Documentation / RelNotes / 2.14.0.txton commit Sync with maint (117ddef)
   1Git 2.14 Release Notes
   2======================
   3
   4Backward compatibility notes.
   5
   6 * Use of an empty string as a pathspec element that is used for
   7   'everything matches' is still warned and Git asks users to use a
   8   more explicit '.' for that instead.  The hope is that existing
   9   users will not mind this change, and eventually the warning can be
  10   turned into a hard error, upgrading the deprecation into removal of
  11   this (mis)feature.  That is not scheduled to happen in the upcoming
  12   release (yet).
  13
  14 * Git now avoids blindly falling back to ".git" when the setup
  15   sequence said we are _not_ in Git repository.  A corner case that
  16   happens to work right now may be broken by a call to die("BUG").
  17   We've tried hard to locate such cases and fixed them, but there
  18   might still be cases that need to be addressed--bug reports are
  19   greatly appreciated.
  20
  21 * The experiment to improve the hunk-boundary selection of textual
  22   diff output has finished, and the "indent heuristics" has now
  23   become the default.
  24
  25
  26Updates since v2.13
  27-------------------
  28
  29UI, Workflows & Features
  30
  31 * The colors in which "git status --short --branch" showed the names
  32   of the current branch and its remote-tracking branch are now
  33   configurable.
  34
  35 * "git clone" learned the "--no-tags" option not to fetch all tags
  36   initially, and also set up the tagopt not to follow any tags in
  37   subsequent fetches.
  38
  39 * "git archive --format=zip" learned to use zip64 extension when
  40   necessary to go beyond the 4GB limit.
  41
  42 * "git reset" learned "--recurse-submodules" option.
  43
  44 * "git diff --submodule=diff" now recurses into nested submodules.
  45
  46 * "git repack" learned to accept the --threads=<n> option and pass it
  47   to pack-objects.
  48
  49 * "git send-email" learned to run sendemail-validate hook to inspect
  50   and reject a message before sending it out.
  51
  52 * There is no good reason why "git fetch $there $sha1" should fail
  53   when the $sha1 names an object at the tip of an advertised ref,
  54   even when the other side hasn't enabled allowTipSHA1InWant.
  55
  56 * The recently introduced "[includeIf "gitdir:$dir"] path=..."
  57   mechanism has further been taught to take symlinks into account.
  58   The directory "$dir" specified in "gitdir:$dir" may be a symlink to
  59   a real location, not something that $(getcwd) may return.  In such
  60   a case, a realpath of "$dir" is compared with the real path of the
  61   current repository to determine if the contents from the named path
  62   should be included.
  63
  64 * Make the "indent" heuristics the default in "diff" and diff.indentHeuristics
  65   configuration variable an escape hatch for those who do no want it.
  66
  67 * Many commands learned to pay attention to submodule.recurse
  68   configuration.
  69
  70 * The convention for a command line is to follow "git cmdname
  71   --options" with revisions followed by an optional "--"
  72   disambiguator and then finally pathspecs.  When "--" is not there,
  73   we make sure early ones are all interpretable as revs (and do not
  74   look like paths) and later ones are the other way around.  A
  75   pathspec with "magic" (e.g. ":/p/a/t/h" that matches p/a/t/h from
  76   the top-level of the working tree, no matter what subdirectory you
  77   are working from) are conservatively judged as "not a path", which
  78   required disambiguation more often.  The command line parser
  79   learned to say "it's a pathspec" a bit more often when the syntax
  80   looks like so.
  81
  82 * Update "perl-compatible regular expression" support to enable JIT
  83   and also allow linking with the newer PCRE v2 library.
  84
  85 * "filter-branch" learned a pseudo filter "--setup" that can be used
  86   to define common functions/variables that can be used by other
  87   filters.
  88
  89 * Using "git add d/i/r" when d/i/r is the top of the working tree of
  90   a separate repository would create a gitlink in the index, which
  91   would appear as a not-quite-initialized submodule to others.  We
  92   learned to give warnings when this happens.
  93
  94 * "git status" learned to optionally give how many stash entries the
  95   user has in its output.
  96
  97 * "git status" has long shown essentially the same message as "git
  98   commit"; the message it gives while preparing for the root commit,
  99   i.e. "Initial commit", was hard to understand for some new users.
 100   Now it says "No commits yet" to stress more on the current status
 101   (rather than the commit the user is preparing for, which is more in
 102   line with the focus of "git commit").
 103
 104 * "git send-email" learned to overcome some SMTP server limitation
 105   that does not allow many pieces of e-mails to be sent over a single
 106   session.
 107
 108 * An old message shown in the commit log template was removed, as it
 109   has outlived its usefulness.
 110
 111
 112Performance, Internal Implementation, Development Support etc.
 113
 114 * The default packed-git limit value has been raised on larger
 115   platforms to save "git fetch" from a (recoverable) failure while
 116   "gc" is running in parallel.
 117
 118 * Code to update the cache-tree has been tightened so that we won't
 119   accidentally write out any 0{40} entry in the tree object.
 120
 121 * Attempt to allow us notice "fishy" situation where we fail to
 122   remove the temporary directory used during the test.
 123
 124 * Travis CI gained a task to format the documentation with both
 125   AsciiDoc and AsciiDoctor.
 126
 127 * Some platforms have ulong that is smaller than time_t, and our
 128   historical use of ulong for timestamp would mean they cannot
 129   represent some timestamp that the platform allows.  Invent a
 130   separate and dedicated timestamp_t (so that we can distingiuish
 131   timestamps and a vanilla ulongs, which along is already a good
 132   move), and then declare uintmax_t is the type to be used as the
 133   timestamp_t.
 134
 135 * We can trigger Windows auto-build tester (credits: Dscho &
 136   Microsoft) from our existing Travis CI tester now.
 137
 138 * Conversion from uchar[20] to struct object_id continues.
 139
 140 * Simplify parse_pathspec() codepath and stop it from looking at the
 141   default in-core index.
 142
 143 * Add perf-test for wildmatch.
 144
 145 * Code from "conversion using external process" codepath has been
 146   extracted to a separate sub-process.[ch] module.
 147
 148 * When "git checkout", "git merge", etc. manipulates the in-core
 149   index, various pieces of information in the index extensions are
 150   discarded from the original state, as it is usually not the case
 151   that they are kept up-to-date and in-sync with the operation on the
 152   main index.  The untracked cache extension is copied across these
 153   operations now, which would speed up "git status" (as long as the
 154   cache is properly invalidated).
 155
 156 * The internal implementation of "git grep" has seen some clean-up.
 157
 158 * Update the C style recommendation for notes for translators, as
 159   recent versions of gettext tools can work with our style of
 160   multi-line comments.
 161
 162 * The implementation of "ref" API around the "packed refs" have been
 163   cleaned up, in preparation for further changes.
 164
 165 * The internal logic used in "git blame" has been libified to make it
 166   easier to use by cgit.
 167
 168 * Our code often opens a path to an optional file, to work on its
 169   contents when we can successfully open it.  We can ignore a failure
 170   to open if such an optional file does not exist, but we do want to
 171   report a failure in opening for other reasons (e.g. we got an I/O
 172   error, or the file is there, but we lack the permission to open).
 173
 174   The exact errors we need to ignore are ENOENT (obviously) and
 175   ENOTDIR (less obvious).  Instead of repeating comparison of errno
 176   with these two constants, introduce a helper function to do so.
 177
 178 * We often try to open a file for reading whose existence is
 179   optional, and silently ignore errors from open/fopen; report such
 180   errors if they are not due to missing files.
 181
 182 * When an existing repository is used for t/perf testing, we first
 183   create bit-for-bit copy of it, which may grab a transient state of
 184   the repository and freeze it into the repository used for testing,
 185   which then may cause Git operations to fail.  Single out "the index
 186   being locked" case and forcibly drop the lock from the copy.
 187
 188 * Three instances of the same helper function have been consolidated
 189   to one.
 190
 191 * "fast-import" uses a default pack chain depth that is consistent
 192   with other parts of the system.
 193
 194 * A new test to show the interaction between the pattern [^a-z]
 195   (which matches '/') and a slash in a path has been added.  The
 196   pattern should not match the slash with "pathmatch", but should
 197   with "wildmatch".
 198
 199 * The 'diff-highlight' program (in contrib/) has been restructured
 200   for easier reuse by an external project 'diff-so-fancy'.
 201
 202 * A common pattern to free a piece of memory and assign NULL to the
 203   pointer that used to point at it has been replaced with a new
 204   FREE_AND_NULL() macro.
 205
 206 * Traditionally, the default die() routine had a code to prevent it
 207   from getting called multiple times, which interacted badly when a
 208   threaded program used it (one downside is that the real error may
 209   be hidden and instead the only error message given to the user may
 210   end up being "die recursion detected", which is not very useful).
 211
 212 * Introduce a "repository" object to eventually make it easier to
 213   work in multiple repositories (the primary focus is to work with
 214   the superproject and its submodules) in a single process.
 215
 216 * Optimize "what are the object names already taken in an alternate
 217   object database?" query that is used to derive the length of prefix
 218   an object name is uniquely abbreviated to.
 219
 220
 221Also contains various documentation updates and code clean-ups.
 222
 223
 224Fixes since v2.13
 225-----------------
 226
 227Unless otherwise noted, all the fixes since v2.13 in the maintenance
 228track are contained in this release (see the maintenance releases'
 229notes for details).
 230
 231 * "git gc" did not interact well with "git worktree"-managed
 232   per-worktree refs.
 233
 234 * "git cherry-pick" and other uses of the sequencer machinery
 235   mishandled a trailer block whose last line is an incomplete line.
 236   This has been fixed so that an additional sign-off etc. are added
 237   after completing the existing incomplete line.
 238
 239 * The codepath in "git am" that is used when running "git rebase"
 240   leaked memory held for the log message of the commits being rebased.
 241
 242 * "git clone --config var=val" is a way to populate the
 243   per-repository configuration file of the new repository, but it did
 244   not work well when val is an empty string.  This has been fixed.
 245
 246 * Setting "log.decorate=false" in the configuration file did not take
 247   effect in v2.13, which has been corrected.
 248
 249 * A few codepaths in "checkout" and "am" working on an unborn branch
 250   tried to access an uninitialized piece of memory.
 251
 252 * The Web interface to gmane news archive is long gone, even though
 253   the articles are still accessible via NTTP.  Replace the links with
 254   ones to public-inbox.org.  Because their message identification is
 255   based on the actual message-id, it is likely that it will be easier
 256   to migrate away from it if/when necessary.
 257
 258 * The receive-pack program now makes sure that the push certificate
 259   records the same set of push options used for pushing.
 260
 261 * Tests have been updated to pass under GETTEXT_POISON (a mechanism
 262   to ensure that output strings that should not be translated are
 263   not translated by mistake), and TravisCI is told to run them.
 264
 265 * "git checkout --recurse-submodules" did not quite work with a
 266   submodule that itself has submodules.
 267
 268 * "pack-objects" can stream a slice of an existing packfile out when
 269   the pack bitmap can tell that the reachable objects are all needed
 270   in the output, without inspecting individual objects.  This
 271   strategy however would not work well when "--local" and other
 272   options are in use, and need to be disabled.
 273
 274 * Fix memory leaks pointed out by Coverity (and people).
 275
 276 * "git read-tree -m" (no tree-ish) gave a nonsense suggestion "use
 277   --empty if you want to clear the index".  With "-m", such a request
 278   will still fail anyway, as you'd need to name at least one tree-ish
 279   to be merged.
 280
 281 * Make sure our tests would pass when the sources are checked out
 282   with "platform native" line ending convention by default on
 283   Windows.  Some "text" files out tests use and the test scripts
 284   themselves that are meant to be run with /bin/sh, ought to be
 285   checked out with eol=LF even on Windows.
 286
 287 * Introduce the BUG() macro to improve die("BUG: ...").
 288
 289 * Clarify documentation for include.path and includeIf.<condition>.path
 290   configuration variables.
 291
 292 * Git sometimes gives an advice in a rhetorical question that does
 293   not require an answer, which can confuse new users and non native
 294   speakers.  Attempt to rephrase them.
 295
 296 * A few http:// links that are redirected to https:// in the
 297   documentation have been updated to https:// links.
 298
 299 * "git for-each-ref --format=..." with %(HEAD) in the format used to
 300   resolve the HEAD symref as many times as it had processed refs,
 301   which was wasteful, and "git branch" shared the same problem.
 302
 303 * Regression fix to topic recently merged to 'master'.
 304
 305 * The shell completion script (in contrib/) learned "git stash" has
 306   a new "push" subcommand.
 307
 308 * "git interpret-trailers", when used as GIT_EDITOR for "git commit
 309   -v", looked for and appended to a trailer block at the very end,
 310   i.e. at the end of the "diff" output.  The command has been
 311   corrected to pay attention to the cut-mark line "commit -v" adds to
 312   the buffer---the real trailer block should appear just before it.
 313
 314 * A test allowed both "git push" and "git receive-pack" on the other
 315   end write their traces into the same file.  This is OK on platforms
 316   that allows atomically appending to a file opened with O_APPEND,
 317   but on other platforms led to a mangled output, causing
 318   intermittent test failures.  This has been fixed by disabling
 319   traces from "receive-pack" in the test.
 320
 321 * Tag objects, which are not reachable from any ref, that point at
 322   missing objects were mishandled by "git gc" and friends (they
 323   should silently be ignored instead)
 324
 325 * "git describe --contains" penalized light-weight tags so much that
 326   they were almost never considered.  Instead, give them about the
 327   same chance to be considered as an annotated tag that is the same
 328   age as the underlying commit would.
 329
 330 * The "run-command" API implementation has been made more robust
 331   against dead-locking in a threaded environment.
 332
 333 * A recent update to t5545-push-options.sh started skipping all the
 334   tests in the script when a web server testing is disabled or
 335   unavailable, not just the ones that require a web server.  Non HTTP
 336   tests have been salvaged to always run in this script.
 337
 338 * "git send-email" now uses Net::SMTP::SSL, which is obsolete, only
 339   when needed.  Recent versions of Net::SMTP can do TLS natively.
 340
 341 * "foo\bar\baz" in "git fetch foo\bar\baz", even though there is no
 342   slashes in it, cannot be a nickname for a remote on Windows, as
 343   that is likely to be a pathname on a local filesystem.
 344
 345 * "git clean -d" used to clean directories that has ignored files,
 346   even though the command should not lose ignored ones without "-x".
 347   "git status --ignored"  did not list ignored and untracked files
 348   without "-uall".  These have been corrected.
 349
 350 * The result from "git diff" that compares two blobs, e.g. "git diff
 351   $commit1:$path $commit2:$path", used to be shown with the full
 352   object name as given on the command line, but it is more natural to
 353   use the $path in the output and use it to look up .gitattributes.
 354
 355 * The "collision detecting" SHA-1 implementation shipped with 2.13
 356   was quite broken on some big-endian platforms and/or platforms that
 357   do not like unaligned fetches.  Update to the upstream code which
 358   has already fixed these issues.
 359
 360 * "git am -h" triggered a BUG().
 361
 362 * The interaction of "url.*.insteadOf" and custom URL scheme's
 363   whitelisting is now documented better.
 364
 365 * The timestamp of the index file is now taken after the file is
 366   closed, to help Windows, on which a stale timestamp is reported by
 367   fstat() on a file that is opened for writing and data was written
 368   but not yet closed.
 369
 370 * "git pull --rebase --autostash" didn't auto-stash when the local history
 371   fast-forwards to the upstream.
 372
 373 * A flaky test has been corrected.
 374
 375 * "git $cmd -h" for builtin commands calls the implementation of the
 376   command (i.e. cmd_$cmd() function) without doing any repository
 377   set-up, and the commands that expect RUN_SETUP is done by the Git
 378   potty needs to be prepared to show the help text without barfing.
 379   (merge d691551192 jk/consistent-h later to maint).
 380
 381 * Help contributors that visit us at GitHub.
 382
 383 * "git stash push <pathspec>" did not work from a subdirectory at all.
 384   Bugfix for a topic in v2.13
 385
 386 * As there is no portable way to pass timezone information to
 387   strftime, some output format from "git log" and friends are
 388   impossible to produce.  Teach our own strbuf_addftime to replace %z
 389   and %Z with caller-supplied values to help working around this.
 390   (merge 6eced3ec5e rs/strbuf-addftime-zZ later to maint).
 391
 392 * "git mergetool" learned to work around a wrapper MacOS X adds
 393   around underlying meld.
 394
 395 * An example in documentation that does not work in multi worktree
 396   configuration has been corrected.
 397
 398 * The pretty-format specifiers like '%h', '%t', etc. had an
 399   optimization that no longer works correctly.  In preparation/hope
 400   of getting it correctly implemented, first discard the optimization
 401   that is broken.
 402
 403 * The code to pick up and execute command alias definition from the
 404   configuration used to switch to the top of the working tree and
 405   then come back when the expanded alias was executed, which was
 406   unnecessarilyl complex.  Attempt to simplify the logic by using the
 407   early-config mechanism that does not chdir around.
 408
 409 * Fix configuration codepath to pay proper attention to commondir
 410   that is used in multi-worktree situation, and isolate config API
 411   into its own header file.
 412   (merge dc8441fdb4 bw/config-h later to maint).
 413
 414 * "git add -p" were updated in 2.12 timeframe to cope with custom
 415   core.commentchar but the implementation was buggy and a
 416   metacharacter like $ and * did not work.
 417
 418 * A recent regression in "git rebase -i" has been fixed and tests
 419   that would have caught it and others have been added.
 420
 421 * An unaligned 32-bit access in pack-bitmap code ahs been corrected.
 422
 423 * Tighten error checks for invalid "git apply" input.
 424
 425 * The split index code did not honor core.sharedrepository setting
 426   correctly.
 427
 428 * The Makefile rule in contrib/subtree for building documentation
 429   learned to honour USE_ASCIIDOCTOR just like the main documentation
 430   set does.
 431
 432 * Update the sha1dc again to fix portability glitches.
 433
 434 * Code clean-up to fix possible buffer over-reading.
 435   (merge 8bc172e5f2 rs/apply-avoid-over-reading later to maint).
 436
 437 * A few tests that tried to verify the contents of push certificates
 438   did not use 'git rev-parse' to formulate the line to look for in
 439   the certificate correctly.
 440
 441 * Update the character width tables.
 442   (merge 7560aacd7c bb/unicode-10.0 later to maint).
 443
 444 * After "git branch --move" of the currently checked out branch, the
 445   code to walk the reflog of HEAD via "log -g" and friends
 446   incorrectly stopped at the reflog entry that records the renaming
 447   of the branch.
 448   (merge e30d463d45 jk/reflog-walk-maint later to maint).
 449
 450 * Other minor doc, test and build updates and code cleanups.
 451   (merge 3f9c637ec7 pw/unquote-path-in-git-pm later to maint).
 452   (merge 669638fe7a ks/typofix-commit-c-comment later to maint).