The following was written by Hongzhe Zhou in August 2022 in response to
a problematic check-in using git. In order for others to avoid this,
please read this and make sure you avoid mistakes.
Axel

So in general I have two suggestions:

One is to use
git pull --rebase, (the rebase option is important)
instead of git merge
as much as possible, especially when one has relatively small changes to the code.
This keeps the history of the code linear and makes it easier to reverted changes.
Personally I only use merge when my private branch is many commits behind the master, and there
is potentially a lot of conflicts I need to solve. This rarely happens when I try to check anything in,
but only when I want to make my local private branch updated (not updating the master branch).

The second is to make use of
pc_auto-test
git diff
gitk
git pc panic
before you push anything to the remote repo. The first command runs some test examples (optionally
at different levels) and compares with the reference data, to make sure the core part of the code
is not broken. The “git diff” will show the difference between the head and the last commit of the branch,
so that one doesn’t check in unnecessary things. The rest two are essentially the same, which displays “git log” in a graphical way so one understands where the head is and relations between branches etc.

Also after checking in anything it would be good to check the auto tests at
http://pencil-code.nordita.org/tests.php
The first one (Travis) will run each time a new commit is in.

Enjoy coding!

Hongzhe