Documentation / RelNotes / 1.8.2.txton commit Merge branch 'mh/maint-ceil-absolute' (3e07d26)
   1Git v1.8.2 Release Notes
   2========================
   3
   4Backward compatibility notes
   5----------------------------
   6
   7In the next major release Git 2.0 (not *this* one), we will change the
   8behavior of the "git push" command.
   9
  10When "git push [$there]" does not say what to push, we have used the
  11traditional "matching" semantics so far (all your branches were sent
  12to the remote as long as there already are branches of the same name
  13over there).  We will use the "simple" semantics that pushes the
  14current branch to the branch with the same name, only when the current
  15branch is set to integrate with that remote branch.  There is a user
  16preference configuration variable "push.default" to change this.
  17
  18"git push $there tag v1.2.3" used to allow replacing a tag v1.2.3
  19that already exists in the repository $there, if the rewritten tag
  20you are pushing points at a commit that is a descendant of a commit
  21that the old tag v1.2.3 points at.  This was found to be error prone
  22and starting with this release, any attempt to update an existing
  23ref under refs/tags/ hierarchy will fail, without "--force".
  24
  25When "git add -u" and "git add -A", that does not specify what paths
  26to add on the command line, is run from inside a subdirectory, the
  27scope of the operation has always been limited to the subdirectory.
  28Many users found this counter-intuitive, given that "git commit -a"
  29and other commands operate on the entire tree regardless of where you
  30are. In this release, these commands give warning in such a case and
  31encourage the user to say "git add -u/-A ." instead when restricting
  32the scope to the current directory. At Git 2.0 (not *this* one), we
  33plan to change these commands without pathspec to operate on the
  34entire tree, and training your fingers to type "." will protect you
  35against the future change.
  36
  37
  38Updates since v1.8.1
  39--------------------
  40
  41UI, Workflows & Features
  42
  43 * Initial ports to QNX and z/OS UNIX System Services have started.
  44
  45 * Output from the tests is coloured using "green is okay, yellow is
  46   questionable, red is bad and blue is informative" scheme.
  47
  48 * Mention of "GIT/Git/git" in the documentation have been updated to
  49   be more uniform and consistent.  The name of the system and the
  50   concept it embodies is "Git"; the command the users type is "git".
  51   All-caps "GIT" was merely a way to imitate "Git" typeset in small
  52   caps in our ASCII text only documentation and to be avoided.
  53
  54 * The completion script (in contrib/completion) used to let the
  55   default completer to suggest pathnames, which gave too many
  56   irrelevant choices (e.g. "git add" would not want to add an
  57   unmodified path).  It learnt to use a more git-aware logic to
  58   enumerate only relevant ones.
  59
  60 * In bare repositories, "git shortlog" and other commands now read
  61   mailmap files from the tip of the history, to help running these
  62   tools in server settings.
  63
  64 * Color specifiers, e.g. "%C(blue)Hello%C(reset)", used in the
  65   "--format=" option of "git log" and friends can be disabled when
  66   the output is not sent to a terminal by prefixing them with
  67   "auto,", e.g. "%C(auto,blue)Hello%C(auto,reset)".
  68
  69 * Scripts can ask Git that wildcard patterns in pathspecs they give do
  70   not have any significance, i.e. take them as literal strings.
  71
  72 * The patterns in .gitignore and .gitattributes files can have **/,
  73   as a pattern that matches 0 or more levels of subdirectory.
  74   E.g. "foo/**/bar" matches "bar" in "foo" itself or in a
  75   subdirectory of "foo".
  76
  77 * When giving arguments without "--" disambiguation, object names
  78   that come earlier on the command line must not be interpretable as
  79   pathspecs and pathspecs that come later on the command line must
  80   not be interpretable as object names.  This disambiguation rule has
  81   been tweaked so that ":/" (no other string before or after) is
  82   always interpreted as a pathspec; "git cmd -- :/" is no longer
  83   needed, you can just say "git cmd :/".
  84
  85 * Various "hint" lines Git gives when it asks the user to edit
  86   messages in the editor are commented out with '#' by default. The
  87   core.commentchar configuration variable can be used to customize
  88   this '#' to a different character.
  89
  90 * "git add -u" and "git add -A" without pathspec issues warning to
  91   make users aware that they are only operating on paths inside the
  92   subdirectory they are in.  Use ":/" (everything from the top) or
  93   "." (everything from the $cwd) to disambiguate.
  94
  95 * "git blame" (and "git diff") learned the "--no-follow" option.
  96
  97 * "git branch" now rejects some nonsense combinations of command line
  98   arguments (e.g. giving more than one branch name to rename) with
  99   more case-specific error messages.
 100
 101 * "git check-ignore" command to help debugging .gitignore files has
 102   been added.
 103
 104 * "git cherry-pick" can be used to replay a root commit to an unborn
 105   branch.
 106
 107 * "git commit" can be told to use --cleanup=whitespace by setting the
 108   configuration variable commit.cleanup to 'whitespace'.
 109
 110 * "git diff" and other Porcelain commands can be told to use a
 111   non-standard algorithm by setting diff.algorithm configuration
 112   variable.
 113
 114 * "git fetch --mirror" and fetch that uses other forms of refspec
 115   with wildcard used to attempt to update a symbolic ref that match
 116   the wildcard on the receiving end, which made little sense (the
 117   real ref that is pointed at by the symbolic ref would be updated
 118   anyway).  Symbolic refs no longer are affected by such a fetch.
 119
 120 * "git format-patch" now detects more cases in which a whole branch
 121   is being exported, and uses the description for the branch, when
 122   asked to write a cover letter for the series.
 123
 124 * "git format-patch" learned "-v $count" option, and prepends a
 125   string "v$count-" to the names of its output files, and also
 126   automatically sets the subject prefix to "PATCH v$count". This
 127   allows patches from rerolled series to be stored under different
 128   names and makes it easier to reuse cover letter messages.
 129
 130 * "git log" and friends can be told with --use-mailmap option to
 131   rewrite the names and email addresses of people using the mailmap
 132   mechanism.
 133
 134 * "git log --cc --graph" now shows the combined diff output with the
 135   ancestry graph.
 136
 137 * "git log --grep=<pattern>" honors i18n.logoutputencoding to look
 138   for the pattern after fixing the log message to the specified
 139   encoding.
 140
 141 * "git mergetool" and "git difftool" learned to list the available
 142   tool backends in a more consistent manner.
 143
 144 * "git mergetool" is aware of TortoiseGitMerge now and uses it over
 145   TortoiseMerge when available.
 146
 147 * "git push" now requires "-f" to update a tag, even if it is a
 148   fast-forward, as tags are meant to be fixed points.
 149
 150 * Error messages from "git push" when it stops to prevent remote refs
 151   from getting overwritten by mistake have been improved to explain
 152   various situations separately.
 153
 154 * "git push" will stop without doing anything if the new "pre-push"
 155   hook exists and exits with a failure.
 156
 157 * When "git rebase" fails to generate patches to be applied (e.g. due
 158   to oom), it failed to detect the failure and instead behaved as if
 159   there were nothing to do.  A workaround to use a temporary file has
 160   been applied, but we probably would want to revisit this later, as
 161   it hurts the common case of not failing at all.
 162
 163 * Input and preconditions to "git reset" has been loosened where
 164   appropriate.  "git reset $fromtree Makefile" requires $fromtree to
 165   be any tree (it used to require it to be a commit), for example.
 166   "git reset" (without options or parameters) used to error out when
 167   you do not have any commits in your history, but it now gives you
 168   an empty index (to match non-existent commit you are not even on).
 169
 170 * "git status" says what branch is being bisected or rebased when
 171   able, not just "bisecting" or "rebasing".
 172
 173 * "git submodule" started learning a new mode to integrate with the
 174   tip of the remote branch (as opposed to integrating with the commit
 175   recorded in the superproject's gitlink).
 176
 177 * "git upload-pack" which implements the service "ls-remote" and
 178   "fetch" talk to can be told to hide ref hierarchies the server
 179   side internally uses (and that clients have no business learning
 180   about) with transfer.hiderefs configuration.
 181
 182
 183Foreign Interface
 184
 185 * "git fast-export" has been updated for its use in the context of
 186   the remote helper interface.
 187
 188 * A new remote helper to interact with bzr has been added to contrib/.
 189
 190 * "git p4" got various bugfixes around its branch handling.  It is
 191   also made usable with Python 2.4/2.5.  In addition, its various
 192   portability issues for Cygwin have been addressed.
 193
 194 * The remote helper to interact with Hg in contrib/ has seen a few
 195   fixes.
 196
 197
 198Performance, Internal Implementation, etc.
 199
 200 * "git fsck" has been taught to be pickier about entries in tree
 201   objects that should not be there, e.g. ".", ".git", and "..".
 202
 203 * Matching paths with common forms of pathspecs that contain wildcard
 204   characters has been optimized further.
 205
 206 * We stopped paying attention to $GIT_CONFIG environment that points
 207   at a single configuration file from any command other than "git config"
 208   quite a while ago, but "git clone" internally set, exported, and
 209   then unexported the variable during its operation unnecessarily.
 210
 211 * "git reset" internals has been reworked and should be faster in
 212   general. We tried to be careful not to break any behaviour but
 213   there could be corner cases, especially when running the command
 214   from a conflicted state, that we may have missed.
 215
 216 * The implementation of "imap-send" has been updated to reuse xml
 217   quoting code from http-push codepath, and lost a lot of unused
 218   code.
 219
 220 * There is a simple-minded checker for the test scripts in t/
 221   directory to catch most common mistakes (it is not enabled by
 222   default).
 223
 224 * You can build with USE_WILDMATCH=YesPlease to use a replacement
 225   implementation of pattern matching logic used for pathname-like
 226   things, e.g. refnames and paths in the repository.  This new
 227   implementation is not expected change the existing behaviour of Git
 228   in this release, except for "git for-each-ref" where you can now
 229   say "refs/**/master" and match with both refs/heads/master and
 230   refs/remotes/origin/master.  We plan to use this new implementation
 231   in wider places (e.g. "git ls-files '**/Makefile' may find Makefile
 232   at the top-level, and "git log '**/t*.sh'" may find commits that
 233   touch a shell script whose name begins with "t" at any level) in
 234   future versions of Git, but we are not there yet.  By building with
 235   USE_WILDMATCH, using the resulting Git daily and reporting when you
 236   find breakages, you can help us get closer to that goal.
 237
 238 * Some reimplementations of Git do not write all the stat info back
 239   to the index due to their implementation limitations (e.g. jgit).
 240   A configuration option can tell Git to ignore changes to most of
 241   the stat fields and only pay attention to mtime and size, which
 242   these implementations can reliably update.  This can be used to
 243   avoid excessive revalidation of contents.
 244
 245 * Some platforms ship with old version of expat where xmlparse.h
 246   needs to be included instead of expat.h; the build procedure has
 247   been taught about this.
 248
 249 * "make clean" on platforms that cannot compute header dependencies
 250   on the fly did not work with implementations of "rm" that do not
 251   like an empty argument list.
 252
 253Also contains minor documentation updates and code clean-ups.
 254
 255
 256Fixes since v1.8.1
 257------------------
 258
 259Unless otherwise noted, all the fixes since v1.8.1 in the maintenance
 260track are contained in this release (see release notes to them for
 261details).
 262
 263 * An element on GIT_CEILING_DIRECTORIES list that does not name the
 264   real path to a directory (i.e. a symbolic link) could have caused
 265   the GIT_DIR discovery logic to escape the ceiling.
 266
 267 * When attempting to read the XDG-style $HOME/.config/git/config and
 268   finding that $HOME/.config/git is a file, we gave a wrong error
 269   message, instead of treating the case as "a custom config file does
 270   not exist there" and moving on.
 271
 272 * The behaviour visible to the end users was confusing, when they
 273   attempt to kill a process spawned in the editor that was in turn
 274   launched by Git with SIGINT (or SIGQUIT), as Git would catch that
 275   signal and die.  We ignore these signals now.
 276   (merge 0398fc34 pf/editor-ignore-sigint later to maint).
 277
 278 * A child process that was killed by a signal (e.g. SIGINT) was
 279   reported in an inconsistent way depending on how the process was
 280   spawned by us, with or without a shell in between.
 281
 282 * After failing to create a temporary file using mkstemp(), failing
 283   pathname was not reported correctly on some platforms.
 284
 285 * We used to stuff "user@" and then append what we read from
 286   /etc/mailname to come up with a default e-mail ident, but a bug
 287   lost the "user@" part.
 288
 289 * The attribute mechanism didn't allow limiting attributes to be
 290   applied to only a single directory itself with "path/" like the
 291   exclude mechanism does.  The initial implementation of this that
 292   was merged to 'maint' and 1.8.1.2 was with a severe performance
 293   degradations and needs to merge a fix-up topic.
 294
 295 * The smart HTTP clients forgot to verify the content-type that comes
 296   back from the server side to make sure that the request is being
 297   handled properly.
 298
 299 * "git am" did not parse datestamp correctly from Hg generated patch,
 300   when it is run in a locale outside C (or en).
 301
 302 * "git apply" misbehaved when fixing whitespace breakages by removing
 303   excess trailing blank lines.
 304
 305 * "git apply --summary" has been taught to make sure the similarity
 306   value shown in its output is sensible, even when the input had a
 307   bogus value.
 308
 309 * A tar archive created by "git archive" recorded a directory in a
 310   way that made NetBSD's implementation of "tar" sometimes unhappy.
 311
 312 * "git archive" did not record uncompressed size in the header when
 313   streaming a zip archive, which confused some implementations of unzip.
 314
 315 * "git archive" did not parse configuration values in tar.* namespace
 316   correctly.
 317   (merge b3873c3 jk/config-parsing-cleanup later to maint).
 318
 319 * Attempt to "branch --edit-description" an existing branch, while
 320   being on a detached HEAD, errored out.
 321
 322 * "git clean" showed what it was going to do, but sometimes end up
 323   finding that it was not allowed to do so, which resulted in a
 324   confusing output (e.g. after saying that it will remove an
 325   untracked directory, it found an embedded git repository there
 326   which it is not allowed to remove).  It now performs the actions
 327   and then reports the outcome more faithfully.
 328
 329 * When "git clone --separate-git-dir=$over_there" is interrupted, it
 330   failed to remove the real location of the $GIT_DIR it created.
 331   This was most visible when interrupting a submodule update.
 332
 333 * "git cvsimport" mishandled timestamps at DST boundary.
 334
 335 * We used to have an arbitrary 32 limit for combined diff input,
 336   resulting in incorrect number of leading colons shown when showing
 337   the "--raw --cc" output.
 338
 339 * "git fetch --depth" was broken in at least three ways.  The
 340   resulting history was deeper than specified by one commit, it was
 341   unclear how to wipe the shallowness of the repository with the
 342   command, and documentation was misleading.
 343   (merge cfb70e1 nd/fetch-depth-is-broken later to maint).
 344
 345 * "git log --all -p" that walked refs/notes/textconv/ ref can later
 346   try to use the textconv data incorrectly after it gets freed.
 347
 348 * We forgot to close the file descriptor reading from "gpg" output,
 349   killing "git log --show-signature" on a long history.
 350
 351 * The way "git svn" asked for password using SSH_ASKPASS and
 352   GIT_ASKPASS was not in line with the rest of the system.
 353
 354 * The --graph code fell into infinite loop when asked to do what the
 355   code did not expect.
 356
 357 * http transport was wrong to ask for the username when the
 358   authentication is done by certificate identity.
 359
 360 * "git pack-refs" that ran in parallel to another process that
 361   created new refs had a nasty race.
 362
 363 * Rebasing the history of superproject with change in the submodule
 364   has been broken since v1.7.12.
 365
 366 * After "git add -N" and then writing a tree object out of the
 367   index, the cache-tree data structure got corrupted.
 368
 369 * "git clone" used to allow --bare and --separate-git-dir=$there
 370   options at the same time, which was nonsensical.
 371
 372 * "git rebase --preserve-merges" lost empty merges in recent versions
 373   of Git.
 374
 375 * "git merge --no-edit" computed who were involved in the work done
 376   on the side branch, even though that information is to be discarded
 377   without getting seen in the editor.
 378
 379 * "git merge" started calling prepare-commit-msg hook like "git
 380   commit" does some time ago, but forgot to pay attention to the exit
 381   status of the hook.
 382
 383 * A failure to push due to non-ff while on an unborn branch
 384   dereferenced a NULL pointer when showing an error message.
 385
 386 * When users spell "cc:" in lowercase in the fake "header" in the
 387   trailer part, "git send-email" failed to pick up the addresses from
 388   there. As e-mail headers field names are case insensitive, this
 389   script should follow suit and treat "cc:" and "Cc:" the same way.
 390
 391 * Output from "git status --ignored" showed an unexpected interaction
 392   with "--untracked".
 393
 394 * "gitweb", when sorting by age to show repositories with new
 395   activities first, used to sort repositories with absolutely
 396   nothing in it early, which was not very useful.
 397
 398 * "gitweb"'s code to sanitize control characters before passing it to
 399   "highlight" filter lost known-to-be-safe control characters by
 400   mistake.
 401
 402 * "gitweb" pages served over HTTPS, when configured to show picon or
 403   gravatar, referred to these external resources to be fetched via
 404   HTTP, resulting in mixed contents warning in browsers.
 405
 406 * When a line to be wrapped has a solid run of non space characters
 407   whose length exactly is the wrap width, "git shortlog -w" failed
 408   to add a newline after such a line.
 409
 410 * Command line completion leaked an unnecessary error message while
 411   looking for possible matches with paths in <tree-ish>.
 412
 413 * Command line completion for "tcsh" emitted an unwanted space
 414   after completing a single directory name.
 415
 416 * Command line completion code was inadvertently made incompatible with
 417   older versions of bash by using a newer array notation.
 418
 419 * "git push" was taught to refuse updating the branch that is
 420   currently checked out long time ago, but the user manual was left
 421   stale.
 422   (merge 50995ed wk/man-deny-current-branch-is-default-these-days later to maint).
 423
 424 * Some shells do not behave correctly when IFS is unset; work it
 425   around by explicitly setting it to the default value.
 426
 427 * Some scripted programs written in Python did not get updated when
 428   PYTHON_PATH changed.
 429   (cherry-pick 96a4647fca54031974cd6ad1 later to maint).
 430
 431 * When autoconf is used, any build on a different commit always ran
 432   "config.status --recheck" even when unnecessary.
 433
 434 * A fix was added to the build procedure to work around buggy
 435   versions of ccache broke the auto-generation of dependencies, which
 436   unfortunately is still relevant because some people use ancient
 437   distros.
 438
 439 * The autoconf subsystem passed --mandir down to generated
 440   config.mak.autogen but forgot to do the same for --htmldir.
 441   (merge 55d9bf0 ct/autoconf-htmldir later to maint).
 442
 443 * A change made on v1.8.1.x maintenance track had a nasty regression
 444   to break the build when autoconf is used.
 445   (merge 7f1b697 jn/less-reconfigure later to maint).
 446
 447 * We have been carrying a translated and long-unmaintained copy of an
 448   old version of the tutorial; removed.
 449
 450 * t0050 had tests expecting failures from a bug that was fixed some
 451   time ago.
 452
 453 * t4014, t9502 and t0200 tests had various portability issues that
 454   broke on OpenBSD.
 455
 456 * t9020 and t3600 tests had various portability issues.
 457
 458 * t9200 runs "cvs init" on a directory that already exists, but a
 459   platform can configure this fail for the current user (e.g. you
 460   need to be in the cvsadmin group on NetBSD 6.0).
 461
 462 * t9020 and t9810 had a few non-portable shell script construct.
 463
 464 * Scripts to test bash completion was inherently flaky as it was
 465   affected by whatever random things the user may have on $PATH.