When committing a code change with git, it will not complain about the commit if any line(s) have trailing whitespace. However, when you do various git rebase type actions, git will complain about that commit containing trailing whitespaces.

The safest way to handle trailing whitespaces with git is to detect it before the commit step and abort with output displaying why it was skipped. By default, a git repo will have a .git/hooks/pre-commit.sample script. Simply rename it to .git/hooks/pre-commit and make sure it is executable. Then any time you do a git commit it will look at the files staged for commit, report if there are any lines with trailing whitespaces, and will skip the commit if there are.

In the following example, I added two lines with trailing whitespaces to a README file:

[tlyons@ivwww01 ~/projects/TEST (master)]$ vim README
[tlyons@ivwww01 ~/projects/TEST (master)]$ git add README
[tlyons@ivwww01 ~/projects/TEST (master)]$ git commit
README:3: trailing whitespace.
+test    
README:6: trailing whitespace.
+test....   
[tlyons@ivwww01 ~/projects/TEST (master)]$

Commit hook problems in git are difficult to diagnose

Git expert shruggar in the #git irc channel on Freenode cautions that

I would severely recommend against making "commit" do any magic of this sort. It invalidates too many assumptions, and will make it pretty much impossible for anyone in this channel to ever help you sanely again.

However, if you still want to do this, you can easily perform this automatic removal of trailing whitespace when making commits in your local repository. Add this script as .git/hooks/pre-commit and make it executable:

if git rev-parse --verify HEAD >/dev/null 2>&1 ; then
   against=HEAD
else
   # Initial commit: diff against an empty tree object
   against=4b825dc642cb6eb9a060e54bf8d69288fbee4904
fi
# Find files with trailing whitespace
# Doubled commands are _linux_version_ || _mac_osx_version_
for FILE in $(
    git diff-index --check --cached $against -- | \
    sed '/^[+-]/d' | \
    ( sed -r 's/:[0-9]+:.*//' 2>/dev/null || \
      sed -E 's/:[0-9]+:.*//' ) | \
    uniq
); do
  # Fix them!
  ( sed -i 's/[[:space:]]*$//' "$FILE" > /dev/null 2>&1 || \
    sed -i '' -E 's/[[:space:]]*$//' "$FILE" )
  echo "Removed trailing whitespace from $FILE"
  git add "$FILE"
done

exit

When you perform a git commit, this script will be called automatically. It finds all files which are staged to be committed and runs a check on them, looking for trailing whitespaces. The script modifies the working copy of any file(s) it finds, does a git add to stage the new change into the cache, and then commits it like normal. All of this is invisible to you the user.

Note that this has a serious flaw: If you made lots of changes to a file but only staged a few of them for this commit, the hook script above will silently add all changes to the file after fixing whitespace. This is a Bad Idea[tm].

If you need to disable the whitespace removal once, you add -n to the git commit commandline, which bypasses the pre-commit and commit-msg hooks.

If you want to completely disable the whitespace removal, you can delete, rename or remove execute permissions from the .git/hooks/pre-commit script.