简体   繁体   中英

Run shell script from python with permissions

I have the most simple script called update.sh

#!/bin/sh
cd /home/pi/circulation_of_circuits
git pull

When I call this from the terminal with ./update.sh I get a Already up-to-date or it updates the files like expected.

I also have a python script, inside that scipt is:

subprocess.call(['./update.sh'])

When that calls the same script I get:

Permission denied (publickey). fatal: Could not read from remote repository.

Please make sure you have the correct access rights and the repository exists.

(I use SSH).

----------------- update --------------------

Someone else had a look for me:

OK so some progress. When I boot your image I can't run git pull in your repo directory and the bash script also fails. It seems to be because the bitbucket repository is private and needs authentication for pull (the one I was using was public so that's why I had no issues). Presumably git remembers this after you type it in the first time, bash somehow tricks git into thinking it's you typing the command subsequently but running it from python isn't the same.

I'm not a git expert but there must be some way of setting this up so python can provide the authentication.

sounds like you need to give your ssh command a public or private key it can access perhaps:

ssh -i /backup/home/user/.ssh/id_dsa user@unixserver1.nixcraft.com

-i tells it where to look for the key

This problem is caused by the git repo authentication failing. You say you are using SSH, and git is complaining about publickey auth failing. Normally you can use git commands on a private repo without inputting a password. All this would imply that git is using ssh, but in the latter case it cannot find the correct private key.

Since the problem only manifests itself when run through another script, it is very likely caused by something messing with the environment variables. Subprocess.call should pass the environment as is, so there are a couple of usual suspects:

  1. sudo.
    • if you are using sudo, it will pass a mostly empty environment to the process
  2. the python script itself
    • if the python script changes its env, those changes will get propagated to the subprocess too.
  3. sh -l or su -
    • these commands set up a login shell, which means their environment gets reset to defaults.

Any of these reasons could hide the environment variables ssh-agent (or some other key management tool) might need to work.

Steps to diagnose and fix:

  1. Isolate the problem.

    • Create a minimal python script that does nothing else than runs subprocess.call(['./update.sh']) . Run both update.sh and the new script.
  2. Diagnose the problem and fix accordingly:

    a) If update.sh works, and the new script doesn't, you are probably experiencing some weird corner case of system misconfiguration. Try upgrading your system and python; if the problem persists, it probably requires additional debugging on the affected system itself.

    b) If both update.sh and the new script work, then the problem lies within the outer python script calling the shell script. Look for occurrences of sudo , su - , sh -l , env and os.environ , one of those is the most likely culprit.

    c) If neither the update.sh nor the new script work, your problem is likely to be with ssh client configuration; a typical cause would be that you are using a non-default identity, did not configure it in ~/.ssh/config but used ssh-add instead, and after that, ssh-agent's cache expired. In this case, run ssh-add identityfile for the identity you used to authenticate to that git repo, and try again.

I believe this answer will help you: https://serverfault.com/questions/497217/automate-git-pull-stuck-with-keychain?answertab=votes#tab-top

I didn't use ssh-agent and it worked: Change your script to the one that follows and try.

#!/bin/bash
cd /home/pi/circulation_of_circuits 
ssh-add /home/yourHomefolderName/.ssh/id_rsa
ssh-add -l
git pull

This assumes that you have configured correctly your ssh key.

It seems like your version control system, need the authentication for the pull so can build the python with use of pexpect ,

import pexpect
child = pexpect.spawn('./update.sh')
child.expect('Password:')
child.sendline('SuperSecretPassword')

Try using the sh package instead of using the subprocess call. https://pypi.python.org/pypi/sh I tried this snippet and it worked for me.

#!/usr/local/bin/python

import sh

sh.cd("/Users/siyer/workspace/scripts")
print sh.git("pull")

Output:

Already up-to-date.

import subprocess 

subprocess.call("sh update.sh", shell=True)

I can reproduce your fault . It has nothing to do with permission , it depends how your ssh are installed on your system. To verify it's the same cause i need the diff output .

Save the following to a file log_shell_env.sh ,

#!/bin/bash

log="shell_env"$1
echo "create shell_env"$1

echo "shell_env" > $log

echo "whoami="$(whoami) >> $log
echo "which git="$(which git) >> $log
echo "git status="$(git status 2>&1) >> $log
echo "git pull="$(git pull 2>&1) >> $log
echo "ssh -vT git@github.com="$(ssh -T git@github.com 2>&1) >> $log

echo "ssh -V="$(ssh -V 2>&1) >> $log
echo "ls -al ~/.ssh="$(ls -a ~/.ssh) >> $log

echo "which ssh-askpass="$(which ssh-askpass) >> $log
echo "ps -e | grep [s]sh-agent="$(ps -e | grep [s]sh-agent ) >> $log
echo "ssh-add -l="$(ssh-add -l) >> $log

echo "set=" >> $log
set  >> $log

set execute permission and run it twice :
1. From the console without parameter
2. From your python script with parameter '.python'
Please, run it realy from the same python script!

   For instance:
    try:
        output= subprocess.check_output(['./log_shell_env.sh', '.python'], stderr=subprocess.STDOUT)
        print(output.decode('utf-8'))

    except subprocess.CalledProcessError as cpe:
        print('[ERROR] check_output: %s' % cpe)

Do a diff shell_env shell_env.python > shell_env.diff The resulting shell_env.diff should show not more than the following diffs:

15,16c15,16  
< BASH_ARGC=()
< BASH_ARGV=()
---
> BASH_ARGC=([0]="1")
> BASH_ARGV=([0]=".python")
48c48
< PPID=2209
---
> PPID=2220
72c72
< log=shell_env
---
> log=shell_env.python

Come back and comment, if you get more diffs update your Question with the diff output.

With Git 1.7.9 or later, you can just use one of the following credential helpers:

With a timeout

git config --global credential.helper cache

... which tells Git to keep your password cached in memory for (by default) 15 minutes. You can set a longer timeout with:

git config --global credential.helper "cache --timeout=3600"

(That example was suggested in the GitHub help page for Linux.) You can also store your credentials permanently if so desired.

Saving indefinitely

You can use the git-credential-store via

git config credential.helper store

GitHub's help also suggests that if you're on Mac OS X and used Homebrew to install Git, you can use the native Mac OS X keystore with:

git config --global credential.helper osxkeychain

For Windows, there is a helper called Git Credential Manager for Windows or wincred in msysgit.

git config --global credential.helper wincred # obsolete

With Git for Windows 2.7.3+ (March 2016):

git config --global credential.helper manager

For Linux, you can use gnome-keyring(or other keyring implementation such as KWallet).

Finally, after executing one of the suggested command one time manually, you can execute your script without changes in it.

Use the following python code. This will import the os module in python and make a system call with sudo permissions.

#!/bin/python
import os 
os.system("sudo ./update.sh")

The technical post webpages of this site follow the CC BY-SA 4.0 protocol. If you need to reprint, please indicate the site URL or the original address.Any question please contact:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM