[PATCH v3 0/7] Gets rid of "if reflog exists, append to it regardless of config settings"

<As discussed in CAFQ2z_Ps3YxycA+NJ9VKt_PEXb+m83JdNB7ujzWw1fTPKyZ=fg@mail.gmail.com

v3:

 * fix show-branch
 * add some more context to commit messages
 * change calling convention for log_ref_setup; could fold into predecessor
   if needed too.

Han-Wen Nienhuys (7):
  show-branch: show reflog message
  refs: trim newline from reflog message
  test-ref-store: tweaks to for-each-reflog-ent format
  t1400: use test-helper ref-store to inspect reflog contents
  refs: drop force_create argument of create_reflog API
  RFC: refs: reflog entries aren't written based on reflog existence.
  refs: change log_ref_setup calling convention

 builtin/checkout.c             |   2 +-
 builtin/show-branch.c          |   7 +-
 reflog-walk.c                  |   6 +-
 refs.c                         |   9 ++-
 refs.h                         |   4 +-
 refs/debug.c                   |   5 +-
 refs/files-backend.c           | 128 +++++++++++++--------------------
 refs/packed-backend.c          |   3 +-
 refs/refs-internal.h           |   2 +-
 t/helper/test-ref-store.c      |   8 +--
 t/t1400-update-ref.sh          |  21 +++---
 t/t1405-main-ref-store.sh      |   6 +-
 t/t1406-submodule-ref-store.sh |   6 +-
 t/t3202-show-branch.sh         |  15 ++++
 14 files changed, 101 insertions(+), 121 deletions(-)

base-commit: e0a2f5cbc585657e757385ad918f167f519cfb96
Published-As: https://github.com/gitgitgadget/git/releases/tag/pr-git-1067%2Fhanwen%2Freflog-touch-v3
Fetch-It-Via: git fetch https://github.com/gitgitgadget/git pr-git-1067/hanwen/reflog-touch-v3
Pull-Request: https://github.com/git/git/pull/1067

Range-diff vs v2:

 -:  ----------- > 1:  e158882812f show-branch: show reflog message
 1:  995d450da42 ! 2:  d16d94164c1 refs: trim newline from reflog message
     @@ Commit message
      
       ## builtin/show-branch.c ##
      @@ builtin/show-branch.c: int cmd_show_branch(int ac, const char **av, const char *prefix)
     - 				show_one_commit(rev[i], 1);
     + 			char *logmsg;
     + 			char *nth_desc;
     + 			const char *msg;
     +-			char *end;
     + 			timestamp_t timestamp;
     + 			int tz;
     + 
     +@@ builtin/show-branch.c: int cmd_show_branch(int ac, const char **av, const char *prefix)
     + 				break;
       			}
     - 			else
     --				puts(reflog_msg[i]);
     -+				puts(reflog_msg[i]); /* XXX - this puts a
     -+							newline. Did we put two
     -+							newlines beforehand? */
       
     - 			if (is_head)
     - 				head_at = i;
     +-			end = strchr(logmsg, '\n');
     +-			if (end)
     +-				*end = '\0';
     +-
     + 			msg = (*logmsg == '\0') ? "(none)" : logmsg;
     + 			reflog_msg[i] = xstrfmt("(%s) %s",
     + 						show_date(timestamp, tz,
      
       ## reflog-walk.c ##
      @@ reflog-walk.c: void get_reflog_message(struct strbuf *sb,
 2:  11b296a55e9 = 3:  e273963216c test-ref-store: tweaks to for-each-reflog-ent format
 3:  9ec09cc64cd = 4:  52093fce57c t1400: use test-helper ref-store to inspect reflog contents
 4:  aa25fd9b7de ! 5:  ce0047028dd refs: drop force_create argument of create_reflog API
     @@ Commit message
          There is only one caller, builtin/checkout.c, and it hardcodes
          force_create=1.
      
     +    This argument was introduced in abd0cd3a301 (refs: new public ref function:
     +    safe_create_reflog, 2015-07-21), which promised to immediately use it in a
     +    follow-on commit, but that never happened.
     +
          Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
      
       ## builtin/checkout.c ##
     @@ refs/files-backend.c: error:
       	int fd;
       
      -	if (log_ref_setup(refs, refname, force_create, &fd, err))
     -+	if (log_ref_setup(refs, refname, /*force_create=*/1, &fd, err))
     ++	if (log_ref_setup(refs, refname, 1, &fd, err))
       		return -1;
       
       	if (fd >= 0)
 5:  f6a7c5ad56e ! 6:  7a030cfd3e2 RFC: refs: reflog entries aren't written based on reflog existence.
     @@ Commit message
          The reftable storage backend cannot distinguish between a non-existing
          reflog, and an empty one. Therefore it cannot mimick this functionality.
      
     -    In CAFQ2z_Ps3YxycA+NJ9VKt_PEXb+m83JdNB7ujzWw1fTPKyZ=fg@mail.gmail.com,
     -    we came to the conclusion that this feature is probably a remnant from
     -    the time that reflogs weren't enabled by default, and it does not need
     -    to be kept.
     +    With this feature, it is possible to mark only specific branches as subject to
     +    reflog updates. When introduced, it presumably served as a cheap substitute for
     +    introducing branch.$NAME.logRefUpdate configuration setting.
     +
     +    Reflogs are small and don't impact the runtime of normal operations, so this
     +    flexibility is not very useful. Since it incurs complexity for alternate ref
     +    backends, we remove it.
     +
     +    Further background to this change is in
     +    <CAFQ2z_Ps3YxycA+NJ9VKt_PEXb+m83JdNB7ujzWw1fTPKyZ=fg@mail.gmail.com>.
      
          Signed-off-by: Han-Wen Nienhuys <hanwen@google.com>
      
 -:  ----------- > 7:  1124dbad594 refs: change log_ref_setup calling convention