't blown up again, it stays at the size of the branch I clone.

Am I doing something wrong while cloning the single branch? How do I avoid first doing a copy of the whole history, only to then delete most of it with git gc?

Answer 1

The option you're looking for is --no-local. From the git-clone(1) documentation option for --local (emphasis mine):

When the repository to clone from is on a local machine, this flag bypasses the normal "Git aware" transport mechanism and clones the repository by making a copy of HEAD and everything under objects and refs directories. The files under .git/objects/ directory are hardlinked to save space when possible.

If the repository is specified as a local path (e.g., /path/to/repo), this is the default, and --local is essentially a no-op. If the repository is specified as a URL, then this flag is ignored (and we never use the local optimizations). Specifying --no-local will override the default when /path/to/repo is given, using the regular Git transport instead.

When you use the --local option, either implicitly or explicitly, Git copies everything because it is much, much faster than walking the history and copying only the objects you asked for. You would need to copy most of the pack files anyway, because in most cases, at least one of the objects you've requested will be in a pack file. And if you use hardlinks, the copy is extremely cheap and adds no increased disk space usage, so there's really no reason not to do so.

If you use --no-local, you'll then get the normal transport output, which tells you that Git is really producing an independent pack and running it over the normal protocols.

Do note that it is possible for Git to unpack objects after a clone or fetch even when it uses the normal transport mechanisms; it depends on the number of objects transferred and the configuration. Small fetches will often have their objects unpacked because having many small, independent packs that don't share many objects aren't an efficient use of space; it's better to have the normal git gc process pack them more efficiently in the future.