How to clone git repository with specific revision/changeset?

How can I clone git repository with specific revision, something like I usu开发者_StackOverflow社区ally do in Mercurial:

hg clone -r 3 /path/to/repository

---

$ git clone $URL
$ cd $PROJECT_NAME
$ git reset --hard $SHA1

To again go back to the most recent commit

$ git pull

---

To save online (remote) the reverted commit, you must to push enforcing origin:

git push origin -f

---

UPDATE 2 Since Git 2.5.0 the feature described below can be enabled on server side with configuration variable uploadpack.allowReachableSHA1InWant, here the GitHub feature request and the GitHub commit enabling this feature. Note that some Git servers activate this option by default, e.g. Bitbucket Server enabled it since version 5.5+. See this answer on Stackexchange for a exmple of how to activate the configuration option.

UPDATE 1 For Git versions 1.7 < v < 2.5 use git clone and git reset, as described in Vaibhav Bajpai's answer

If you don't want to fetch the full repository then you probably shouldn't be using clone. You can always just use fetch to choose the branch that you want to fetch. I'm not an hg expert so I don't know the details of -r but in git you can do something like this.

# make a new blank repository in the current directory
git init

# add a remote
git remote add origin url://to/source/repository

# fetch a commit (or branch or tag) of interest
# Note: the full history up to this commit will be retrieved unless 
#       you limit it with '--depth=...' or '--shallow-since=...'
git fetch origin <sha1-of-commit-of-interest>

# reset this repository's master branch to the commit of interest
git reset --hard FETCH_HEAD

---

To clone only one single specific commit on a particular branch or tag use:

git clone --depth=1 --branch NAME https://github.com/your/repo.git

Unfortunately, NAME can only be branch name or tag name (not commit SHA).

Omit the --depth flag to download the whole history and then checkout that branch or tag:

git clone --branch NAME https://github.com/your/repo.git

This works with recent version of git (I did it with version 2.18.0).

---

Cloning a git repository, aptly, clones the entire repository: there isn't a way to select only one revision to clone. However, once you perform git clone, you can checkout a specific revision by doing checkout <rev>.

---

You Can use simply git checkout <commit hash>

in this sequence

bash git clone [URLTORepository] git checkout [commithash]

commit hash looks like this "45ef55ac20ce2389c9180658fdba35f4a663d204"

---

If you mean you want to fetch everything from the beginning up to a particular point, Charles Bailey's answer is perfect. If you want to do the reverse and retrieve a subset of the history going back from the current date, you can use git clone --depth [N] where N is the number of revs of history you want. However:

--depth

Create a shallow clone with a history truncated to the specified number of revisions. A shallow repository has a number of limitations (you cannot clone or fetch from it, nor push from nor into it), but is adequate if you are only interested in the recent history of a large project with a long history, and would want to send in fixes as patches.

---

Just to sum things up (git v. 1.7.2.1):

1. do a regular git clone where you want the repo (gets everything to date — I know, not what is wanted, we're getting there)
2. git checkout <sha1 rev> of the rev you want
3. git reset --hard
4. git checkout -b master

---

TL;DR - Just create a tag in the source repository against the commit you want to clone up to and use the tag in the fetch command. You can delete the tag from the original repo later to clean up.

Well, its 2014 and it looks like Charles Bailey's accepted answer from 2010 is well and truly outdated by now and most (all?) of the other answers involve cloning, which many people are hoping to avoid.

The following solution achieves what the OP and many others are looking for, which is a way to create a copy of a repository, including history, but only up to a certain commit.

Here are the commands I used with git version 2.1.2 to clone a local repo (ie. a repository in another directory) up to a certain point:

# in the source repository, create a tag against the commit you want to check out
git tag -m "Temporary tag" tmptag <sha1>

# create a new directory and change into that directory
cd somewhere_else;mkdir newdir;cd newdir

# ...and create a new repository
git init

# add the source repository as a remote (this can be a URL or a directory)
git remote add origin /path/to/original/repo

# fetch the tag, which will include the entire repo and history up to that point
git fetch origin refs/tags/tmptag

# reset the head of the repository
git reset --hard FETCH_HEAD

# you can now change back to the original repository and remove the temporary tag
cd original_repo
git tag -d tmptag

Hopefully this solution keeps working for a few more years! :-)

---

No need to download the whole history, and no need to call git init:

git clone --depth=1 URL
git fetch --depth=1 origin SHA1
git checkout SHA1
git branch -D @{-1}  # if you want to tidy up the fetched branch

This has the disadvantage, to CB Baileys answer, that you will still download 1 unnecessary revision. But it's technically a git clone (which the OP wants), and it does not force you to download the whole history of some branch.

---

# clone special tag/branch without history
git clone  --branch=<tag/branch> --depth=1 <repository>


# clone special revision with minimal histories
git clone --branch <branch> <repository> --shallow-since=yyyy-MM-ddTHH:mm:ss  # get the commit time
cd <dir>
git reset --hard <revision> 

you can't get a revision without histories if not set uploadpack.allowReachableSHA1InWant=true on server side, while you can create a tag for it and clone the special tag instead.

---

I was able to accomplish this using the git clone --config option, which I learned from this answer: https://stackoverflow.com/a/43759576/1330650

My scenario involves a sparse checkout in an Azure DevOps pipeline, where I need to clone a repo using a commit hash, not a branch name. The clone command doesn't accept a commit hash as a parameter. The workaround is to set a configuration variable (-c) containing a refspec, because that refspec can use a commit hash instead of a branch name:

git clone -c remote.origin.fetch=+<commit hash>:refs/remotes/origin/<commit hash> <repo_url> --no-checkout --progress --depth 1
git sparse-checkout init --cone
git sparse-checkout set <file list>
git checkout <commit hash>

---

Using 2 of the above answers (How to clone git repository with specific revision/changeset? and How to clone git repository with specific revision/changeset?) Helped me to come up with a definative. If you want to clone up to a point, then that point has to be a tag/branch not simply an SHA or the FETCH_HEAD gets confused. Following the git fetch set, if you use a branch or tag name, you get a response, if you simply use an SHA-1 you get not response.
Here's what I did:- create a full working clone of the full repo, from the actual origin

cd <path to create repo>
git clone git@<our gitlab server>:ui-developers/ui.git 

Then create a local branch, at the point that's interesting

git checkout 2050c8829c67f04b0db81e6247bb589c950afb14
git checkout -b origin_point

Then create my new blank repo, with my local copy as its origin

cd <path to create repo>
mkdir reduced-repo
cd reduced-repo
git init
git remote add local_copy <path to create repo>/ui
git fetch local_copy origin_point

At that point I got this response. I note it because if you use a SHA-1 in place of the branch above, nothing happens, so the response, means it worked

/var/www/html/ui-hacking$ git fetch local_copy origin_point
remote: Counting objects: 45493, done.
remote: Compressing objects: 100% (15928/15928), done.
remote: Total 45493 (delta 27508), reused 45387 (delta 27463)
Receiving objects: 100% (45493/45493), 53.64 MiB | 50.59 MiB/s, done.
Resolving deltas: 100% (27508/27508), done.
From /var/www/html/ui
 * branch            origin_point -> FETCH_HEAD
 * [new branch]      origin_point -> origin/origin_point

Now in my case, I then needed to put that back onto gitlab, as a fresh repo so I did

git remote add origin git@<our gitlab server>:ui-developers/new-ui.git

Which meant I could rebuild my repo from the origin_point by using git --git-dir=../ui/.git format-patch -k -1 --stdout <sha1> | git am -3 -k to cherry pick remotely then use git push origin to upload the whole lot back to its new home.

Hope that helps someone

---

My version was a combination of accepted and most upvoted answers. But it's a little bit different, because everyone uses SHA1 but nobody tells you how to get it

$ git init
$ git remote add <remote_url>
$ git fetch --all

now you can see all branches & commits

$ git branch -a
$ git log remotes/origin/master <-- or any other branch

Finally you know SHA1 of desired commit

git reset --hard <sha1>

---

Full workflow for cloning a single branch, choosing a commit, then checking out that specific commit... This method requires git version 2.28.0 or higher to use the option --no-write-fetch-head, tested with version 2.35.3. (If you already know the full sha1 hash of the commit you want, please skip ahead to the second method available in the final code block)

#Create empty git repo
mkdir repo && cd repo && git init

#add remote, configure it to track <branch>
git remote add --no-tags -t <branch> -m <branch> origin <url>

#fetch objects from remote repo
git fetch --no-write-fetch-head

#examine commits and logs to decide which one we will use
git log --oneline origin

#Once you have found the commit of interest copy the abbreviated hash or save as variable
commit=<sha1>

#rename our default branch to match remote branch
git branch -m <branch>

#set branch head to desired commit
git branch <branch> $commit

#set remote branch as upstream for <branch>
git branch -u origin <branch>

#All done time to checkout
git checkout

To optionally truncate the history of the local branch execute :

git fetch --no-write-fetch-head --depth <n> ./ <branch>

To truncate the remote branch history you can execute the following, but keep in mind that if you truncate history to a commit newer than the commit you checked out git status will tell you that you have diverged from the remote by <n> commits

git fetch --no-write-fetch-head --depth <n>

If you don't need remote tracking and already know the the full commit hash :

mkdir repo && cd repo && git init
git remote --no-tags add origin <url>
git fetch --depth 1 --no-write-fetch-head origin <sha1>
#Set default local branch (master in this case) head to <sha1>
git branch master <sha1>
git checkout

What makes this method better in my opinion is that it truly fetches only a single commit. We also avoid creating a FETCH_HEAD or ORIG_HEAD leaving our .git directory squeaky clean. This also leaves the reflog clean (with a single entry) as opposed to having two entries due to a git reset --hard commit Without the need for remote tracking and using fetch --depth 1 it creates the smallest possible clone (shallow clone).

---

mkdir linux-4.3.20151106
cd linux-4.3.20151106/
git init
git fetch git@github.com:torvalds/linux.git 9154301a47b33bdc273d8254c407792524367558

error: unknown option `no-write-fetch-head'        
usage: git fetch [<options>] [<repository> [<refspec>...]]   or: git fetch [<options>] <group>                                      
   or: git fetch --multiple [<options>] [(<repository> | <group>)...]   or: git fetch --all [<options>]       



 git --version
git version 2.17.1

export https_proxy=http://192.168.1.3:1080;export http_proxy=http://192.168.1.3:1080
add-apt-repository ppa:git-core/ppa
apt update
apt-get install --only-upgrade git

 git --version
git version 2.38.0


 git fetch git@github.com:torvalds/linux.git 9154301a47b33bdc273d8254c407792524367558 --no-write-fetch-head --depth=1 
remote: Enumerating objects: 54692, done.
remote: Counting objects: 100% (54692/54692), done.
remote: Compressing objects: 100% (50960/50960), done.
remote: Total 54692 (delta 3828), reused 29210 (delta 2966), pack-reused 0
Receiving objects: 100% (54692/54692), 147.35 MiB | 2.85 MiB/s, done.
Resolving deltas: 100% (3828/3828), done.
 


git branch master 9154301a47b33bdc273d8254c407792524367558


git checkout
 

---

I use this snippet with GNU make to close any revision tag, branch or hash

it was tested on git version 2.17.1

${dir}:
    mkdir -p ${@D}
    git clone --recursive --depth 1 --branch ${revison} ${url} ${@} \
 || git clone --recursive --branch ${revison} ${url} ${@} \
 || git clone ${url} ${@}
    cd ${@} && git reset --hard ${revison}
    ls $@

---

git clone https://github.com/ORGANIZATION/repository.git (clone the repository)

cd repository (navigate to the repository)

git fetch origin 2600f4f928773d79164964137d514b85400b09b2

git checkout FETCH_HEAD

---

For single files and when the commit number is known, one can use a wget onliner:

wget https://raw.githubusercontent.com/torvalds/linux/896066ee1cf4d653057dac4e952f49c96ad16fa7/README

---

git clone -o <sha1-of-the-commit> <repository-url> <local-dir-name>

git uses the word origin in stead of popularly known revision

Following is a snippet from the manual $ git help clone

--origin <name>, -o <name>
    Instead of using the remote name origin to keep track of the upstream repository, use <name>.