You should not use git branch when writing scripts. Git provides a “plumbing” interface that is explicitly designed for use in scripting (many current and historical implementations of normal Git commands (add, checkout, merge, etc.) use this same interface).
The plumbing command you want is git for-each-ref:
for-each-ref --shell \
--format='git log --oneline %(refname) ^origin/master' \
refs/heads/
Note: You do not need the remotes/
prefix on the remote ref unless you have other refs that cause origin/master
to match multiple places in the ref name search path (see “A symbolic ref name. …” in the Specifying Revisions section of git-rev-parse(1)). If you are trying to explictly avoid ambiguity, then go with the full ref name: refs/remotes/origin/master
.
You will get output like this:
git log --oneline 'refs/heads/master' ^origin/master
git log --oneline 'refs/heads/other' ^origin/master
git log --oneline 'refs/heads/pu' ^origin/master
You can pipe it into sh.
If you do not like the idea of generating the shell code, you could give up a bit of robustness* and do this:
for branch in $(git for-each-ref --format='%(refname)' refs/heads/); do
git log --oneline "$branch" ^origin/master
done
*
Ref names should be safe from the shell’s word splitting (see git-check-ref-format(1)). Personally I would stick with the former version (generated shell code); I am more confident that nothing inappropriate can happen with it.
Since you specified bash and it supports arrays, you could maintain safety and still avoid generating the guts of your loop:
branches=()
eval "$(git for-each-ref --shell --format='branches+=(%(refname))' refs/heads/)"
for branch in "${branches[@]}"; do
…
done
You could do something similar with $@
if you are not using a shell that supports arrays (set --
to initialize and set -- "$@" %(refname)
to add elements).