If I run a 'git clone' from the Windows 10 command line to a Linux remote repo in our office, there seem to be no permission issues; I'm presented with an ssh login prompt to the remote server, my password is accepted, and the clone runs to completion without a problem.
If I run the same 'git clone' from Cygwin, I don't even get the login prompt and the git command fails with:
Permission denied (publickey,password)
The thing is, I think I know why the problem is happening, but not what to do about it.
Image my name is John Smith. My credentials as stored by the remote Linux server are 'john.smith'. My Windows user folder is 'C:\Users\john.smith'. However, when I start a Cygwin session, the prompt is actually 'John.Smith#myhostname'. So because Windows is case-insensitive, the git clone works because the Windows copy of 'known-hosts' can be located (no case issues). However, the copy of 'known-hosts' in Cygwin isn't being located because of the mismatch between the Unix username (john.smith) and the Cygwin username (John.Smith). That's my theory anyway !
What do I need to do in Cygwin to lower-case my username so that everything aligns ?
Thanks
Can you try to issue the clone command with the format: john.smith#linuxhost:path/to/repo?
Check where your git configuration is
git config --list --show-origin
cygwin probably uses /home/john.smith as home/profile folder(directory).
But here is main issue ssh configuration.
You could copy your ssh private and host keys from windows location to cygwin home .ssh directory.
Check in your Cygwin install folder under <install-path>/Cygwin/home/<username>
You should find:
.gitconfig - edit this to change your username and email address etc...
.ssh - this contains your known-hosts and id_rsa ssh keys. You probably need to copy your windows ones into here.
You may also find other issues in Cygwin like auto complete not working? and a few other gotchas...?
Related
So recently I've been trying to set up a Git server to manage personal projects and work. I created a bare Git repo on a spare laptop running Ubuntu and I installed Git Bash on my Windows machine. Right now I'm having some issues in actually being able to clone my Ubuntu git server to my Windows computer.
I used git remote add origin myserver#myserver-Dell-System:/home/myserver/CentralRepo.git but
whenever I try to clone to my Windows machine, Git bash spits out this error:
ssh: Could not resolve hostname myserver#myserver-Dell-System: Name or service not known
fatal: Could not read from remote repository.
Please make sure you have the correct access rights and repository exists.
I double checked my add to make sure the address of my Git server was correct and I assume Git bash for Windows comes with OpenSSH.
Does anyone know why this is happenning and how I can get my Windows machine to sucessfully clone from a server to itself? Thanks!
A valid ssh url would be
auser#myserver-Dell-System
You can check if you can resolve and connect to that server with
ssh -T auser#myserver-Dell-System
That supposes that you have generated a public/private ssh key pair, and added your public key to ~auser/.ssh/authorized_keys file first.
See "Adding SSH Keys to authorized_keys"
when I make git clone with ssh from a user prompt it works properly.
git clone ssh://URL.com/soft.git soft_git
the ssh key id_rsa and id_rsa.pub are under /home/user/.ssh
my purpose is the execute git with sudo but I got the following error
Cloning into '/home/user/git/soft'...
Permission denied (publickey,keyboard-interactive).
fatal: Could not read from remote repository.
Please make sure you have the correct access rights
and the repository exists.
I create a folder /root/.ssh and I copy the ssh keys into it but I got the same error
how to execute git with sudo properly.
When you run git using sudo, git will run as root. Because git is running as root, ssh is running as root. Because ssh is running as root, it is trying to log on to the remote server as root. The remote server is not okay with this (as it should be!)
You will need to do two things:
Put the username in your URL: ssh://myusername#URL.com/soft.git.
Make your SSH key available to the root user, because it will look under /root/.ssh instead of /home/user/.ssh. (You could also probably point SSH at the correct key, but I don't know how to do this, and SSH is picky about permissions.)
On my computer (Ubunutu 18.04), adding SSH_AUTH_SOCK=$SSH_AUTH_SOCK after sudo and before git fixed the problem:
sudo SSH_AUTH_SOCK=$SSH_AUTH_SOCK git clone git#github.com:my-github-account/my-repo.git
Normally, sudo's SSH_AUTH_SOCK environment variable won't be set properly. Executing the git clone with SSH_AUTH_SOCK=$SSH_AUTH_SOCK sets sudo's SSH_AUTH_SOCK environment variable to whatever it is for you.
This way, you don't need to add an extra .ssh dir for sudo with copies of your keys, which is what I think one of the other answers suggests.
The solution is more fully explained in this rather old github gist:
https://gist.github.com/scottjacobsen/4281310
P.S. I'm adding a new answer several years later; I googled a solution to this problem, and this SO Q/A is one of the first things that comes up.
Normally the default remote ssh user is the same as your user name. If you're using sudo this will be root which probably isnt' going to work. You need to supply the remote username.
sudo git clone ssh://username#URL.com/soft.git soft_git
You can generally resolve git ssh issues easier by trying to login to the remote with plain ssh. You'll get better diagnostics and can see what's going wrong.
sudo ssh ssh://URL.com/
"Failed to find 'git' on your PATH. Please ensure 'git' is executable by the Go Server and on the Go Agents where this material will be used."
I'm getting this when attempting to use GIT when creating a new build pipeline.
I'm not that experienced with Linux to know how to debug this and I can't find anything in the docs or via a google search.
GO was installed as root, so was GIT, the server has been rebooted since and I can run git in bash. "which git" shows that it is in /usr/bin. "export $path" shows that /usr/bin is in the PATH variable.
Can anyone recommend anything else to try?
Thanks.
The Go server and agent runs under user 'go' and therefore you will need to ensure git is in path for this user.
Change the user to go. If you get a prompt for password, put the password of the current logged in user.
$ sudo su - go
Then check if git is in path or simply check for the version
$ which git
$ git --version
If you don't get expected output for the above commands you might have to provide git executables to the user 'go', that is, you will need to add git in this user's PATH. Alternatively, you can provide the path in /etc/default/go-server
export PATH=$PATH:/path/to/git
I recall having this problem too, for me it was related to a local git repository that I did not share with the git daemon. Running the repository with the git daemon solved this problem for me.
URL for me starts with git://localhost, not sure if http will work too
I am trying to make a new branch in Gitlab by using Gitolite. I complete the installation steps. when i come across "setting up gitolite" section i have a trouble. I followed this link.
When i run
gitolite setup -pk alice.pub
command i got "bash: gitolite: command not found" error message. I don't know what is the problem.. Any one please help me.
This step comes after the Gitolite installation, which supposes you have chosen one of three possibilities:
Keep the sources anywhere and use the full path to run the gitolite command.
Keep the sources anywhere and symlink just the gitolite program to some directory on your $PATH.
Copy the sources somewhere and use that path to run the gitolite command.
So make sure gitolite is in your PATH, and that command will work.
I prefer a local installation of gitolite (in a local directory, as opposed to /usr/local, which requires root privileges.).
See, for illustration, "install_or_update_gitolite.sh"
"${github}/install" -to "${gtl}/bin" # Note: "${gtl}/bin" is in my $PATH
GITOLITE_HTTP_HOME= gitolite setup -pk "${H}/.ssh/gitoliteadm.pub"
Note that for gitolite setup to properly work, you might want to set GITOLITE_HTTP_HOME to an empty string first.
As I also faced the same problem, I found the solution(s) as below.
First way is ...,
Open your terminal and key in below code
$ PATH=$PATH:~/bin
It is because the value of $PATH variable is point to incorrected path.
So I just modify this variable.
To be more detail click here.
Second way is ...,
Edit .bashrc file going to the end and insert below line.
PATH=/home/git/bin:$PATH
To be more detail click here.
On debian, there is no /usr/bin/gitolite
Linux debian-srv 3.2.0-4-amd64 #1 SMP Debian 3.2.60-1+deb7u3 x86_64 GNU/Linux ls: cannot access /home/gitolite/bin: No such file or directory
ls: cannot access /usr/bin/gito*: No such file or directory
Here installing gitolite3 helped:
apt-get install gitolite3
root#debian-srv:# which gitolite
/usr/bin/gitolite
I have a problem with the PATH variable on Debian Linux when executing commands via SSH. This happens for example when I use GIT or Mercurial and I have them installed in /opt instead of /usr/local. But I can also reproduce this easily like this:
When I login to the server via SSH in a normal way and then do echo $PATH then I get the PATH which I have configured in /etc/profile:
/usr/local/bin:/usr/bin:/bin:/usr/games:/opt/maven/bin:/opt/ant/bin:/opt/mercurial/bin:/opt/git/bin
But when I do ssh user#server 'echo $PATH' instead then I get this:
/usr/local/bin:/usr/bin:/bin:/usr/bin/X11:/usr/games
The same happens when I use GIT or Mercurial with an ssh URL. It can't find hg or git executable on the server because it don't get the PATH from /etc/profile.
So the question is: Is there a "more global" way to configure the PATH so it also works with SSH remote execution? Configuring the PATH per user is not an option. Specifying the full path to git/hg executables when using git/hg on the client is also not an option. And I know that I could use symlinks or wrapper scripts in /usr/local/bin to get it working but I'm searching explicitly for a possibility to have a correct PATH when executing commands remotely via SSH.
On Debian, the standard environment is setup through pam_env (in /etc/pam.d/sshd), which will read /etc/environment and /etc/security/pam_env.conf. You can either edit those, or you add another pam_env line to the pam configuration, pointing to an environment file specific to SSH logins.
While not an answer to your problem, from this link, it looks like ssh has a path settings of its own.
The ssh server set some environment variables at the start of the session. You can create a ~/.ssh/environment file on the ssh server to set additional variables (assuming there is a BSD openssh server on the server), but the server must be configured to allow to change the environment.
If the above doesn't work, you can set the remote command. In mercurial this is the --remotecmd switch for push and pull.
I solved this issue by including all necessary files to the .bashrc
For instance:
. /etc/profile.d/rvm.sh
. /etc/profile.d/bash_completion.sh