msysGit-based Git for Windows 1.x was superseded by Git for Windows 2.x

[git/mingw/4msysgit.git] / Documentation / RelNotes / 2.1.0.txt

Git v2.1 Release Notes
======================

Backward compatibility notes
----------------------------

 * The default value we give to the environment variable LESS has been
   changed from "FRSX" to "FRX", losing "S" (chop long lines instead
   of wrapping).  Existing users who prefer not to see line-wrapped
   output may want to set

     $ git config core.pager "less -S"

   to restore the traditional behaviour.  It is expected that people
   find output from most subcommands easier to read with the new
   default, except for "blame" which tends to produce really long
   lines.  To override the new default only for "git blame", you can
   do this:

     $ git config pager.blame "less -S"

 * A few disused directories in contrib/ have been retired.


Updates since v2.0
------------------

UI, Workflows & Features

 * Since the very beginning of Git, we gave the LESS environment a
   default value "FRSX" when we spawn "less" as the pager.  "S" (chop
   long lines instead of wrapping) has been removed from this default
   set of options, because it is more or less a personal taste thing,
   as opposed to the others that have good justifications (i.e. "R" is
   very much justified because many kinds of output we produce are
   colored and "FX" is justified because output we produce is often
   shorter than a page).

 * The logic and data used to compute the display width needed for
   UTF-8 strings have been updated to match Unicode 7.0 better.

 * HTTP-based transports learned to better propagate the error messages from
   the webserver to the client coming over the HTTP transport.

 * The completion script for bash (in contrib/) has been updated to
   better handle aliases that define a complex sequence of commands.

 * The "core.preloadindex" configuration variable is enabled by default,
   allowing modern platforms to take advantage of their
   multiple cores.

 * "git clone" applies the "if cloning from a local disk, physically
   copy the repository using hardlinks, unless otherwise told not to with
   --no-local" optimization when the url.*.insteadOf mechanism rewrites a
   remote-repository "git clone $URL" into a
   clone from a local disk.

 * "git commit --date=<date>" option learned more
   timestamp formats, including "--date=now".

 * The `core.commentChar` configuration variable is used to specify a
   custom comment character (other than the default "#") for
   the commit message editor.  This can be set to `auto` to attempt to
   choose a different character that does not conflict with any that
   already starts a line in the message being edited, for cases like
   "git commit --amend".

 * "git format-patch" learned --signature-file=<file> to add the contents
   of a file as a signature to the mail message it produces.

 * "git grep" learned the grep.fullname configuration variable to force
   "--full-name" to be the default.  This may cause regressions for
   scripted users who do not expect this new behaviour.

 * "git imap-send" learned to ask the credential helper for auth
   material.

 * "git log" and friends now understand the value "auto" for the
   "log.decorate" configuration variable to enable the "--decorate"
   option automatically when the output is sent to tty.

 * "git merge" without an argument, even when there is an upstream
   defined for the current branch, refused to run until
   merge.defaultToUpstream is set to true.  Flip the default of that
   configuration variable to true.

 * "git mergetool" learned to drive the vimdiff3 backend.

 * mergetool.prompt used to default to 'true', always asking "do you
   really want to run the tool on this path?".  The default has been
   changed to 'false'.  However, the prompt will still appear if
   mergetool used its autodetection system to guess which tool to use.
   Users who explicitly specify or configure a tool will no longer see
   the prompt by default.

   Strictly speaking, this is a backward incompatible change and
   users need to explicitly set the variable to 'true' if they want
   to be prompted to confirm running the tool on each path.

 * "git replace" learned the "--edit" subcommand to create a
   replacement by editing an existing object.

 * "git replace" learned a "--graft" option to rewrite the parents of a
   commit.

 * "git send-email" learned "--to-cover" and "--cc-cover" options, to
   tell it to copy To: and Cc: headers found in the first input file
   when emitting later input files.

 * "git svn" learned to cope with malformed timestamps with only one
   digit in the hour part, e.g. 2014-01-07T5:01:02.048176Z, emitted
   by some broken subversion server implementations.

 * "git tag" when editing the tag message shows the name of the tag
   being edited as a comment in the editor.

 * "git tag" learned to pay attention to "tag.sort" configuration, to
   be used as the default sort order when no --sort=<value> option
   is given.

 * A new "git verify-commit" command, to check GPG signatures in signed
   commits, in a way similar to "git verify-tag" is used to check
   signed tags, was added.


Performance, Internal Implementation, etc.

 * Build procedure for 'subtree' (in contrib/) has been cleaned up.

 * Support for the profile-feedback build, which has
   bit-rotted for quite a while, has been updated.

 * An experimental format to use two files (the base file and
   incremental changes relative to it) to represent the index has been
   introduced; this may reduce I/O cost of rewriting a large index
   when only small part of the working tree changes.

 * Effort to shrink the size of patches Windows folks maintain on top
   by upstreaming them continues.  More tests that are not applicable
   to the Windows environment are identified and either skipped or
   made more portable.

 * Eradication of "test $condition -a $condition" from our scripts
   continues.

 * The `core.deltabasecachelimit` used to default to 16 MiB , but this
   proved to be too small, and has been bumped to 96 MiB.

 * "git blame" has been optimized greatly by reorganising the data
   structure that is used to keep track of the work to be done.

 * "git diff" that compares 3-or-more trees (e.g. parents and the
   result of a merge) has been optimized.

 * The API to update/delete references are being converted to handle
   updates to multiple references in a transactional way.  As an
   example, "update-ref --stdin [-z]" has been updated to use this
   API.

 * skip_prefix() and strip_suffix() API functions are used a lot more
   widely throughout the codebase now.

 * Parts of the test scripts can be skipped by using a range notation,
   e.g. "sh t1234-test.sh --run='1-4 6 8-'" to omit test piece 5 and 7
   and run everything else.


Also contains various documentation updates and code clean-ups.


Fixes since v2.0
----------------

Unless otherwise noted, all the fixes since v2.0 in the maintenance
track are contained in this release (see the maintenance releases'
notes for details).

 * We used to unconditionally disable the pager in the pager process
   we spawn to feed out output, but that prevented people who want to
   run "less" within "less" from doing so.
   (merge c0459ca je/pager-do-not-recurse later to maint).

 * Tools that read diagnostic output in our standard error stream do
   not want to see terminal control sequence (e.g. erase-to-eol).
   Detect them by checking if the standard error stream is connected
   to a tty.
   (merge 38de156 mn/sideband-no-ansi later to maint).

 * Mishandling of patterns in .gitignore that have trailing SPs quoted
   with backslashes (e.g. ones that end with "\ ") has been
   corrected.
   (merge 97c1364be6b pb/trim-trailing-spaces later to maint).

 * Reworded the error message given upon a failure to open an existing
   loose object file due to e.g. permission issues; it was reported as
   the object being corrupt, but that is not quite true.
   (merge d6c8a05 jk/report-fail-to-read-objects-better later to maint).

 * "git log -2master" is a common typo that shows two commits starting
   from whichever random branch that is not 'master' that happens to
   be checked out currently.
   (merge e3fa568 jc/revision-dash-count-parsing later to maint).

 * Code to avoid adding the same alternate object store twice was
   subtly broken for a long time, but nobody seems to have noticed.
   (merge 80b4785 rs/fix-alt-odb-path-comparison later to maint).
   (merge 539e750 ek/alt-odb-entry-fix later to maint).

 * The "%<(10,trunc)%s" pretty format specifier in the log family of
   commands is used to truncate the string to a given length (e.g. 10
   in the example) with padding to column-align the output, but did
   not take into account that number of bytes and number of display
   columns are different.
   (merge 7d50987 as/pretty-truncate later to maint).

 * "%G" (nothing after G) is an invalid pretty format specifier, but
   the parser did not notice it as garbage.
   (merge 958b2eb jk/pretty-G-format-fixes later to maint).

 * A handful of code paths had to read the commit object more than
   once when showing header fields that are usually not parsed.  The
   internal data structure to keep track of the contents of the commit
   object has been updated to reduce the need for this double-reading,
   and to allow the caller find the length of the object.
   (merge 218aa3a jk/commit-buffer-length later to maint).

 * The "mailmap.file" configuration option did not support tilde
   expansion (i.e. ~user/path and ~/path).
   (merge 9352fd5 ow/config-mailmap-pathname later to maint).

 * The completion scripts (in contrib/) did not know about quite a few
   options that are common between "git merge" and "git pull", and a
   couple of options unique to "git merge".
   (merge 8fee872 jk/complete-merge-pull later to maint).

 * The unix-domain socket used by the sample credential cache daemon
   tried to unlink an existing stale one at a wrong path, if the path
   to the socket was given as an overlong path that does not fit in
   the sun_path member of the sockaddr_un structure.
   (merge 2869b3e rs/fix-unlink-unix-socket later to maint).

 * An ancient rewrite passed a wrong pointer to a curl library
   function in a rarely used code path.
   (merge 479eaa8 ah/fix-http-push later to maint).

 * "--ignore-space-change" option of "git apply" ignored the spaces
   at the beginning of lines too aggressively, which is inconsistent
   with the option of the same name that "diff" and "git diff" have.
   (merge 14d3bb4 jc/apply-ignore-whitespace later to maint).

 * "git blame" miscounted the number of columns needed to show localized
   timestamps, resulting in a jaggy left-side-edge for the source code
   lines in its output.
   (merge dd75553 jx/blame-align-relative-time later to maint).

 * "git blame" assigned the blame to the copy in the working-tree if
   the repository is set to core.autocrlf=input and the file used CRLF
   line endings.
   (merge 4d4813a bc/blame-crlf-test later to maint).

 * "git clone -b brefs/tags/bar" would have mistakenly thought we were
   following a single tag, even though it was a name of the branch,
   because it incorrectly used strstr().
   (merge 60a5f5f jc/fix-clone-single-starting-at-a-tag later to maint).

 * "git commit --allow-empty-message -C $commit" did not work when the
   commit did not have any log message.
   (merge 076cbd6 jk/commit-C-pick-empty later to maint).

 * "git diff --find-copies-harder" sometimes pretended as if the mode
   bits have changed for paths that are marked with the assume-unchanged
   bit.
   (merge 5304810 jk/diff-files-assume-unchanged later to maint).

 * "filter-branch" left an empty single-parent commit that results when
   all parents of a merge commit get mapped to the same commit, even
   under "--prune-empty".
   (merge 79bc4ef cb/filter-branch-prune-empty-degenerate-merges later to maint).

 * "git format-patch" did not enforce the rule that the "--follow"
   option from the log/diff family of commands must be used with
   exactly one pathspec.
   (merge dd63f16 jk/diff-follow-must-take-one-pathspec later to maint).

 * "git gc --auto" was recently changed to run in the background to
   give control back early to the end-user sitting in front of the
   terminal, but it forgot that housekeeping involving reflogs should
   be done without other processes competing for accesses to the refs.
   (merge 62aad18 nd/daemonize-gc later to maint).

 * "git grep -O" to show the lines that hit in the pager did not work
   well with case insensitive search.  We now spawn "less" with its
   "-I" option when it is used as the pager (which is the default).
   (merge f7febbe sk/spawn-less-case-insensitively-from-grep-O-i later to maint).

 * We used to disable threaded "git index-pack" on platforms without
   thread-safe pread(); use a different workaround for such
   platforms to allow threaded "git index-pack".
   (merge 3953949 nd/index-pack-one-fd-per-thread later to maint).

 * The error reporting from "git index-pack" has been improved to
   distinguish missing objects from type errors.
   (merge 77583e7 jk/index-pack-report-missing later to maint).

 * "log --show-signature" incorrectly decided the color to paint a
   mergetag that was and was not correctly validated.
   (merge 42c55ce mg/fix-log-mergetag-color later to maint).

 * "log --show-signature" did not pay attention to the "--graph" option.
   (merge cf3983d zk/log-graph-showsig later to maint).

 * "git mailinfo" used to read beyond the ends of header strings while
   parsing an incoming e-mail message to extract the patch.
   (merge b1a013d rs/mailinfo-header-cmp later to maint).

 * On a case insensitive filesystem, merge-recursive incorrectly
   deleted the file that is to be renamed to a name that is the same
   except for case differences.
   (merge baa37bf dt/merge-recursive-case-insensitive later to maint).

 * Merging changes into a file that ends in an incomplete line made the
   last line into a complete one, even when the other branch did not
   change anything around the end of file.
   (merge ba31180 mk/merge-incomplete-files later to maint).

 * "git pack-objects" unnecessarily copied the previous contents when
   extending the hashtable, even though it will populate the table
   from scratch anyway.
   (merge fb79947 rs/pack-objects-no-unnecessary-realloc later to maint).

 * Recent updates to "git repack" started to duplicate objects that
   are in packfiles marked with the .keep flag into the new packfile by
   mistake.
   (merge d078d85 jk/repack-pack-keep-objects later to maint).

 * "git rerere forget" did not work well when merge.conflictstyle
   was set to a non-default value.
   (merge de3d8bb fc/rerere-conflict-style later to maint).

 * "git remote rm" and "git remote prune" can involve removing many
   refs at once, which is not a very efficient thing to do when very
   many refs exist in the packed-refs file.
   (merge e6bea66 jl/remote-rm-prune later to maint).

 * "git log --exclude=<glob> --all | git shortlog" worked as expected,
   but "git shortlog --exclude=<glob> --all", which is supposed to be
   identical to the above pipeline, was not accepted at the command
   line argument parser level.
   (merge eb07774 jc/shortlog-ref-exclude later to maint).

 * The autostash mode of "git rebase -i" did not restore the dirty
   working tree state if the user aborted the interactive rebase by
   emptying the insn sheet.
   (merge ddb5432 rr/rebase-autostash-fix later to maint).

 * "git rebase --fork-point" did not filter out patch-identical
   commits correctly.

 * During "git rebase --merge", a conflicted patch could not be
   skipped with "--skip" if the next one also conflicted.
   (merge 95104c7 bc/fix-rebase-merge-skip later to maint).

 * "git show -s" (i.e. show log message only) used to incorrectly emit
   an extra blank line after a merge commit.
   (merge ad2f725 mk/show-s-no-extra-blank-line-for-merges later to maint).

 * "git status", even though it is a read-only operation, tries to
   update the index with refreshed lstat(2) info to optimize future
   accesses to the working tree opportunistically, but this could
   race with a "read-write" operation that modifies the index while it
   is running.  Detect such a race and avoid overwriting the index.
   (merge 426ddee ym/fix-opportunistic-index-update-race later to maint).

 * "git status" (and "git commit") behaved as if changes in a modified
   submodule are not there if submodule.*.ignore configuration is set,
   which was misleading.  The configuration is only to unclutter diff
   output during the course of development, and not to hide
   changes in the "status" output to cause the users forget to commit
   them.
   (merge c215d3d jl/status-added-submodule-is-never-ignored later to maint).

 * Documentation for "git submodule sync" forgot to say that the subcommand
   can take the "--recursive" option.
   (merge 9393ae7 mc/doc-submodule-sync-recurse later to maint).

 * "git update-index --cacheinfo" in 2.0 release crashed on a
   malformed command line.
   (merge c8e1ee4 jc/rev-parse-argh-dashed-multi-words later to maint).

 * The mode to run tests with HTTP server tests disabled was broken.
   (merge afa53fe na/no-http-test-in-the-middle later to maint).