Documentation / RelNotes / 1.8.4.txton commit Merge branch 'jk/in-pack-size-measurement' (802f878)
   1Git v1.8.4 Release Notes
   2========================
   3
   4Backward compatibility notes (for Git 2.0)
   5------------------------------------------
   6
   7When "git push [$there]" does not say what to push, we have used the
   8traditional "matching" semantics so far (all your branches were sent
   9to the remote as long as there already are branches of the same name
  10over there).  In Git 2.0, the default will change to the "simple"
  11semantics that pushes:
  12
  13 - only the current branch to the branch with the same name, and only
  14   when the current branch is set to integrate with that remote
  15   branch, if you are pushing to the same remote as you fetch from; or
  16
  17 - only the current branch to the branch with the same name, if you
  18   are pushing to a remote that is not where you usually fetch from.
  19
  20Use the user preference configuration variable "push.default" to
  21change this.  If you are an old-timer who is used to the "matching"
  22semantics, you can set the variable to "matching" to keep the
  23traditional behaviour.  If you want to live in the future early, you
  24can set it to "simple" today without waiting for Git 2.0.
  25
  26When "git add -u" (and "git add -A") is run inside a subdirectory and
  27does not specify which paths to add on the command line, it
  28will operate on the entire tree in Git 2.0 for consistency
  29with "git commit -a" and other commands.  There will be no
  30mechanism to make plain "git add -u" behave like "git add -u .".
  31Current users of "git add -u" (without a pathspec) should start
  32training their fingers to explicitly say "git add -u ."
  33before Git 2.0 comes.  A warning is issued when these commands are
  34run without a pathspec and when you have local changes outside the
  35current directory, because the behaviour in Git 2.0 will be different
  36from today's version in such a situation.
  37
  38In Git 2.0, "git add <path>" will behave as "git add -A <path>", so
  39that "git add dir/" will notice paths you removed from the directory
  40and record the removal.  Versions before Git 2.0, including this
  41release, will keep ignoring removals, but the users who rely on this
  42behaviour are encouraged to start using "git add --ignore-removal <path>"
  43now before 2.0 is released.
  44
  45
  46Updates since v1.8.3
  47--------------------
  48
  49Foreign interfaces, subsystems and ports.
  50
  51 * "git rebase -i" now honors --strategy and -X options.
  52
  53 * Git-gui has been updated to its 0.18.0 version.
  54
  55 * MediaWiki remote helper (in contrib/) has been updated to use the
  56   credential helper interface from Git.pm.
  57
  58 * Update build for Cygwin 1.[57].  Torsten Bögershausen reports that
  59   this is fine with Cygwin 1.7 ($gmane/225824) so let's try moving it
  60   ahead.
  61
  62 * The credential helper to talk to keychain on OS X (in contrib/) has
  63   been updated to kick in not just when talking http/https but also
  64   imap(s) and smtp.
  65
  66 * Remote transport helper has been updated to report errors and
  67   maintain ref hierarchy used to keep track of its own state better.
  68
  69 * With "export" remote-helper protocol, (1) a push that tries to
  70   update a remote ref whose name is different from the pushing side
  71   does not work yet, and (2) the helper may not know how to do
  72   --dry-run; these problematic cases are disabled for now.
  73
  74 * git-remote-hg/bzr (in contrib/) updates.
  75
  76 * git-remote-mw (in contrib/) hints users to check the certificate,
  77   when https:// connection failed.
  78
  79
  80UI, Workflows & Features
  81
  82 * In addition to the choice from "rebase, merge, or checkout-detach",
  83   "submodule update" can allow a custom command to be used in to
  84   update the working tree of submodules via the "submodule.*.update"
  85   configuration variable.
  86
  87 * "git submodule update" can optionally clone the submodule
  88   repositories shallowly.
  89
  90 * "git format-patch" learned "--from[=whom]" option, which sets the
  91   "From: " header to the specified person (or the person who runs the
  92   command, if "=whom" part is missing) and move the original author
  93   information to an in-body From: header as necessary.
  94
  95 * The configuration variable "merge.ff" was cleary a tri-state to
  96   choose one from "favor fast-forward when possible", "always create
  97   a merge even when the history could fast-forward" and "do not
  98   create any merge, only update when the history fast-forwards", but
  99   the command line parser did not implement the usual convention of
 100   "last one wins, and command line overrides the configuration"
 101   correctly.
 102
 103 * "gitweb" learned to optionally place extra links that point at the
 104   levels higher than the Gitweb pages themselves in the breadcrumbs,
 105   so that it can be used as part of a larger installation.
 106
 107 * "git log --format=" now honors i18n.logoutputencoding configuration
 108   variable.
 109
 110 * The "push.default=simple" mode of "git push" has been updated to
 111   behave like "current" without requiring a remote tracking
 112   information, when you push to a remote that is different from where
 113   you fetch from (i.e. a triangular workflow).
 114
 115 * Having multiple "fixup!" on a line in the rebase instruction sheet
 116   did not work very well with "git rebase -i --autosquash".
 117
 118 * "git log" learned the "--author-date-order" option, with which the
 119   output is topologically sorted and commits in parallel histories
 120   are shown intermixed together based on the author timestamp.
 121
 122 * Various subcommands of "git submodule" refused to run from anywhere
 123   other than the top of the working tree of the superproject, but
 124   they have been taught to let you run from a subdirectory.
 125
 126 * "git diff" learned a mode that ignores hunks whose change consists
 127   only of additions and removals of blank lines, which is the same as
 128   "diff -B" (ignore blank lines) of GNU diff.
 129
 130 * "git rm" gives a single message followed by list of paths to report
 131   multiple paths that cannot be removed.
 132
 133 * "git rebase" can be told with ":/look for this string" syntax commits
 134   to replay the changes onto and where the work to be replayed begins.
 135
 136 * Many tutorials teach users to set "color.ui" to "auto" as the first
 137   thing after you set "user.name/email" to introduce yourselves to
 138   Git.  Now the variable defaults to "auto".
 139
 140 * On Cygwin, "cygstart" is now recognised as a possible way to start
 141   a web browser (used in "help -w" and "instaweb" among others).
 142
 143 * "git status" learned status.branch and status.short configuration
 144   variables to use --branch and --short options by default (override
 145   with --no-branch and --no-short options from the command line).
 146
 147 * "git cmd <name>", when <name> happens to be a 40-hex string,
 148   directly uses the 40-hex string as an object name, even if a ref
 149   "refs/<some hierarchy>/<name>" exists.  This disambiguation order
 150   is unlikely to change, but we should warn about the ambiguity just
 151   like we warn when more than one refs/ hierachies share the same
 152   name.
 153
 154 * "git rebase" learned "--[no-]autostash" option to save local
 155   changes instead of refusing to run (to which people's normal
 156   response was to stash them and re-run).  This introduced a corner
 157   case breakage to "git am --abort" but it has been fixed.
 158
 159 * Instead of typing four capital letters "HEAD", you can say "@" now,
 160   e.g. "git log @".
 161
 162 * "check-ignore" (new feature since 1.8.2) has been updated to work
 163   more like "check-attr" over bidi-pipes.
 164
 165 * "git describe" learned "--first-parent" option to limit its closest
 166   tagged commit search to the first-parent chain.
 167
 168 * "git merge foo" that might have meant "git merge origin/foo" is
 169   diagnosed with a more informative error message.
 170
 171 * "git log -L<line>,<range>:<filename>" has been added.  This may
 172   still have leaks and rough edges, though.
 173
 174 * We used the approxidate() parser for "--expire=<timestamp>" options
 175   of various commands, but it is better to treat --expire=all and
 176   --expire=now a bit more specially than using the current timestamp.
 177   "git gc" and "git reflog" have been updated with a new parsing
 178   function for expiry dates.
 179
 180 * Updates to completion (both bash and zsh) helpers.
 181
 182 * The behaviour of the "--chain-reply-to" option of "git send-email"
 183   have changed at 1.7.0, and we added a warning/advice message to
 184   help users adjust to the new behaviour back then, but we kept it
 185   around for too long.  The message has finally been removed.
 186
 187 * "git fetch origin master" unlike "git fetch origin" or "git fetch"
 188   did not update "refs/remotes/origin/master"; this was an early
 189   design decision to keep the update of remote tracking branches
 190   predictable, but in practice it turns out that people find it more
 191   convenient to opportunistically update them whenever we have a
 192   chance, and we have been updating them when we run "git push" which
 193   already breaks the original "predictability" anyway.
 194
 195 * The configuration variable core.checkstat was advertised in the
 196   documentation but the code expected core.statinfo instead.
 197   For now, we accept both core.checkstat and core.statinfo, but the
 198   latter will be removed in the longer term.
 199
 200
 201Performance, Internal Implementation, etc.
 202
 203 * Fetching between repositories with many refs employed O(n^2)
 204   algorithm to match up the common objects, which has been corrected.
 205
 206 * The original way to specify remote repository using .git/branches/
 207   used to have a nifty feature.  The code to support the feature was
 208   still in a function but the caller was changed not to call it 5
 209   years ago, breaking that feature and leaving the supporting code
 210   unreachable.  The dead code has been removed.
 211
 212 * "git pack-refs" that races with new ref creation or deletion have
 213   been susceptible to lossage of refs under right conditions, which
 214   has been tightened up.
 215
 216 * We read loose and packed rerferences in two steps, but after
 217   deciding to read a loose ref but before actually opening it to read
 218   it, another process racing with us can unlink it, which would cause
 219   us to barf.  The codepath has been updated to retry when such a
 220   race is detected, instead of outright failing.
 221
 222 * Uses of the platform fnmatch(3) function (many places in the code,
 223   matching pathspec, .gitignore and .gitattributes to name a few)
 224   have been replaced with wildmatch, allowing "foo/**/bar" that would
 225   match foo/bar, foo/a/bar, foo/a/b/bar, etc.
 226
 227 * Memory ownership and lifetime rules for what for-each-ref feeds to
 228   its callbacks have been clarified (in short, "you do not own it, so
 229   make a copy if you want to keep it").
 230
 231 * The revision traversal logic to improve culling of irrelevant
 232   parents while traversing a mergy history has been updated.
 233
 234 * Some leaks in unpack-trees (used in merge, cherry-pick and other
 235   codepaths) have been plugged.
 236
 237 * The codepath to read from marks files in fast-import/export did not
 238   have to accept anything but 40-hex representation of the object
 239   name.  Further, fast-export did not need full in-core object
 240   representation to have parsed wen reading from them.  These
 241   codepaths have been optimized by taking advantage of these access
 242   patterns.
 243
 244 * Object lookup logic, when the object hashtable starts to become
 245   crowded, has been optimized.
 246
 247 * When TEST_OUTPUT_DIRECTORY setting is used, it was handled somewhat
 248   inconsistently between the test framework and t/Makefile, and logic
 249   to summarize the results looked at a wrong place.
 250
 251 * "git clone" uses a lighter-weight implementation when making sure
 252   that the history behind refs are complete.
 253
 254 * Many warnings from sparse source checker in compat/ area has been
 255   squelched.
 256
 257 * The code to reading and updating packed-refs file has been updated,
 258   correcting corner case bugs.
 259
 260
 261Also contains various documentation updates and code clean-ups.
 262
 263
 264Fixes since v1.8.3
 265------------------
 266
 267Unless otherwise noted, all the fixes since v1.8.3 in the maintenance
 268track are contained in this release (see release notes to them for
 269details).
 270
 271 * Pass port number as a separate argument when "send-email" initializes
 272   Net::SMTP, instead of as a part of the hostname, i.e. host:port.
 273   This allows GSSAPI codepath to match with the hostname given.
 274   (merge 1a741bf bc/send-email-use-port-as-separate-param later to maint).
 275
 276 * "git diff" refused to even show difference when core.safecrlf is
 277   set to true (i.e. error out) and there are offending lines in the
 278   working tree files.
 279   (merge 5430bb2 jc/maint-diff-core-safecrlf later to maint).
 280
 281 * A test that should have failed but didn't revealed a bug that needs
 282   to be corrected.
 283   (merge 94d75d1 jc/t1512-fix later to maint).
 284
 285 * An overlong path to a .git directory may have overflown the
 286   temporary path buffer used to create a name for lockfiles.
 287   (merge 2fbd4f9 mh/maint-lockfile-overflow later to maint).
 288
 289 * Invocations of "git checkout" used internally by "git rebase" were
 290   counted as "checkout", and affected later "git checkout -" to the
 291   the user to an unexpected place.
 292   (merge 3bed291 rr/rebase-checkout-reflog later to maint).
 293
 294 * "git stash save", when your local change turns a tracked file into
 295   a directory, has to remove files in that directory in order to
 296   revert your working tree to a pristine state.  This will lose
 297   untracked files in such a directory, and the command now requires
 298   you to "--force" it.
 299
 300 * The configuration variable column.ui was poorly documented.
 301   (merge 5e62cc1 rr/column-doc later to maint).
 302
 303 * "git name-rev --refs=tags/v*" were forbidden, which was a bit
 304   inconvenient (you had to give a pattern to match refs fully, like
 305   --refs=refs/tags/v*).
 306   (merge 98c5c4a nk/name-rev-abbreviated-refs later to maint).
 307
 308 * "git apply" parsed patches that add new files, generated by
 309   programs other than Git, incorrectly.  This is an old breakage in
 310   v1.7.11 and will need to be merged down to the maintanance tracks.
 311
 312 * Older cURL wanted piece of memory we call it with to be stable, but
 313   we updated the auth material after handing it to a call.
 314
 315 * "git pull" into nothing trashed "local changes" that were in the
 316   index, and this avoids it.
 317
 318 * Many "git submodule" operations do not work on a submodule at a
 319   path whose name is not in ASCII.
 320
 321 * "cherry-pick" had a small leak in an error codepath.
 322
 323 * Logic used by git-send-email to suppress cc mishandled names like
 324   "A U. Thor" <author@example.xz>, where the human readable part
 325   needs to be quoted (the user input may not have the double quotes
 326   around the name, and comparison was done between quoted and
 327   unquoted strings).  It also mishandled names that need RFC2047
 328   quoting.
 329
 330 * Call to discard_cache/discard_index (used when we use different
 331   contents of the index in-core, in many operations like commit,
 332   apply, and merge) used to leak memory that held the array of index
 333   entries, which has been plugged.
 334   (merge a0fc4db rs/discard-index-discard-array later to maint).
 335
 336 * "gitweb" forgot to clear a global variable $search_regexp upon each
 337   request, mistakenly carrying over the previous search to a new one
 338   when used as a persistent CGI.
 339
 340 * The wildmatch engine did not honor WM_CASEFOLD option correctly.
 341
 342 * "git log -c --follow $path" segfaulted upon hitting the commit that
 343   renamed the $path being followed.
 344
 345 * When a reflog notation is used for implicit "current branch", we
 346   did not say which branch and worse said "branch ''".
 347
 348 * "difftool --dir-diff" did not copy back changes made by the
 349   end-user in the diff tool backend to the working tree in some
 350   cases.
 351
 352 * "git push $there HEAD:branch" did not resolve HEAD early enough, so
 353   it was easy to flip it around while push is still going on and push
 354   out a branch that the user did not originally intended when the
 355   command was started.
 356
 357 * The bash prompt code (in contrib/) displayed the name of the branch
 358   being rebased when "rebase -i/-m/-p" modes are in use, but not the
 359   plain vanilla "rebase".
 360
 361 * Handling of negative exclude pattern for directories "!dir" was
 362   broken in the update to v1.8.3.
 363
 364 * zsh prompt script that borrowed from bash prompt script did not
 365   work due to slight differences in array variable notation between
 366   these two shells.
 367
 368 * An entry for "file://" scheme in the enumeration of URL types Git
 369   can take in the HTML documentation was made into a clickable link
 370   by mistake.
 371
 372 * "git push --[no-]verify" was not documented.
 373
 374 * Stop installing the git-remote-testpy script that is only used for
 375   testing.
 376
 377 * "git commit --allow-empty-message -m ''" should not start an
 378   editor.
 379
 380 * "git merge @{-1}~22" was rewritten to "git merge frotz@{1}~22"
 381   incorrectly when your previous branch was "frotz" (it should be
 382   rewritten to "git merge frotz~22" instead).
 383
 384 * "git diff -c -p" was not showing a deleted line from a hunk when
 385   another hunk immediately begins where the earlier one ends.
 386
 387 * "git log --ancestry-path A...B" did not work as expected, as it did
 388   not pay attention to the fact that the merge base between A and B
 389   was the bottom of the range being specified.
 390
 391 * Mac OS X does not like to write(2) more than INT_MAX number of
 392   bytes; work it around by chopping write(2) into smaller pieces.
 393
 394 * Newer MacOS X encourages the programs to compile and link with
 395   their CommonCrypto, not with OpenSSL.
 396
 397 * "git clone foo/bar:baz" cannot be a request to clone from a remote
 398   over git-over-ssh specified in the scp style.  This case is now
 399   detected and clones from a local repository at "foo/bar:baz".
 400
 401 * When $HOME is misconfigured to point at an unreadable directory, we
 402   used to complain and die. Loosen the check.
 403
 404 * "git subtree" (in contrib/) had one codepath with loose error
 405   checks to lose data at the remote side.
 406
 407 * "git fetch" into a shallow repository from a repository that does
 408   not know about the shallow boundary commits (e.g. a different fork
 409   from the repository the current shallow repository was cloned from)
 410   did not work correctly.
 411
 412 * "git checkout foo" DWIMs the intended "upstream" and turns it into
 413   "git checkout -t -b foo remotes/origin/foo". This codepath has been
 414   updated to correctly take existing remote definitions into account.