Git v2.3 Release Notes
======================
+This one ended up to be a release with lots of small corrections and
+improvements without big uncomfortably exciting features. The recent
+security fix that went to 2.2.1 and older maintenance tracks is also
+contained in this update.
+
+
Updates since v2.2
------------------
Ports
- *
+ * Recent gcc toolchain on Cygwin started throwing compilation warning,
+ which has been squelched.
+
+ * A few updates to build on platforms that lack tv_nsec,
+ clock_gettime, CLOCK_MONOTONIC and HMAC_CTX_cleanup (e.g. older
+ RHEL) have been added.
+
UI, Workflows & Features
note from the object but with --allow-empty we will store a
(surprise!) note that is empty.
+ * "git interpret-trailers" learned to properly handle the
+ "Conflicts:" block at the end.
+
+ * "git am" learned "--message-id" option to copy the message ID of
+ the incoming e-mail to the log message of resulting commit.
+
+ * "git clone --reference=<over there>" learned the "--dissociate"
+ option to go with it; it borrows objects from the reference object
+ store while cloning only to reduce network traffic and then
+ dissociates the resulting clone from the reference by performing
+ local copies of borrowed objects.
+
+ * "git send-email" learned "--transfer-encoding" option to force a
+ non-fault Content-Transfer-Encoding header (e.g. base64).
+
+ * "git send-email" normally identifies itself via X-Mailer: header in
+ the message it sends out. A new command line flag --no-xmailer
+ allows the user to squelch the header.
+
+ * "git push" into a repository with a working tree normally refuses
+ to modify the branch that is checked out. The command learned to
+ optionally do an equivalent of "git reset --hard" only when there
+ is no change to the working tree and the index instead, which would
+ be useful to "deploy" by pushing into a repository.
+
+ * "git new-workdir" (in contrib/) can be used to populate an empty
+ and existing directory now.
+
+ * Credential helpers are asked in turn until one of them give
+ positive response, which is cumbersome to turn off when you need to
+ run Git in an automated setting. The credential helper interface
+ learned to allow a helper to say "stop, don't ask other helpers."
+ Also GIT_TERMINAL_PROMPT environment can be set to false to disable
+ our built-in prompt mechanism for passwords.
+
+ * "git branch -d" (delete) and "git branch -m" (move) learned to
+ honor "-f" (force) flag; unlike many other subcommands, the way to
+ force these have been with separate "-D/-M" options, which was
+ inconsistent.
+
+ * "diff-highlight" filter (in contrib/) allows its color output to be
+ customized via configuration variables.
+
+ * "git imap-send" learned to take "-v" (verbose) and "-q" (quiet)
+ command line options.
+
+ * "git remote add $name $URL" is now allowed when "url.$URL.insteadOf"
+ is already defined.
+
+ * "git imap-send" now can be built to use cURL library to talk to
+ IMAP servers (if the library is recent enough, of course).
+ This allows you to use authenticate method other than CRAM-MD5,
+ among other things.
+
+ * "git imap-send" now allows GIT_CURL_VERBOSE environment variable to
+ control the verbosity when talking via the cURL library.
+
+ * The prompt script (in contrib/) learned to optionally hide prompt
+ when in an ignored directory by setting GIT_PS1_HIDE_IF_PWD_IGNORED
+ shell variable.
+
Performance, Internal Implementation, Development Support etc.
- *
+ * Earlier we made "rev-list --object-edge" more aggressively list the
+ objects at the edge commits, in order to reduce number of objects　
+ fetched into a shallow repository, but the change affected cases
+ other than "fetching into a shallow repository" and made it
+ unusably slow (e.g. fetching into a normal repository should not
+ have to suffer the overhead from extra processing). Limit it to a
+ more specific case by introducing --objects-edge-aggressive, a new
+ option to rev-list.
+
+ * Squelched useless compiler warnings on Mac OS X regarding the
+ crypto API.
+
+ * The procedure to generate unicode table has been simplified.
+
+ * Some filesystems assign filemodes in a strange way, fooling then
+ automatic "filemode trustability" check done during a new
+ repository creation. The initialization codepath has been hardened
+ against this issue.
+
+ * The codepath in "git remote update --prune" to drop many refs has
+ been optimized.
+
+ * The API into get_merge_bases*() family of functions was easy to
+ misuse, which has been corrected to make it harder to do so.
+
+ * Long overdue departure from the assumption that S_IFMT is shared by
+ everybody made in 2005, which was necessary to port to z/OS.
+
+ * "git push" and "git fetch" did not communicate an overlong refname
+ correctly. Now it uses 64kB sideband to accommodate longer ones.
+
+ * Recent GPG changes the keyring format and drops support for RFC1991
+ formatted signatures, breaking our existing tests.
+
+ * "git-prompt" (in contrib/) used a variable from the global scope,
+ possibly contaminating end-user's namespace.
Also contains various documentation updates and code clean-ups.
track are contained in this release (see the maintenance releases'
notes for details).
+ * "git http-push" over WebDAV (aka dumb http-push) was broken in
+ v2.2.2 when parsing a symbolic ref, resulting in a bogus request
+ that gets rejected by recent versions of cURL library.
+ (merge f6786c8 jk/http-push-symref-fix later to maint).
+
+ * The logic in "git bisect bad HEAD" etc. to avoid forcing the test
+ of the common ancestor of bad and good commits was broken.
+ (merge 07913d5 cc/bisect-rev-parsing later to maint).
+
+ * "git checkout-index --temp=$target $path" did not work correctly
+ for paths outside the current subdirectory in the project.
+ (merge 74c4de5 es/checkout-index-temp later to maint).
+
+ * The report from "git checkout" on a branch that builds on another
+ local branch by setting its branch.*.merge to branch name (not a
+ full refname) incorrectly said that the upstream is gone.
+ (merge 05e7368 jc/checkout-local-track-report later to maint).
+
+ * With The git-prompt support (in contrib/), using the exit status of
+ the last command in the prompt, e.g. PS1='$(__git_ps1) $? ', did
+ not work well, because the helper function stomped on the exit
+ status.
+ (merge 6babe76 tf/prompt-preserve-exit-status later to maint).
+
+ * Recent update to "git commit" broke amending an existing commit
+ with bogus author/committer lines without a valid e-mail address.
+ (merge c83a509 jk/commit-date-approxidate later to maint).
+
+ * The lockfile API used to get confused which file to clean up when
+ the process moved the $cwd after creating a lockfile.
+ (merge fa137f6 nd/lockfile-absolute later to maint).
+
+ * Traditionally we tried to avoid interpreting date strings given by
+ the user as future dates, e.g. GIT_COMMITTER_DATE=2014-12-10 when
+ used early November 2014 was taken as "October 12, 2014" because it
+ is likely that a date in the future, December 10, is a mistake.
+ This heuristics has been loosened to allow people to express future
+ dates (most notably, --until=<date> may want to be far in the
+ future) and we no longer tiebreak by future-ness of the date when
+
+ (1) ISO-like format is used, and
+ (2) the string can make sense interpreted as both y-m-d and y-d-m.
+
+ Git may still have to use the heuristics to tiebreak between dd/mm/yy
+ and mm/dd/yy, though.
+ (merge d372395 jk/approxidate-avoid-y-d-m-over-future-dates later to maint).
+
+ * Git did not correctly read an overlong refname from a packed refs
+ file.
+ (merge ea41783 jk/read-packed-refs-without-path-max later to maint).
+
+ * "git apply" was described in the documentation to take --ignore-date
+ option, which it does not.
+ (merge 0cef4e7 rw/apply-does-not-take-ignore-date later to maint).
+
+ * "git add -i" did not notice when the interactive command input
+ stream went away and kept asking the same question.
+ (merge a8bec7a jk/add-i-read-error later to maint).
+
+ * "git send-email" did not handle RFC 2047 encoded headers quite
+ right.
+ (merge ab47e2a rd/send-email-2047-fix later to maint).
+
+ * New tag object format validation added in 2.2 showed garbage after
+ a tagname it reported in its error message.
+ (merge a1e920a js/fsck-tag-validation later to maint).
+
+ * The code that reads the reflog from the newer to the older entries
+ did not handle an entry that crosses a boundary of block it uses to
+ read them correctly.
+ (merge 69216bf jk/for-each-reflog-ent-reverse later to maint).
+
+ * "git diff -B -M" after making a new copy B out of an existing file
+ A and then editing A extensively ought to report that B was created
+ by copying A and A was modified, which is what "git diff -C"
+ reports, but it instead said A was renamed to B and A was edited
+ heavily in place. This was not just incoherent but also failed to
+ apply with "git apply". The report has been corrected to match what
+ "git diff -C" produces for this case.
+ (merge 6936b58 jc/diff-b-m later to maint).
+
+ * In files we pre-populate for the user to edit with commented hints,
+ a line of hint that is indented with a tab used to show as '#' (or
+ any comment char), ' ' (space), and then the hint text that began
+ with the tab, which some editors flag as an indentation error (tab
+ following space). We now omit the space after the comment char in
+ such a case.
+ (merge d55aeb7 jc/strbuf-add-lines-avoid-sp-ht-sequence later to maint).
+
+ * "git ls-tree" does not support path selection based on negative
+ pathspecs, but did not error out when negative pathspecs are given.
+ (merge f1f6224 nd/ls-tree-pathspec later to maint).
+
+ * The function sometimes returned a non-freeable memory and some
+ other times returned a piece of memory that must be freed, leading
+ to inevitable leaks.
+ (merge 59362e5 jc/exec-cmd-system-path-leak-fix later to maint).
+
+ * The code to abbreviate an object name to its short unique prefix
+ has been optimized when no abbreviation was requested.
+ (merge 61e704e mh/find-uniq-abbrev later to maint).
+
+ * "git add --ignore-errors ..." did not ignore an error to
+ give a file that did not exist.
+ (merge 1d31e5a mg/add-ignore-errors later to maint).
+
* "git checkout $treeish $path", when $path in the index and the
working tree already matched what is in $treeish at the $path,
still overwrote the $path unnecessarily.
configuration file .git/config with x-bit by mistake.
(merge 1f32ecf mh/config-flip-xbit-back-after-checking later to maint).
+ * Recent update in Git 2.2 started creating objects/info/packs and
+ info/refs files with permission bits tighter than user's umask.
+ (merge d91175b jk/prune-packed-server-info later to maint).
+
* Git 2.0 was supposed to make the "simple" mode for the default of
"git push", but it didn't.
(merge 00a6fa0 jk/push-simple later to maint).
* The build procedure did not bother fixing perl and python scripts
when NO_PERL and NO_PYTHON build-time configuration changed.
(merge ca2051d jk/rebuild-perl-scripts-with-no-perl-seting-change later to maint).
+
+ * The usage string of "git log" command was marked incorrectly for
+ l10n.
+ (merge e66dc0c km/log-usage-string-i18n later to maint).
+
+ * "git for-each-ref" mishandled --format="%(upstream:track)" when a
+ branch is marked to have forked from a non-existing branch.
+ (merge b6160d9 rc/for-each-ref-tracking later to maint).