Committing involves the following steps:

1. Determine the current value of HEAD (if any).
2. Create the new commit object.
3. Update HEAD.

Please note that step 2 can take arbitrarily long, because it might
involve the user editing a commit message.

If a second process sneaks in a commit during step 2, then the first
commit process should fail. This is usually done correctly, because
step 3 verifies that HEAD still points at the same commit that it
pointed to during step 1.

However, if there is a race when creating an *orphan* commit, then the
test in step 3 is skipped.

Add tests for proper handling of such races. One of the new tests
fails. It will be fixed in a moment.
Signed-off-by: NMichael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Instead, verify the reference's old value if and only if old_sha1 is
non-NULL.
Signed-off-by: NMichael Haggerty <mhagger@alum.mit.edu>
Reviewed-by: NStefan Beller <sbeller@google.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Instead, verify the reference's old value if and only if old_sha1 is
non-NULL.

ref_transaction_delete() will get the same treatment in a moment.
Signed-off-by: NMichael Haggerty <mhagger@alum.mit.edu>
Reviewed-by: NStefan Beller <sbeller@google.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Instead of having a separate have_old field, record this boolean value
as a bit in the "flags" field.
Signed-off-by: NMichael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Change the following functions' "flags" arguments from "int" to
"unsigned int":

 * ref_transaction_update()
 * ref_transaction_create()
 * ref_transaction_delete()
 * update_ref()
 * delete_ref()
 * lock_ref_sha1_basic()

Also change the "flags" member in "struct ref_update" to unsigned.
Suggested-by: NJunio C Hamano <gitster@pobox.com>
Signed-off-by: NMichael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

There is no reason to "reserve" a gap between the public and private
flags values.
Signed-off-by: NMichael Haggerty <mhagger@alum.mit.edu>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

It is only used internally now. Document it a little bit better, too.
Signed-off-by: NMichael Haggerty <mhagger@alum.mit.edu>
Reviewed-by: NStefan Beller <sbeller@google.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

Signed-off-by: NJunio C Hamano <gitster@pobox.com>

* ah/usage-strings:
  standardize usage info string format

* jc/pretty-format-doc:
  "log --pretty" documentation: do not forget "tformat:"

Mark file-local symbols as "static", and drop functions that nobody
uses.

* jc/unused-symbols:
  shallow.c: make check_shallow_file_for_update() static
  remote.c: make clear_cas_option() static
  urlmatch.c: make match_urls() static
  revision.c: make save_parents() and free_saved_parents() static
  line-log.c: make line_log_data_init() static
  pack-bitmap.c: make pack_bitmap_filename() static
  prompt.c: remove git_getpass() nobody uses
  http.c: make finish_active_slot() and handle_curl_result() static

Extending the js/push-to-deploy topic, the behaviour of "git push"
when updating the working tree and the index with an update to the
branch that is checked out can be tweaked by push-to-checkout hook.

* jc/push-to-checkout:
  receive-pack: support push-to-checkout hook
  receive-pack: refactor updateInstead codepath

"git push" has been taught a "--atomic" option that makes push to
update more than one ref an "all-or-none" affair.

* sb/atomic-push:
  Document receive.advertiseatomic
  t5543-atomic-push.sh: add basic tests for atomic pushes
  push.c: add an --atomic argument
  send-pack.c: add --atomic command line argument
  send-pack: rename ref_update_to_be_sent to check_to_send_update
  receive-pack.c: negotiate atomic push support
  receive-pack.c: add execute_commands_atomic function
  receive-pack.c: move transaction handling in a central place
  receive-pack.c: move iterating over all commands outside execute_commands
  receive-pack.c: die instead of error in case of possible future bug
  receive-pack.c: shorten the execute_commands loop over all commands

Restructure "reflog expire" to fit the reflogs better with the
recently updated ref API.

Looked reasonable (except that some shortlog entries stood out like
a sore thumb).

* mh/reflog-expire: (24 commits)
  refs.c: let fprintf handle the formatting
  refs.c: don't expose the internal struct ref_lock in the header file
  lock_any_ref_for_update(): inline function
  refs.c: remove unlock_ref/close_ref/commit_ref from the refs api
  reflog_expire(): new function in the reference API
  expire_reflog(): treat the policy callback data as opaque
  Move newlog and last_kept_sha1 to "struct expire_reflog_cb"
  expire_reflog(): move rewrite to flags argument
  expire_reflog(): move verbose to flags argument
  expire_reflog(): pass flags through to expire_reflog_ent()
  struct expire_reflog_cb: a new callback data type
  Rename expire_reflog_cb to expire_reflog_policy_cb
  expire_reflog(): move updateref to flags argument
  expire_reflog(): move dry_run to flags argument
  expire_reflog(): add a "flags" argument
  expire_reflog(): extract two policy-related functions
  Extract function should_expire_reflog_ent()
  expire_reflog(): use a lock_file for rewriting the reflog file
  expire_reflog(): return early if the reference has no reflog
  expire_reflog(): rename "ref" parameter to "refname"
  ...

"git log --invert-grep --grep=WIP" will show only commits that do
not have the string "WIP" in their messages.

* cj/log-invert-grep:
  log: teach --invert-grep option

* km/gettext-n:
  gettext.h: add parentheses around N_ expansion if supported

After attempting and failing a password-less authentication
(e.g. kerberos), libcURL refuses to fall back to password based
Basic authentication without a bit of help/encouragement.

* bc/http-fallback-to-password-after-krb-fails:
  remote-curl: fall back to Basic auth if Negotiate fails

Setting diff.submodule to 'log' made "git format-patch" produce
broken patches.

* dk/format-patch-ignore-diff-submodule:
  format-patch: ignore diff.submodule setting
  t4255: test am submodule with diff.submodule

"git rerere" (invoked internally from many mergy operations) did
not correctly signal errors when told to update the working tree
files and failed to do so for whatever reason.

* jn/rerere-fail-on-auto-update-failure:
  rerere: error out on autoupdate failure

"git blame HEAD -- missing" failed to correctly say "HEAD" when it
tried to say "No such path 'missing' in HEAD".

* jk/blame-commit-label:
  blame.c: fix garbled error message
  use xstrdup_or_null to replace ternary conditionals
  builtin/commit.c: use xstrdup_or_null instead of envdup
  builtin/apply.c: use xstrdup_or_null instead of null_strdup
  git-compat-util: add xstrdup_or_null helper

* ld/p4-submit-hint:
  git-p4: correct --prepare-p4-only instructions

Like the "clone" subcommand, allow excluding subdirectories in the
"sync" subcommand.

* ld/p4-exclude-in-sync:
  git-p4: support excluding paths on sync

The clone subcommand has long had support for excluding
subdirectories, but sync has not. This is a nuisance,
since as soon as you do a sync, any changed files that
were initially excluded start showing up.

Move the "exclude" command-line option into the parent
class; the actual behavior was already present there so
it simply had to be exposed.
Signed-off-by: NLuke Diamand <luke@diamand.org>
Reviewed-by: NPete Wyckoff <pw@padd.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

* jc/coding-guidelines:
  CodingGuidelines: clarify C #include rules

* ak/typofixes:
  t/lib-terminal.sh: fix typo
  pack-bitmap: fix typo

Signed-off-by: NJunio C Hamano <gitster@pobox.com>

* git://github.com/git-l10n/git-po:
  l10n: ca.po: Fix trailing whitespace

Signed-off-by: NAlex Henrie <alexhenrie24@gmail.com>

Signed-off-by: NJunio C Hamano <gitster@pobox.com>

* git://github.com/git-l10n/git-po:
  l10n: de.po: correct singular form
  l10n: de.po: translate "leave behind" correctly
  l10n: de.po: fix typo
  l10n: ca.po: update translation

* 'master' of git://github.com/alexhenrie/git-po:
  l10n: ca.po: update translation

Signed-off-by: NMichael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: NRalf Thielow <ralf.thielow@gmail.com>

This message is about leaving orphaned commits behind, not about
behind an upstream branch. Try to make this clear.
Signed-off-by: NMichael J Gruber <git@drmicha.warpmail.net>
Signed-off-by: NRalf Thielow <ralf.thielow@gmail.com>

Signed-off-by: NBenedikt Heine <bebe@bebehei.de>
Signed-off-by: NRalf Thielow <ralf.thielow@gmail.com>

Signed-off-by: NAlex Henrie <alexhenrie24@gmail.com>

If you use git-p4 with the "--prepare-p4-only" option, then
it prints the p4 command line to use. However, the command
line was incorrect: the changelist specification must be
supplied on standard input, not as an argument to p4.
Signed-off-by: NLuke Diamand <luke@diamand.org>
Acked-by: NPete Wyckoff <pw@padd.com>
Signed-off-by: NJunio C Hamano <gitster@pobox.com>

* js/t1050:
  t1050-large: generate large files without dd

* ak/cat-file-clean-up:
  cat-file: use "type" and "size" from outer scope

* git://github.com/git-l10n/git-po:
  l10n: correct indentation of show-branch usage
  l10n: de.po: translate 3 messages
  l10n: zh_CN: various fixes on command arguments
  l10n: vi.po(2298t): Updated 3 new strings
  l10n: sv.po: Update Swedish translation (2298t0f0u)
  l10n: fr.po v2.3.0 round 2
  l10n: git.pot: v2.3.0 round 2 (3 updated)
  l10n: de.po: translate 13 new messages
  l10n: de.po: fix typo
  l10n: de.po: translate "track" as "versionieren"
  l10n: zh_CN: translations for git v2.3.0-rc0
  l10n: sv.po: Update Swedish translation (2298t0f0u)
  l10n: fr.po v2.3.0 round 1
  l10n: vi.po(2298t): Updated and change Plural-Forms
  l10n: git.pot: v2.3.0 round 1 (13 new, 11 removed)
  l10n: ca.po: various fixes

* sh/asciidoc-git-version-fix:
  Documentation: fix version numbering