[PATCH] branch: description for non-existent branch errors

["Rubén Justo" <rjusto@gmail.com>]

When the repository does not yet has commits, some errors describe that
there is no branch:

    $ git init -b first

    $ git --edit-description first
    fatal: branch 'first' does not exist

    $ git --set-upstream-to=upstream
    fatal: branch 'first' does not exist

    $ git branch -c second
    error: refname refs/heads/first not found
    fatal: Branch copy failed

That "first" branch is unborn but to say it doesn't exists is confusing.

Options "-c" (copy) and "-m" (rename) show the same error when the
origin branch doesn't exists:

    $ git branch -c non-existent-branch second
    error: refname refs/heads/non-existent-branch not found
    fatal: Branch copy failed

    $ git branch -m non-existent-branch second
    error: refname refs/heads/non-existent-branch not found
    fatal: Branch rename failed

Note that "--edit-description" without an explicit argument is already
considering the _empty repository_ circumstance in its error.  Also note
that "-m" on the initial branch it is an allowed operation.

This commit makes the error descriptions for those branch operations
with unborn or non-existent branches, more informative.

This is the result of the change:

    $ git init -b first

    $ git --edit-description first
    fatal: No commit on branch 'first' yet.

    $ git --set-upstream-to=upstream
    fatal: No commit on branch 'first' yet.

    $ git -c second
    fatal: No commit on branch 'first' yet.

    $ git [-c/-m] non-existent-branch second
    fatal: No branch named 'non-existent-branch'.

---

Hi all.

I would appreciate some guidance here.

This patch is based on master but there is already in 'seen' a patch
submitted for 'branch' that produces conflicts with this.  I don't know
if it is more convenient to submit this patch based on 'seen'
considering that.

I submitted that other patch too and prefer to keep the changes separate
but maybe it's better not to and do a rollup with this on that other
patch.. I don't know how disturbing it is to submit a new patch when it
is "waiting for review".

Another approach?

Thank you.


Un saludo.

 builtin/branch.c | 14 ++++++++++++--
 1 file changed, 12 insertions(+), 2 deletions(-)

diff --git a/builtin/branch.c b/builtin/branch.c
index 55cd9a6e99..5ca35064f3 100644
--- a/builtin/branch.c
+++ b/builtin/branch.c
@@ -538,6 +538,13 @@ static void copy_or_rename_branch(const char *oldname, const char *newname, int
 			die(_("Invalid branch name: '%s'"), oldname);
 	}
 
+	if (copy && !ref_exists(oldref.buf)) {
+		if (!strcmp(head, oldname))
+			die(_("No commit on branch '%s' yet."), oldname);
+		else
+			die(_("No branch named '%s'."), oldname);
+	}
+
 	/*
 	 * A command like "git branch -M currentbranch currentbranch" cannot
 	 * cause the worktree to become inconsistent with HEAD, so allow it.
@@ -805,7 +812,7 @@ int cmd_branch(int argc, const char **argv, const char *prefix)
 		if (!ref_exists(branch_ref.buf)) {
 			strbuf_release(&branch_ref);
 
-			if (!argc)
+			if (!argc || !strcmp(head, branch_name))
 				return error(_("No commit on branch '%s' yet."),
 					     branch_name);
 			else
@@ -848,8 +855,11 @@ int cmd_branch(int argc, const char **argv, const char *prefix)
 			die(_("no such branch '%s'"), argv[0]);
 		}
 
-		if (!ref_exists(branch->refname))
+		if (!ref_exists(branch->refname)) {
+			if (!argc || !strcmp(head, branch->name))
+				die(_("No commit on branch '%s' yet."), branch->name);
 			die(_("branch '%s' does not exist"), branch->name);
+		}
 
 		dwim_and_setup_tracking(the_repository, branch->name,
 					new_upstream, BRANCH_TRACK_OVERRIDE,
-- 
2.36.1