Git v1.8.3-rc1

A release candidate Git v1.8.3-rc1 is now available for testing
at the usual places.

The release tarballs are found at:

and their SHA-1 checksums are:

68160a9e9246a4857ccab0e68b466e0e442c1da5  git-1.8.3.rc1.tar.gz
eecfe00a46a4b26b1f1a4a83f66aca2ab8b264f2  git-htmldocs-1.8.3.rc1.tar.gz
4ea043cdf8c716dddb21f60f6941cda6a12fad9e  git-manpages-1.8.3.rc1.tar.gz

Also the following public repositories all have a copy of the v1.8.3-rc1
tag and the master branch that the tag points at:

url =
url = git://
url =
url = git://
url = git://
url =

Git v1.8.3 Release Notes (draft)

Backward compatibility notes (for Git 2.0)

When “git push [$there]” does not say what to push, we have used the
traditional “matching” semantics so far (all your branches were sent
to the remote as long as there already are branches of the same name
over there). In Git 2.0, the default will change to the “simple”
semantics that pushes only the current branch to the branch with the same
name, and only when the current branch is set to integrate with that
remote branch. Use the user preference configuration variable
“push.default” to change this. If you are an old-timer who is used
to the “matching” semantics, you can set the variable to “matching”
to keep the traditional behaviour. If you want to live in the future
early, you can set it to “simple” today without waiting for Git 2.0.

When “git add -u” (and “git add -A”) is run inside a subdirectory and
does not specify which paths to add on the command line, it
will operate on the entire tree in Git 2.0 for consistency
with “git commit -a” and other commands. There will be no
mechanism to make plain “git add -u” behave like “git add -u .”.
Current users of “git add -u” (without a pathspec) should start
training their fingers to explicitly say “git add -u .”
before Git 2.0 comes. A warning is issued when these commands are
run without a pathspec and when you have local changes outside the
current directory, because the behaviour in Git 2.0 will be different
from today’s version in such a situation.

In Git 2.0, “git add <path>” will behave as “git add -A <path>”, so
that “git add dir/” will notice paths you removed from the directory
and record the removal. Versions before Git 2.0, including this
release, will keep ignoring removals, but the users who rely on this
behaviour are encouraged to start using “git add –ignore-removal <path>”
now before 2.0 is released.

Updates since v1.8.2

Foreign interface

  • remote-hg and remote-bzr helpers (in contrib/) have been updated.

UI, Workflows & Features

  • The prompt string generator (in contrib/completion/) learned to
    show how many changes there are in total and how many have been
    replayed during a “git rebase” session.

  • “git branch –vv” learned to paint the name of the branch it
    integrates with in a different color (color.branch.upstream,
    which defaults to blue).

  • In a sparsely populated working tree, “git checkout <pathspec>” no
    longer unmarks paths that match the given pathspec that were
    originally ignored with “–sparse” (use –ignore-skip-worktree-bits
    option to resurrect these paths out of the index if you really want

  • “git log –format” specifier learned %C(auto) token that tells Git
    to use color when interpolating %d (decoration), %h (short commit
    object name), etc. for terminal output.

  • “git bisect” leaves the final outcome as a comment in its bisect
    log file.

  • “git clone –reference” can now refer to a gitfile “textual symlink”
    that points at the real location of the repository.

  • “git count-objects” learned “–human-readable” aka “-H” option to
    show various large numbers in Ki/Mi/GiB scaled as necessary.

  • “git cherry-pick $blob” and “git cherry-pick $tree” are nonsense,
    and a more readable error message e.g. “can’t cherry-pick a tree”
    is given (we used to say “expected exactly one commit”).

  • The “–annotate” option to “git send-email” can be turned on (or
    off) by default with sendemail.annotate configuration variable (you
    can use –no-annotate from the command line to override it).

  • The “–cover-letter” option to “git format-patch” can be turned on
    (or off) by default with format.coverLetter configuration
    variable. By setting it to auto, you can turn it on only for a
    series with two or more patches.

  • The bash completion support (in contrib/) learned that cherry-pick
    takes a few more options than it already knew about.

  • “git help” learned “-g” option to show the list of guides just like
    list of commands are given with “-a”.

  • A triangular “pull from one place, push to another place” workflow
    is supported better by new remote.pushdefault (overrides the
    “origin” thing) and branch..pushremote (overrides the
    .remote) configuration variables.

  • “git status” learned to report that you are in the middle of a
    revert session, just like it does for a cherry-pick and a bisect

  • The handling by “git branch –set-upstream-to” against various forms
    of erroneous inputs was suboptimal and has been improved.

  • When the interactive access to git-shell is not enabled, it issues
    a message meant to help the system administrator to enable it.
    An explicit way to help the end users who connect to the service by
    issuing custom messages to refuse such an access has been added.

  • In addition to the case where the user edits the log message with
    the “e)dit” option of “am -i”, replace the “Applying: this patch”
    message with the final log message contents after applymsg hook
    munges it.

  • “git status” suggests users to look into using –untracked=no option
    when it takes too long.

  • “git status” shows a bit more information during a
    rebase/bisect session.

  • “git fetch” learned to fetch a commit at the tip of an unadvertised
    ref by specifying a raw object name from the command line when the
    server side supports this feature.

  • Output from “git log –graph” works better with submodule log
    output now.

  • “git count-objects -v” learned to report leftover temporary
    packfiles and other garbage in the object store.

  • A new read-only credential helper (in contrib/) to interact with
    the .netrc/.authinfo files has been added.

  • “git send-email” can be used with the credential helper system.

  • There was no Porcelain way to say “I no longer am interested in
    this submodule”, once you express your interest in a submodule with
    “submodule init”. “submodule deinit” is the way to do so.

  • “git pull –rebase” learned to pass “-v/-q” options to underlying
    “git rebase”.

  • The new “–follow-tags” option tells “git push” to push relevant
    annotated tags when pushing branches out.

  • “git merge” and “git pull” can optionally be told to inspect and
    reject when merging a commit that does not carry a trusted GPG

  • “git mergetool” now feeds files to the “p4merge” backend in the
    order that matches the p4 convention, where “theirs” is usually
    shown on the left side, which is the opposite from what other backends

  • “show/log” now honors gpg.program configuration just like other
    parts of the code that use GnuPG.

  • “git log” that shows the difference between the parent and the
    child has been optimized somewhat.

  • “git difftool” allows the user to write into the temporary files
    being shown; if the user makes changes to the working tree at the
    same time, it now refrains from overwriting the copy in the working
    tree and leaves the temporary file so that changes can be merged

  • There was no good way to ask “I have a random string that came from
    outside world. I want to turn it into a 40-hex object name while
    making sure such an object exists”. A new peeling suffix ^{object}
    can be used for that purpose, together with “rev-parse –verify”.

Performance, Internal Implementation, etc.

  • Updates for building under msvc.

  • A handful of issues in the code that traverses the working tree to find
    untracked and/or ignored files have been fixed, and the general
    codepath involved in “status -u” and “clean” have been cleaned up
    and optimized.

  • The stack footprint of some codepaths that access an object from a
    pack has been shrunk.

  • The logic to coalesce the same lines removed from the parents in
    the output from “diff -c/–cc” has been updated, but with O(n^2)
    complexity, so this might turn out to be undesirable.

  • The code to enforce permission bits on files in $GIT_DIR/ for
    shared repositories has been simplified.

  • A few codepaths know how much data they need to put in the
    hashtables they use when they start, but still began with small tables
    and repeatedly grew and rehashed them.

  • The API to walk reflog entries from the latest to older, which was
    necessary for operations such as “git checkout -“, was cumbersome
    to use correctly and also inefficient.

  • Codepaths that inspect log-message-to-be and decide when to add a
    new Signed-off-by line in various commands have been consolidated.

  • The pkt-line API, implementation and its callers have been cleaned
    up to make them more robust.

  • The Cygwin port has a faster-but-lying lstat(2) emulation whose
    incorrectness does not matter in practice except for a few
    codepaths, and setting permission bits on directories is a codepath
    that needs to use a more correct one.

  • “git checkout” had repeated pathspec matches on the same paths,
    which have been consolidated. Also a bug in “git checkout dir/”
    that is started from an unmerged index has been fixed.

  • A few bugfixes to “git rerere” working on corner case merge
    conflicts have been applied.

Also contains various documentation updates and code clean-ups.

Fixes since v1.8.2

Unless otherwise noted, all the fixes since v1.8.2 in the maintenance
track are contained in this release (see release notes to them for

  • Recent versions of File::Temp (used by “git svn”) started blowing
    up when its tempfile sub is called as a class method; updated the
    callsite to call it as a plain vanilla function to fix it.
    (merge eafc2dd hb/git-pm-tempfile later to maint).

  • Various subcommands of “git remote” simply ignored extraneous
    command line arguments instead of diagnosing them as errors.
    (merge b17dd3f tr/remote-tighten-commandline-parsing later to maint).

  • When receive-pack detects an error in the pack header it received in
    order to decide which of unpack-objects or index-pack to run, it
    returned without closing the error stream, which led to a hung
    sideband thread.

  • Zsh completion forgot that the % character used to signal untracked
    files needs to be escaped with another %.

  • A commit object whose author or committer ident are malformed
    crashed some code that trusted that a name, an email and a
    timestamp can always be found in it.

  • When “upload-pack” fails while generating a pack in response to
    “git fetch” (or “git clone”), the receiving side had
    a programming error that triggered the die handler

  • “rev-list –stdin” and friends kept bogus pointers into the input
    buffer around as human readable object names. This was not a huge
    problem but was exposed by a new change that uses these names in
    error output.
    (merge 70d26c6 tr/copy-revisions-from-stdin later to maint).

  • Smart-capable HTTP servers were not restricted via the
    GIT_NAMESPACE mechanism when talking with commit-walking clients,
    like they are when talking with smart HTTP clients.
    (merge 6130f86 jk/http-dumb-namespaces later to maint).

  • “git merge-tree” did not omit a merge result that is identical to
    the “our” side in certain cases.
    (merge aacecc3 jk/merge-tree-added-identically later to maint).

  • Perl scripts like “git-svn” closed (instead of redirecting to /dev/null)
    the standard error stream, which is not a very smart thing to do.
    A later open may return file descriptor #2 for an unrelated purpose, and
    error reporting code may write into it.

  • “git show-branch” was not prepared to show a very long run of
    ancestor operators e.g. foobar22222…^24 correctly.

  • “git diff –diff-algorithm algo” is also understood as “git diff

  • The new core.commentchar configuration was not applied in a few

  • “git bundle” erroneously bailed out when parsing a valid bundle
    containing a prerequisite commit without a commit message.

  • “git log -S/-G” started paying attention to textconv filter, but
    there was no way to disable this. Make it honor the –no-textconv

  • When used with the “-d temporary-directory” option, “git filter-branch”
    failed to come back to the original working tree to perform the
    final clean-up procedure.

  • “git merge $(git rev-parse v1.8.2)” behaved quite differently from
    “git merge v1.8.2”, as if v1.8.2 were written as v1.8.2^0 and did
    not pay much attention to the annotated tag payload. Make the code
    notice the type of the tag object, in addition to the dwim_ref()
    based classification the current code uses (i.e. the name appears
    in refs/tags/) to decide when to special-case tag merging.

  • Fix a 1.8.1.x regression that stopped matching “dir” (without a
    trailing slash) to a directory “dir”.
    (merge efa5f82 jc/directory-attrs-regression-fix later to maint-1.8.1).

  • “git apply –whitespace=fix” was not prepared to see a line getting
    longer after fixing whitespaces (e.g. tab-in-indent aka Python).
    (merge 329b26e jc/apply-ws-fix-tab-in-indent later to maint-1.8.1).

  • The prompt string generator (in contrib/completion/) did not notice
    when we are in a middle of a “git revert” session.

  • “submodule summary –summary-limit” option did not support the
    “–option=value” form.

  • “index-pack –fix-thin” used an uninitialized value to compute
    the delta depths of objects it appends to the resulting pack.

  • “index-pack –verify-stat” used a few counters outside the protection
    of a mutex, possibly showing incorrect numbers.

  • The code to keep track of what directory names are known to Git on
    platforms with case insensitive filesystems could get confused upon a
    hash collision between these pathnames and would loop forever.

  • Annotated tags outside the refs/tags/ hierarchy were not advertised
    correctly to ls-remote and fetch with recent versions of Git.

  • Recent optimizations broke shallow clones.

  • “git cmd — :(top” was not diagnosed as an invalid syntax, and
    instead the parser kept reading beyond the end of the string.

  • “git tag -f <tag>” always said “Updated tag <tag>” even when
    creating a new tag (i.e. neither overwriting nor updating).

  • “git p4” did not behave well when the path to the root of the P4
    client was not its real path.
    (merge bbd8486 pw/p4-symlinked-root later to maint).

  • “git archive” reported a failure when asked to create an archive out
    of an empty tree. It is more intuitive to give an empty
    archive back in such a case.

  • When “format-patch” quoted a non-ascii string in header files,
    it incorrectly applied rfc2047 and chopped a single character in
    the middle of the string.

  • An aliased command spawned from a bare repository that does not say
    it is bare with “core.bare = yes” was treated as non-bare by mistake.

  • In “git reflog expire”, the REACHABLE bit was not cleared from the
    correct objects.

  • The logic used by “git diff -M –stat” to shorten the names of
    files before and after a rename did not work correctly when the
    common prefix and suffix between the two filenames overlapped.

  • The “–match=<pattern>” option of “git describe”, when used with
    “–all” to allow refs that are not annotated tags to be a
    base of description, did not restrict the output from the command
    to those refs that match the given pattern.

  • Clarify in the documentation “what” gets pushed to “where” when the
    command line to “git push” does not say these explicitly.

  • The “–color=<when>” argument to the commands in the diff family
    was described poorly.

  • The arguments given to the pre-rebase hook were not documented.

  • The v4 index format was not documented.

  • The “–match=<pattern>” argument “git describe” takes uses glob
    pattern but it wasn’t obvious from the documentation.

  • Some sources failed to compile on systems that lack NI_MAXHOST in
    their system header (e.g. z/OS).

  • Add an example use of “–env-filter” in “filter-branch”

  • “git bundle verify” did not say “records a complete history” for a
    bundle that does not have any prerequisites.

  • In the v1.8.0 era, we changed symbols that do not have to be global
    to file scope static, but a few functions in graph.c were used by
    CGit sideways, bypassing the entry points of the API the
    in-tree users use.

  • “git update-index -h” did not do the usual “-h(elp)” thing.

  • “git index-pack” had a buffer-overflow while preparing an
    informational message when the translated version of it was too

  • git commit -m “$msg” used to add an extra newline even when
    $msg already ended with one.

  • The SSL peer verification done by “git imap-send” did not ask for
    Server Name Indication (RFC 4366), failing to connect to SSL/TLS
    sites that serve multiple hostnames on a single IP.

  • perl/ slurped everything in core only to write it
    out to a file descriptor, which was not a very smart thing to do.

  • “git branch” did not bother to check nonsense command line
    parameters. It now issues errors in many cases.

  • Verification of signed tags was not done correctly when not in C
    or en/US locale.

  • Some platforms and users spell UTF-8 differently; retry with the
    most official “UTF-8” when the system does not understand the
    user-supplied encoding name that is a common alternative
    spelling of UTF-8.

  • When export-subst is used, “zip” output recorded an incorrect
    size of the file.

  • “git am $maildir/” applied messages in an unexpected order; sort
    filenames read from the maildir/ in a way that is more likely to
    sort the messages in the order the writing MUA meant to, by sorting
    numeric segments in numeric order and non-numeric segments in
    alphabetical order.

  • “git submodule update”, when recursed into sub-submodules, did not
    accumulate the prefix paths.

Changes since v1.8.3-rc0 are as follows:

Anders Granskogen Bjørnstad (1):
      Documentation/git-commit: Typo under --edit

Felipe Contreras (5):
      completion: add missing format-patch options
      documentation: trivial whitespace cleanups
      complete: zsh: trivial simplification
      complete: zsh: use zsh completion for the main cmd
      completion: zsh: don't override suffix on _detault

H. Merijn Brand (1): call tempfile from File::Temp as a regular function

John Keeping (2):
      merge-tree: fix typo in "both changed identically"
      t/Makefile: remove smoke test targets

Junio C Hamano (2):
      Update draft release notes to 1.8.3
      Git 1.8.3-rc1

Marc Branchaud (1):
      Fix grammar in the 1.8.3 release notes.

Ramkumar Ramachandra (5):
      git-completion.bash: lexical sorting for diff.statGraphWidth
      git-completion.bash: add diff.submodule to config list
      git-completion.bash: complete branch.*.rebase as boolean
      git-completion.bash: add branch.*.pushremote to config list
      git-completion.bash: add remote.pushdefault to config list

Ramsay Allan Jones (2):
      clone: Make the 'junk_mode' symbol a file static
      pretty: Fix bug in truncation support for %>, %< and %><

René Scharfe (3):
      pretty: simplify input line length calculation in pp_user_info()
      pretty: simplify output line length calculation in pp_user_info()
      pretty: remove intermediate strbufs from pp_user_info()

Thomas Rast (4):
      remote: add a test for extra arguments, according to docs
      remote: check for superfluous arguments in 'git remote add'
      remote: 'show' and 'prune' can take more than one remote
      unpack_entry: avoid freeing objects in base cache

Zoltan Klinger (1): show where rebase is at when stopped

Junio C Hamano wrote on 03 May 2013