path: root/Documentation/RelNotes/2.20.0.txt

Git 2.20 Release Notes
======================

Backward Compatibility Notes
----------------------------

 * "git branch -l <foo>" used to be a way to ask a reflog to be
   created while creating a new branch, but that is no longer the
   case.  It is a short-hand for "git branch --list <foo>" now.

 * "git push" into refs/tags/* hierarchy is rejected without getting
   forced, but "git fetch" (misguidedly) used the "fast forwarding"
   rule used for the refs/heads/* hierarchy; this has been corrected,
   which means some fetches of tags that did not fail with older
   version of Git will fail without "--force" with this version.

 * "git help -a" now gives verbose output (same as "git help -av").
   Those who want the old output may say "git help --no-verbose -a"..

 * "git cpn --help", when "cpn" is an alias to, say, "cherry-pick -n",
   reported only the alias expansion of "cpn" in earlier versions of
   Git.  It now runs "git cherry-pick --help" to show the manual page
   of the command, while sending the alias expansion to the standard
   error stream.

 * "git send-email" learned to grab address-looking string on any
   trailer whose name ends with "-by". This is a backward-incompatible
   change.  Adding "--suppress-cc=misc-by" on the command line, or
   setting sendemail.suppresscc configuration variable to "misc-by",
   can be used to disable this behaviour.


Updates since v2.19
-------------------

UI, Workflows & Features

 * Running "git clone" against a project that contain two files with
   pathnames that differ only in cases on a case insensitive
   filesystem would result in one of the files lost because the
   underlying filesystem is incapable of holding both at the same
   time.  An attempt is made to detect such a case and warn.

 * "git checkout -b newbranch [HEAD]" should not have to do as much as
   checking out a commit different from HEAD.  An attempt is made to
   optimize this special case.

 * "git rev-list --stdin </dev/null" used to be an error; it now shows
   no output without an error.  "git rev-list --stdin --default HEAD"
   still falls back to the given default when nothing is given on the
   standard input.

 * Lift code from GitHub to restrict delta computation so that an
   object that exists in one fork is not made into a delta against
   another object that does not appear in the same forked repository.

 * "git format-patch" learned new "--interdiff" and "--range-diff"
   options to explain the difference between this version and the
   previous attempt in the cover letter (or after the three-dashes as
   a comment).

 * "git mailinfo" used in "git am" learned to make a best-effort
   recovery of a patch corrupted by MUA that sends text/plain with
   format=flawed option.
   (merge 3aa4d81f88 rs/mailinfo-format-flowed later to maint).

 * The rules used by "git push" and "git fetch" to determine if a ref
   can or cannot be updated were inconsistent; specifically, fetching
   to update existing tags were allowed even though tags are supposed
   to be unmoving anchoring points.  "git fetch" was taught to forbid
   updates to existing tags without the "--force" option.

 * "git multi-pack-index" learned to detect corruption in the .midx
   file it uses, and this feature has been integrated into "git fsck".

 * Generation of (experimental) commit-graph files have so far been
   fairly silent, even though it takes noticeable amount of time in a
   meaningfully large repository.  The users will now see progress
   output.

 * The minimum version of Windows supported by Windows port of Git is
   now set to Vista.

 * The completion script (in contrib/) learned to complete a handful of
   options "git stash list" command takes.

 * The completion script (in contrib/) learned that "git fetch
   --multiple" only takes remote names as arguments and no refspecs.

 * "git status" learns to show progress bar when refreshing the index
   takes a long time.
   (merge ae9af12287 nd/status-refresh-progress later to maint).

 * "git help -a" and "git help -av" give different pieces of
   information, and generally the "verbose" version is more friendly
   to the new users.  "git help -a" by default now uses the more
   verbose output (with "--no-verbose", you can go back to the
   original).  Also "git help -av" now lists aliases and external
   commands, which it did not used to.

 * Unlike "grep", "git grep" by default recurses to the whole tree.
   The command learned "git grep --recursive" option, so that "git
   grep --no-recursive" can serve as a synonym to setting the
   max-depth to 0.

 * When pushing into a repository that borrows its objects from an
   alternate object store, "git receive-pack" that responds to the
   push request on the other side lists the tips of refs in the
   alternate to reduce the amount of objects transferred.  This
   sometimes is detrimental when the number of refs in the alternate
   is absurdly large, in which case the bandwidth saved in potentially
   fewer objects transferred is wasted in excessively large ref
   advertisement.  The alternate refs that are advertised are now
   configurable with a pair of configuration variables.

 * "git cmd --help" when "cmd" is aliased used to only say "cmd is
   aliased to ...".  Now it shows that to the standard error stream
   and runs "git $cmd --help" where $cmd is the first word of the
   alias expansion.

 * The documentation of "git gc" has been updated to mention that it
   is no longer limited to "pruning away cruft" but also updates
   ancillary files like commit-graph as a part of repository
   optimization.

 * "git p4 unshelve" improvements.

 * The logic to select the default user name and e-mail on Windows has
   been improved.
   (merge 501afcb8b0 js/mingw-default-ident later to maint).

 * The "rev-list --filter" feature learned to exclude all trees via
   "tree:0" filter.

 * "git send-email" learned to grab address-looking string on any
   trailer whose name ends with "-by"; --suppress-cc=misc-by on the
   command line, or setting sendemail.suppresscc configuration
   variable to "misc-by", can be used to disable this behaviour.

 * "git mergetool" learned to take the "--[no-]gui" option, just like
   "git difftool" does.

 * "git rebase -i" learned a new insn, 'break', that the user can
   insert in the to-do list.  Upon hitting it, the command returns
   control back to the user.

 * New "--pretty=format:" placeholders %GF and %GP that show the GPG
   key fingerprints have been invented.

 * On platforms with recent cURL library, http.sslBackend configuration
   variable can be used to choose a different SSL backend at runtime.
   The Windows port uses this mechanism to switch between OpenSSL and
   Secure Channel while talking over the HTTPS protocol.

 * "git send-email" learned to disable SMTP authentication via the
   "--smtp-auth=none" option, even when the smtp username is given
   (which turns the authentication on by default).

 * A fourth class of configuration files (in addition to the
   traditional "system wide", "per user in the $HOME directory" and
   "per repository in the $GIT_DIR/config") has been introduced so
   that different worktrees that share the same repository (hence the
   same $GIT_DIR/config file) can use different customization.

 * A pattern with '**' that does not have a slash on either side used
   to be an invalid one, but the code now treats such double-asterisks
   the same way as two normal asterisks that happen to be adjacent to
   each other.
   (merge e5bbe09e88 nd/wildmatch-double-asterisk later to maint).

 * The "--no-patch" option, which can be used to get a high-level
   overview without the actual line-by-line patch difference shown, of
   the "range-diff" command was earlier broken, which has been
   corrected.

 * The recently merged "rebase in C" has an escape hatch to use the
   scripted version when necessary, but it hasn't been documented,
   which has been corrected.


Performance, Internal Implementation, Development Support etc.

 * Developer builds now use -Wunused-function compilation option.

 * One of our CI tests to run with "unusual/experimental/random"
   settings now also uses commit-graph and midx.

 * When there are too many packfiles in a repository (which is not
   recommended), looking up an object in these would require
   consulting many pack .idx files; a new mechanism to have a single
   file that consolidates all of these .idx files is introduced.

 * "git submodule update" is getting rewritten piece-by-piece into C.

 * The code for computing history reachability has been shuffled,
   obtained a bunch of new tests to cover them, and then being
   improved.

 * The unpack_trees() API used in checking out a branch and merging
   walks one or more trees along with the index.  When the cache-tree
   in the index tells us that we are walking a tree whose flattened
   contents is known (i.e. matches a span in the index), as linearly
   scanning a span in the index is much more efficient than having to
   open tree objects recursively and listing their entries, the walk
   can be optimized, which has been done.

 * When creating a thin pack, which allows objects to be made into a
   delta against another object that is not in the resulting pack but
   is known to be present on the receiving end, the code learned to
   take advantage of the reachability bitmap; this allows the server
   to send a delta against a base beyond the "boundary" commit.

 * spatch transformation to replace boolean uses of !hashcmp() to
   newly introduced oideq() is added, and applied, to regain
   performance lost due to support of multiple hash algorithms.

 * Fix a bug in which the same path could be registered under multiple
   worktree entries if the path was missing (for instance, was removed
   manually).  Also, as a convenience, expand the number of cases in
   which --force is applicable.

 * Split Documentation/config.txt for easier maintenance.
   (merge 6014363f0b nd/config-split later to maint).

 * Test helper binaries clean-up.
   (merge c9a1f4161f nd/test-tool later to maint).

 * Various tests have been updated to make it easier to swap the
   hash function used for object identification.
   (merge ae0c89d41b bc/hash-independent-tests later to maint).

 * Update fsck.skipList implementation and documentation.
   (merge 371a655074 ab/fsck-skiplist later to maint).

 * An alias that expands to another alias has so far been forbidden,
   but now it is allowed to create such an alias.

 * Various test scripts have been updated for style and also correct
   handling of exit status of various commands.

 * "gc --auto" ended up calling exit(-1) upon error, which has been
   corrected to use exit(1).  Also the error reporting behaviour when
   daemonized has been updated to exit with zero status when stopping
   due to a previously discovered error (which implies there is no
   point running gc to improve the situation); we used to exit with
   failure in such a case.

 * Various codepaths in the core-ish part learned to work on an
   arbitrary in-core index structure, not necessarily the default
   instance "the_index".
   (merge b3c7eef9b0 nd/the-index later to maint).

 * Code clean-up in the internal machinery used by "git status" and
   "git commit --dry-run".
   (merge 73ba5d78b4 ss/wt-status-committable later to maint).

 * Some environment variables that control the runtime options of Git
   used during tests are getting renamed for consistency.
   (merge 4231d1ba99 bp/rename-test-env-var later to maint).

 * A pair of new extensions to the index file have been introduced.
   They allow the index file to be read in parallel for performance.

 * The oidset API was built on top of the oidmap API which in turn is
   on the hashmap API.  Replace the implementation to build on top of
   the khash API and gain performance.

 * Over some transports, fetching objects with an exact commit object
   name can be done without first seeing the ref advertisements.  The
   code has been optimized to exploit this.

 * In a partial clone that will lazily be hydrated from the
   originating repository, we generally want to avoid "does this
   object exist (locally)?" on objects that we deliberately omitted
   when we created the clone.  The cache-tree codepath (which is used
   to write a tree object out of the index) however insisted that the
   object exists, even for paths that are outside of the partial
   checkout area.  The code has been updated to avoid such a check.

 * To help developers, an EditorConfig file that attempts to follow
   the project convention has been added.
   (merge b548d698a0 bc/editorconfig later to maint).

 * The result of coverage test can be combined with "git blame" to
   check the test coverage of code introduced recently with a new
   'coverage-diff' tool (in contrib/).
   (merge 783faedd65 ds/coverage-diff later to maint).

 * An experiment to fuzz test a few areas, hopefully we can gain more
   coverage to various areas.

 * More codepaths are moving away from hardcoded hash sizes.

 * The way the Windows port figures out the current directory has been
   improved.

 * The way DLLs are loaded on the Windows port has been improved.

 * Some tests have been reorganized and renamed; "ls t/" now gives a
   better overview of what is tested for these scripts than before.

 * "git rebase" and "git rebase -i" have been reimplemented in C.

 * Windows port learned to use nano-second resolution file timestamps.

 * The overly large Documentation/config.txt file have been split into
   million little pieces.  This potentially allows each individual piece
   to be included into the manual page of the command it affects more easily.

 * Replace three string-list instances used as look-up tables in "git
   fetch" with hashmaps.

 * Unify code to read the author-script used in "git am" and the
   commands that use the sequencer machinery, e.g. "git rebase -i".

 * In preparation to the day when we can deprecate and remove the
   "rebase -p", make sure we can skip and later remove tests for
   it.

 * The history traversal used to implement the tag-following has been
   optimized by introducing a new helper.

 * The helper function to refresh the cached stat information in the
   in-core index has learned to perform the lstat() part of the
   operation in parallel on multi-core platforms.

 * The code to traverse objects for reachability, used to decide what
   objects are unreferenced and expendable, have been taught to also
   consider per-worktree refs of other worktrees as starting points to
   prevent data loss.

 * "git add" needs to internally run "diff-files" equivalent, and the
   codepath learned the same optimization as "diff-files" has to run
   lstat(2) in parallel to find which paths have been updated in the
   working tree.

 * The procedure to install dependencies before testing at Travis CI
   is getting revamped for both simplicity and flexibility, taking
   advantage of the recent move to the vm-based environment.

 * The support for format-patch (and send-email) by the command-line
   completion script (in contrib/) has been simplified a bit.

 * The revision walker machinery learned to take advantage of the
   commit generation numbers stored in the commit-graph file.

 * The codebase has been cleaned up to reduce "#ifndef NO_PTHREADS".

 * The way -lcurl library gets linked has been simplified by taking
   advantage of the fact that we can just ask curl-config command how.

 * Various functions have been audited for "-Wunused-parameter" warnings
   and bugs in them got fixed.

 * A sanity check for start-up sequence has been added in the config
   API codepath.

 * The build procedure to link for fuzzing test has been made
   customizable with a new Makefile variable.

 * The way "git rebase" parses and forwards the command line options
   meant for underlying "git am" has been revamped, which fixed for
   options with parameters that were not passed correctly.

 * Our testing framework uses a special i18n "poisoned localization"
   feature to find messages that ought to stay constant but are
   incorrectly marked to be translated.  This feature has been made
   into a runtime option (it used to be a compile-time option).

 * "git push" used to check ambiguities between object-names and
   refnames while processing the list of refs' old and new values,
   which was unnecessary (as it knew that it is feeding raw object
   names).  This has been optimized out.

 * The xcurl_off_t() helper function is used to cast size_t to
   curl_off_t, but some compilers gave warnings against the code to
   ensure the casting is done without wraparound, when size_t is
   narrower than curl_off_t.  This warning has been squelched.

 * Code preparation to replace ulong vars with size_t vars where
   appropriate continues.

 * The "test installed Git" mode of our test suite has been updated to
   work better.

 * A coding convention around the Coccinelle semantic patches to have
   two classes to ease code migration process has been proposed and
   its support has been added to the Makefile.

 * The "container" mode of TravisCI is going away.  Our .travis.yml
   file is getting prepared for the transition.
   (merge 32ee384be8 ss/travis-ci-force-vm-mode later to maint).

 * Our test scripts can now take the '-V' option as a synonym for the
   '--verbose-log' option.
   (merge a5f52c6dab sg/test-verbose-log later to maint).


Fixes since v2.19
-----------------

 * "git interpret-trailers" and its underlying machinery had a buggy
   code that attempted to ignore patch text after commit log message,
   which triggered in various codepaths that will always get the log
   message alone and never get such an input.
   (merge 66e83d9b41 jk/trailer-fixes later to maint).

 * Malformed or crafted data in packstream can make our code attempt
   to read or write past the allocated buffer and abort, i