I am well aware that when writing a pipeline script for a Jenkins job, good practice is to reclone a repo every time you run the job, however I do find myself desiring to just pull the branch content or checkout a revision and just run the job on that (in some cases at least).
However I want to let the user specify either a hash or a branch name to check out.
If it is a branch name, then a git pull is in order, since the user probably pushed new content to try out with the job.
But if the specified revision is a hash then we will be in detached HEAD mode, and git pull will fail with
You are not currently on a branch.
Please specify which branch you want to merge with.
See git-pull(1) for details.
git pull <remote> <branch>
How to silence this error or only run git pull if it was a branch that was specified ? Or circumvent this problem altogether ?
I have found some code to check if the passed revision is a branch but I find it a bit fidgety :
git show-ref --verify --quiet refs/heads/${params.revision}
if %ERRORLEVEL% EQU 0 git pull
Any more natural suggestions ?
1
Note: throughout this answer I will use <commit-ID-or-branch-name> to indicate what is being passed in. Since we don’t know if it’s a commit ID or branch name, all commands below need to work for both cases. This answer also assumes the shell is Bash. (Adjust the syntax accordingly if you’re using a different shell.)
In the case where a branch name is provided, let’s first make sure you don’t have a local copy of that branch. We can use Rahul Sharma’s comment’s syntax for returning true even if the command fails, which it would do when a commit ID is used instead of a branch name:
git switch --detach
git branch -D <commit-ID-or-branch-name> | true
If a branch was passed in, the above commands simply delete that local branch. If the local branch didn’t already exist, or if a commit ID was passed in, then nothing happens.
The following set of 2 commands will now work for both a commit ID and a branch name:
git fetch origin <commit-ID-or-branch-name>
git checkout <commit-ID-or-branch-name>
Note that the checkout command can be used on both a commit ID and a branch name without specifying the --detach option. If your pipeline doesn’t need to create new commits on a branch, then it may be (slightly) better to always use a commit ID even if you are checking out a branch, in which case you should use the --detach option with checkout, or, you can also use the switch command:
git fetch origin <commit-ID-or-branch-name>
git switch --detach <commit-ID-or-branch-name>
Side Notes:
- Checking out a specific commit is slightly different than
pull. Since the premise of your question is comparing pulling with a fresh clone, then this would be functionally equivalent, and in fact conceptually “cleaner” thanpull, since it omits the possibility of unknown local commits tainting the build. If they did exist, they will also be removed above with the deletion of that branch. - The above commands assume a clean status. If you find the pipeline failing because the status does not start out clean, consider using a clean command first, or the force option with checkout.
2