[PATCH v3 00/23] Delete directories left empty after ref deletion

This is a re-roll of an old patch series. v1 [1] got some feedback, which I think was all addressed in v2 [2]. But it seems that v2 fell on the floor, and I didn't bother following up because it was in the same area of code that was undergoing heavy changes due to the pluggable reference backend work. Sorry for the long delay before getting back to it.

It turns out that this patch series is still relevant and didn't even need all that must adjustment. While rebasing onto the current `master`, I tidied up a bit, tightened up some code, and improved some commit messages. But the spirit of the patch series and most of its code are unchanged.

This patch series is also available from my GitHub account [3] as branch delete-empty-refs-dirs.

Brief summary (see v1 [1] for more details):

Previously, we were pretty sloppy about leaving empty directories behind (under both $GIT_DIR/refs and $GIT_DIR/logs) when deleting references. Such directories could accumulate essentially forever. It's true that `pack-refs` deletes directories that it empties, but if a directory is *already* empty, then `pack-refs` doesn't remove it. It is also true that if an empty directory gets in the way of the creation of a *new* reference, then it is deleted. But otherwise there is no systematic cleanup of empty directories.

A reason for the old behavior was that the code paths *creating* new files in these hierarchies were not always robust against a race with another process that was cleaning up empty directories. So most of this patch series is dedicated to hardening up the creation code paths (via a new function, `raceproof_create_file()`). The last several patches teach `files_transaction_commit()` to delete empty directories when deleting references and reflogs.

Michael

[1] http://public-inbox.org/git/cover.1455626201.git.mhagger@alum.mit.edu/T/#u [2] http://public-inbox.org/git/cover.1456405698.git.mhagger@alum.mit.edu/T/#u [3] http://github.com/mhagger/git

Michael Haggerty (23):
  files_rename_ref(): tidy up whitespace
  t5505: use "for-each-ref" to test for the non-existence of references
  safe_create_leading_directories_const(): preserve errno
  safe_create_leading_directories(): set errno on SCLD_EXISTS
  raceproof_create_file(): new function
  lock_ref_sha1_basic(): inline constant
  lock_ref_sha1_basic(): use raceproof_create_file()
  rename_tmp_log(): use raceproof_create_file()
  rename_tmp_log(): improve error reporting
  log_ref_write(): inline function
  log_ref_setup(): separate code for create vs non-create
  log_ref_setup(): improve robustness against races
  log_ref_setup(): pass the open file descriptor back to the caller
  log_ref_write_1(): don't depend on logfile argument
  log_ref_setup(): manage the name of the reflog file internally
  log_ref_write_1(): inline function
  delete_ref_loose(): derive loose reference path from lock
  delete_ref_loose(): inline function
  try_remove_empty_parents(): rename parameter "name" -> "refname"
  try_remove_empty_parents(): don't trash argument contents
  try_remove_empty_parents(): don't accommodate consecutive slashes
  try_remove_empty_parents(): teach to remove parents of reflogs, too
  files_transaction_commit(): clean up empty directories

 cache.h               |  48 ++++++-
 refs/files-backend.c  | 375 +++++++++++++++++++++++++-------------------------
 refs/refs-internal.h  |  11 +-
 sha1_file.c           |  76 +++++++++-
 t/t1400-update-ref.sh |  27 ++++
 t/t5505-remote.sh     |   2 +-
 6 files changed, 346 insertions(+), 193 deletions(-)