Documentation/RelNotes/2.19.0.txt

   1 Git 2.19 Release Notes
   2 ======================
   3
   4 Updates since v2.18
   5 -------------------
   6
   7 UI, Workflows & Features
   8
   9  * "git diff" compares the index and the working tree.  For paths
  10    added with intent-to-add bit, the command shows the full contents
  11    of them as added, but the paths themselves were not marked as new
  12    files.  They are now shown as new by default.
  13
  14    "git apply" learned the "--intent-to-add" option so that an
  15    otherwise working-tree-only application of a patch will add new
  16    paths to the index marked with the "intent-to-add" bit.
  17
  18  * "git grep" learned the "--column" option that gives not just the
  19    line number but the column number of the hit.
  20
  21  * The "-l" option in "git branch -l" is an unfortunate short-hand for
  22    "--create-reflog", but many users, both old and new, somehow expect
  23    it to be something else, perhaps "--list".  This step warns when "-l"
  24    is used as a short-hand for "--create-reflog" and warns about the
  25    future repurposing of the it when it is used.
  26
  27  * The userdiff pattern for .php has been updated.
  28
  29  * The content-transfer-encoding of the message "git send-email" sends
  30    out by default was 8bit, which can cause trouble when there is an
  31    overlong line to bust RFC 5322/2822 limit.  A new option 'auto' to
  32    automatically switch to quoted-printable when there is such a line
  33    in the payload has been introduced and is made the default.
  34
  35  * "git checkout" and "git worktree add" learned to honor
  36    checkout.defaultRemote when auto-vivifying a local branch out of a
  37    remote tracking branch in a repository with multiple remotes that
  38    have tracking branches that share the same names.
  39    (merge 8d7b558bae ab/checkout-default-remote later to maint).
  40
  41  * "git grep" learned the "--only-matching" option.
  42
  43  * "git rebase --rebase-merges" mode now handles octopus merges as
  44    well.
  45
  46  * Add a server-side knob to skip commits in exponential/fibbonacci
  47    stride in an attempt to cover wider swath of history with a smaller
  48    number of iterations, potentially accepting a larger packfile
  49    transfer, instead of going back one commit a time during common
  50    ancestor discovery during the "git fetch" transaction.
  51    (merge 42cc7485a2 jt/fetch-negotiator-skipping later to maint).
  52
  53  * A new configuration variable core.usereplacerefs has been added,
  54    primarily to help server installations that want to ignore the
  55    replace mechanism altogether.
  56
  57  * Teach "git tag -s" etc. a few configuration variables (gpg.format
  58    that can be set to "openpgp" or "x509", and gpg.<format>.program
  59    that is used to specify what program to use to deal with the format)
  60    to allow x.509 certs with CMS via "gpgsm" to be used instead of
  61    openpgp via "gnupg".
  62
  63  * Many more strings are prepared for l10n.
  64
  65  * "git p4 submit" learns to ask its own pre-submit hook if it should
  66    continue with submitting.
  67
  68  * The test performed at the receiving end of "git push" to prevent
  69    bad objects from entering repository can be customized via
  70    receive.fsck.* configuration variables; we now have gained a
  71    counterpart to do the same on the "git fetch" side, with
  72    fetch.fsck.* configuration variables.
  73
  74  * "git pull --rebase=interactive" learned "i" as a short-hand for
  75    "interactive".
  76
  77 Performance, Internal Implementation, Development Support etc.
  78
  79  * The bulk of "git submodule foreach" has been rewritten in C.
  80
  81  * The in-core "commit" object had an all-purpose "void *util" field,
  82    which was tricky to use especially in library-ish part of the
  83    code.  All of the existing uses of the field has been migrated to a
  84    more dedicated "commit-slab" mechanism and the field is eliminated.
  85
  86  * A less often used command "git show-index" has been modernized.
  87    (merge fb3010c31f jk/show-index later to maint).
  88
  89  * The conversion to pass "the_repository" and then "a_repository"
  90    throughout the object access API continues.
  91
  92  * Continuing with the idea to programatically enumerate various
  93    pieces of data required for command line completion, teach the
  94    codebase to report the list of configuration variables
  95    subcommands care about to help complete them.
  96
  97  * Separate "rebase -p" codepath out of "rebase -i" implementation to
  98    slim down the latter and make it easier to manage.
  99
 100  * Make refspec parsing codepath more robust.
 101
 102  * Some flaky tests have been fixed.
 103
 104  * Continuing with the idea to programmatically enumerate various
 105    pieces of data required for command line completion, the codebase
 106    has been taught to enumerate options prefixed with "--no-" to
 107    negate them.
 108
 109  * Build and test procedure for netrc credential helper (in contrib/)
 110    has been updated.
 111
 112  * The conversion to pass "the_repository" and then "a_repository"
 113    throughout the object access API continues.
 114
 115  * Remove unused function definitions and declarations from ewah
 116    bitmap subsystem.
 117
 118  * Code preparation to make "git p4" closer to be usable with Python 3.
 119
 120  * Tighten the API to make it harder to misuse in-tree .gitmodules
 121    file, even though it shares the same syntax with configuration
 122    files, to read random configuration items from it.
 123
 124  * "git fast-import" has been updated to avoid attempting to create
 125    delta against a zero-byte-long string, which is pointless.
 126
 127  * The codebase has been updated to compile cleanly with -pedantic
 128    option.
 129    (merge 2b647a05d7 bb/pedantic later to maint).
 130
 131  * The character display width table has been updated to match the
 132    latest Unicode standard.
 133    (merge 570951eea2 bb/unicode-11-width later to maint).
 134
 135  * test-lint now looks for broken use of "VAR=VAL shell_func" in test
 136    scripts.
 137
 138  * Conversion from uchar[40] to struct object_id continues.
 139
 140  * Recent "security fix" to pay attention to contents of ".gitmodules"
 141    while accepting "git push" was a bit overly strict than necessary,
 142    which has been adjusted.
 143
 144  * "git fsck" learns to make sure the optional commit-graph file is in
 145    a sane state.
 146
 147  * "git diff --color-moved" feature has further been tweaked.
 148
 149  * Code restructuring and a small fix to transport protocol v2 during
 150    fetching.
 151
 152  * Parsing of -L[<N>][,[<M>]] parameters "git blame" and "git log"
 153    take has been tweaked.
 154
 155  * lookup_commit_reference() and friends have been updated to find
 156    in-core object for a specific in-core repository instance.
 157
 158  * Various glitches in the heuristics of merge-recursive strategy have
 159    been documented in new tests.
 160
 161  * "git fetch" learned a new option "--negotiation-tip" to limit the
 162    set of commits it tells the other end as "have", to reduce wasted
 163    bandwidth and cycles, which would be helpful when the receiving
 164    repository has a lot of refs that have little to do with the
 165    history at the remote it is fetching from.
 166
 167  * For a large tree, the index needs to hold many cache entries
 168    allocated on heap.  These cache entries are now allocated out of a
 169    dedicated memory pool to amortize malloc(3) overhead.
 170
 171  * Tests to cover various conflicting cases have been added for
 172    merge-recursive.
 173
 174  * Tests to cover conflict cases that involve submodules have been
 175    added for merge-recursive.
 176
 177  * Look for broken "&&" chains that are hidden in subshell, many of
 178    which have been found and corrected.
 179
 180  * The singleton commit-graph in-core instance is made per in-core
 181    repository instance.
 182
 183  * "make DEVELOPER=1 DEVOPTS=pedantic" allows developers to compile
 184    with -pedantic option, which may catch more problematic program
 185    constructs and potential bugs.
 186
 187  * Preparatory code to later add json output for telemetry data has
 188    been added.
 189
 190  * Update the way we use Coccinelle to find out-of-style code that
 191    need to be modernised.
 192
 193  * It is too easy to misuse system API functions such as strcat();
 194    these selected functions are now forbidden in this codebase and
 195    will cause a compilation failure.
 196
 197  * Add a script (in contrib/) to help users of VSCode work better with
 198    our codebase.
 199
 200  * The Travis CI scripts were taught to ship back the test data from
 201    failed tests.
 202    (merge aea8879a6a sg/travis-retrieve-trash-upon-failure later to maint).
 203
 204  * The parse-options machinery learned to refrain from enclosing
 205    placeholder string inside a "<bra" and "ket>" pair automatically
 206    without PARSE_OPT_LITERAL_ARGHELP.  Existing help text for option
 207    arguments that are not formatted correctly have been identified and
 208    fixed.
 209    (merge 5f0df44cd7 rs/parse-opt-lithelp later to maint).
 210
 211  * Noiseword "extern" has been removed from function decls in the
 212    header files.
 213
 214  * A few atoms like %(objecttype) and %(objectsize) in the format
 215    specifier of "for-each-ref --format=<format>" can be filled without
 216    getting the full contents of the object, but just with the object
 217    header.  These cases have been optimized by calling
 218    oid_object_info() API (instead of reading and inspecting the data).
 219
 220  * The end result of documentation update has been made to be
 221    inspected more easily to help developers.
 222
 223
 224 Fixes since v2.18
 225 -----------------
 226
 227  * "git remote update" can take both a single remote nickname and a
 228    nickname for remote groups, and the completion script (in contrib/)
 229    has been taught about it.
 230    (merge 9cd4382ad5 ls/complete-remote-update-names later to maint).
 231
 232  * "git fetch --shallow-since=<cutoff>" that specifies the cut-off
 233    point that is newer than the existing history used to end up
 234    grabbing the entire history.  Such a request now errors out.
 235    (merge e34de73c56 nd/reject-empty-shallow-request later to maint).
 236
 237  * Fix for 2.17-era regression around `core.safecrlf`.
 238    (merge 6cb09125be as/safecrlf-quiet-fix later to maint).
 239
 240  * The recent addition of "partial clone" experimental feature kicked
 241    in when it shouldn't, namely, when there is no partial-clone filter
 242    defined even if extensions.partialclone is set.
 243    (merge cac1137dc4 jh/partial-clone later to maint).
 244
 245  * "git send-pack --signed" (hence "git push --signed" over the http
 246    transport) did not read user ident from the config mechanism to
 247    determine whom to sign the push certificate as, which has been
 248    corrected.
 249    (merge d067d98887 ms/send-pack-honor-config later to maint).
 250
 251  * "git fetch-pack --all" used to unnecessarily fail upon seeing an
 252    annotated tag that points at an object other than a commit.
 253    (merge c12c9df527 jk/fetch-all-peeled-fix later to maint).
 254
 255  * When user edits the patch in "git add -p" and the user's editor is
 256    set to strip trailing whitespaces indiscriminately, an empty line
 257    that is unchanged in the patch would become completely empty
 258    (instead of a line with a sole SP on it).  The code introduced in
 259    Git 2.17 timeframe failed to parse such a patch, but now it learned
 260    to notice the situation and cope with it.
 261    (merge f4d35a6b49 pw/add-p-recount later to maint).
 262
 263  * The code to try seeing if a fetch is necessary in a submodule
 264    during a fetch with --recurse-submodules got confused when the path
 265    to the submodule was changed in the range of commits in the
 266    superproject, sometimes showing "(null)".  This has been corrected.
 267
 268  * "git submodule" did not correctly adjust core.worktree setting that
 269    indicates whether/where a submodule repository has its associated
 270    working tree across various state transitions, which has been
 271    corrected.
 272    (merge 984cd77ddb sb/submodule-core-worktree later to maint).
 273
 274  * Bugfix for "rebase -i" corner case regression.
 275    (merge a9279c6785 pw/rebase-i-keep-reword-after-conflict later to maint).
 276
 277  * Recently added "--base" option to "git format-patch" command did
 278    not correctly generate prereq patch ids.
 279    (merge 15b76c1fb3 xy/format-patch-prereq-patch-id-fix later to maint).
 280
 281  * POSIX portability fix in Makefile to fix a glitch introduced a few
 282    releases ago.
 283    (merge 6600054e9b dj/runtime-prefix later to maint).
 284
 285  * "git filter-branch" when used with the "--state-branch" option
 286    still attempted to rewrite the commits whose filtered result is
 287    known from the previous attempt (which is recorded on the state
 288    branch); the command has been corrected not to waste cycles doing
 289    so.
 290    (merge 709cfe848a mb/filter-branch-optim later to maint).
 291
 292  * Clarify that setting core.ignoreCase to deviate from reality would
 293    not turn a case-incapable filesystem into a case-capable one.
 294    (merge 48294b512a ms/core-icase-doc later to maint).
 295
 296  * "fsck.skipList" did not prevent a blob object listed there from
 297    being inspected for is contents (e.g. we recently started to
 298    inspect the contents of ".gitmodules" for certain malicious
 299    patterns), which has been corrected.
 300    (merge fb16287719 rj/submodule-fsck-skip later to maint).
 301
 302  * "git checkout --recurse-submodules another-branch" did not report
 303    in which submodule it failed to update the working tree, which
 304    resulted in an unhelpful error message.
 305    (merge ba95d4e4bd sb/submodule-move-head-error-msg later to maint).
 306
 307  * "git rebase" behaved slightly differently depending on which one of
 308    the three backends gets used; this has been documented and an
 309    effort to make them more uniform has begun.
 310    (merge b00bf1c9a8 en/rebase-consistency later to maint).
 311
 312  * The "--ignore-case" option of "git for-each-ref" (and its friends)
 313    did not work correctly, which has been fixed.
 314    (merge e674eb2528 jk/for-each-ref-icase later to maint).
 315
 316  * "git fetch" failed to correctly validate the set of objects it
 317    received when making a shallow history deeper, which has been
 318    corrected.
 319    (merge cf1e7c0770 jt/connectivity-check-after-unshallow later to maint).
 320
 321  * Partial clone support of "git clone" has been updated to correctly
 322    validate the objects it receives from the other side.  The server
 323    side has been corrected to send objects that are directly
 324    requested, even if they may match the filtering criteria (e.g. when
 325    doing a "lazy blob" partial clone).
 326    (merge a7e67c11b8 jt/partial-clone-fsck-connectivity later to maint).
 327
 328  * Handling of an empty range by "git cherry-pick" was inconsistent
 329    depending on how the range ended up to be empty, which has been
 330    corrected.
 331    (merge c5e358d073 jk/empty-pick-fix later to maint).
 332
 333  * "git reset --merge" (hence "git merge ---abort") and "git reset --hard"
 334    had trouble working correctly in a sparsely checked out working
 335    tree after a conflict, which has been corrected.
 336    (merge b33fdfc34c mk/merge-in-sparse-checkout later to maint).
 337
 338  * Correct a broken use of "VAR=VAL shell_func" in a test.
 339    (merge 650161a277 jc/t3404-one-shot-export-fix later to maint).
 340
 341  * "git rev-parse ':/substring'" did not consider the history leading
 342    only to HEAD when looking for a commit with the given substring,
 343    when the HEAD is detached.  This has been fixed.
 344    (merge 6b3351e799 wc/find-commit-with-pattern-on-detached-head later to maint).
 345
 346  * Build doc update for Windows.
 347    (merge ede8d89bb1 nd/command-list later to maint).
 348
 349  * core.commentchar is now honored when preparing the list of commits
 350    to replay in "rebase -i".
 351
 352  * "git pull --rebase" on a corrupt HEAD caused a segfault.  In
 353    general we substitute an empty tree object when running the in-core
 354    equivalent of the diff-index command, and the codepath has been
 355    corrected to do so as well to fix this issue.
 356    (merge 3506dc9445 jk/has-uncommitted-changes-fix later to maint).
 357
 358  * httpd tests saw occasional breakage due to the way its access log
 359    gets inspected by the tests, which has been updated to make them
 360    less flaky.
 361    (merge e8b3b2e275 sg/httpd-test-unflake later to maint).
 362
 363  * Tests to cover more D/F conflict cases have been added for
 364    merge-recursive.
 365
 366  * "git gc --auto" opens file descriptors for the packfiles before
 367    spawning "git repack/prune", which would upset Windows that does
 368    not want a process to work on a file that is open by another
 369    process.  The issue has been worked around.
 370    (merge 12e73a3ce4 kg/gc-auto-windows-workaround later to maint).
 371
 372  * The recursive merge strategy did not properly ensure there was no
 373    change between HEAD and the index before performing its operation,
 374    which has been corrected.
 375    (merge 55f39cf755 en/dirty-merge-fixes later to maint).
 376
 377  * "git rebase" started exporting GIT_DIR environment variable and
 378    exposing it to hook scripts when part of it got rewritten in C.
 379    Instead of matching the old scripted Porcelains' behaviour,
 380    compensate by also exporting GIT_WORK_TREE environment as well to
 381    lessen the damage.  This can harm existing hooks that want to
 382    operate on different repository, but the current behaviour is
 383    already broken for them anyway.
 384    (merge ab5e67d751 bc/sequencer-export-work-tree-as-well later to maint).
 385
 386  * "git send-email" when using in a batched mode that limits the
 387    number of messages sent in a single SMTP session lost the contents
 388    of the variable used to choose between tls/ssl, unable to send the
 389    second and later batches, which has been fixed.
 390    (merge 636f3d7ac5 jm/send-email-tls-auth-on-batch later to maint).
 391
 392  * The lazy clone support had a few places where missing but promised
 393    objects were not correctly tolerated, which have been fixed.
 394
 395  * One of the "diff --color-moved" mode "dimmed_zebra" that was named
 396    in an unusual way has been deprecated and replaced by
 397    "dimmed-zebra".
 398    (merge e3f2f5f9cd es/diff-color-moved-fix later to maint).
 399
 400  * The wire-protocol v2 relies on the client to send "ref prefixes" to
 401    limit the bandwidth spent on the initial ref advertisement.  "git
 402    clone" when learned to speak v2 forgot to do so, which has been
 403    corrected.
 404    (merge 402c47d939 bw/clone-ref-prefixes later to maint).
 405
 406  * "git diff --histogram" had a bad memory usage pattern, which has
 407    been rearranged to reduce the peak usage.
 408    (merge 79cb2ebb92 sb/histogram-less-memory later to maint).
 409
 410  * Code clean-up to use size_t/ssize_t when they are the right type.
 411    (merge 7726d360b5 jk/size-t later to maint).
 412
 413  * The wire-protocol v2 relies on the client to send "ref prefixes" to
 414    limit the bandwidth spent on the initial ref advertisement.  "git
 415    fetch $remote branch:branch" that asks tags that point into the
 416    history leading to the "branch" automatically followed sent to
 417    narrow prefix and broke the tag following, which has been fixed.
 418    (merge 2b554353a5 jt/tag-following-with-proto-v2-fix later to maint).
 419
 420  * When the sparse checkout feature is in use, "git cherry-pick" and
 421    other mergy operations lost the skip_worktree bit when a path that
 422    is excluded from checkout requires content level merge, which is
 423    resolved as the same as the HEAD version, without materializing the
 424    merge result in the working tree, which made the path appear as
 425    deleted.  This has been corrected by preserving the skip_worktree
 426    bit (and not materializing the file in the working tree).
 427    (merge 2b75fb601c en/merge-recursive-skip-fix later to maint).
 428
 429  * The "author-script" file "git rebase -i" creates got broken when
 430    we started to move the command away from shell script, which is
 431    getting fixed now.
 432    (merge 5522bbac20 es/rebase-i-author-script-fix later to maint).
 433
 434  * The automatic tree-matching in "git merge -s subtree" was broken 5
 435    years ago and nobody has noticed since then, which is now fixed.
 436    (merge 2ec4150713 jk/merge-subtree-heuristics later to maint).
 437
 438  * "git fetch $there refs/heads/s" ought to fetch the tip of the
 439    branch 's', but when "refs/heads/refs/heads/s", i.e. a branch whose
 440    name is "refs/heads/s" exists at the same time, fetched that one
 441    instead by mistake.  This has been corrected to honor the usual
 442    disambiguation rules for abbreviated refnames.
 443    (merge 60650a48c0 jt/refspec-dwim-precedence-fix later to maint).
 444
 445  * Futureproofing a helper function that can easily be misused.
 446    (merge 65bb21e77e es/want-color-fd-defensive later to maint).
 447
 448  * The http-backend (used for smart-http transport) used to slurp the
 449    whole input until EOF, without paying attention to CONTENT_LENGTH
 450    that is supplied in the environment and instead expecting the Web
 451    server to close the input stream.  This has been fixed.
 452    (merge eebfe40962 mk/http-backend-content-length later to maint).
 453
 454  * "git merge --abort" etc. did not clean things up properly when
 455    there were conflicted entries in the index in certain order that
 456    are involved in D/F conflicts.  This has been corrected.
 457    (merge ad3762042a en/abort-df-conflict-fixes later to maint).
 458
 459  * "git diff --indent-heuristic" had a bad corner case performance.
 460    (merge 301ef85401 sb/indent-heuristic-optim later to maint).
 461
 462  * Code cleanup, docfix, build fix, etc.
 463    (merge aee9be2ebe sg/update-ref-stdin-cleanup later to maint).
 464    (merge 037714252f jc/clean-after-sanity-tests later to maint).
 465    (merge 5b26c3c941 en/merge-recursive-cleanup later to maint).
 466    (merge 0dcbc0392e bw/config-refer-to-gitsubmodules-doc later to maint).
 467    (merge bb4d000e87 bw/protocol-v2 later to maint).
 468    (merge 928f0ab4ba vs/typofixes later to maint).
 469    (merge d7f590be84 en/rebase-i-microfixes later to maint).
 470    (merge 81d395cc85 js/rebase-recreate-merge later to maint).
 471    (merge 51d1863168 tz/exclude-doc-smallfixes later to maint).
 472    (merge a9aa3c0927 ds/commit-graph later to maint).
 473    (merge 5cf8e06474 js/enhanced-version-info later to maint).
 474    (merge 6aaded5509 tb/config-default later to maint).
 475    (merge 022d2ac1f3 sb/blame-color later to maint).
 476    (merge 5a06a20e0c bp/test-drop-caches-for-windows later to maint).
 477    (merge dd61cc1c2e jk/ui-color-always-to-auto later to maint).
 478    (merge 1e83b9bfdd sb/trailers-docfix later to maint).
 479    (merge ab29f1b329 sg/fast-import-dump-refs-on-checkpoint-fix later to maint).
 480    (merge 6a8ad880f0 jn/subtree-test-fixes later to maint).
 481    (merge ffbd51cc60 nd/pack-objects-threading-doc later to maint).
 482    (merge e9dac7be60 es/mw-to-git-chain-fix later to maint).
 483    (merge fe583c6c7a rs/remote-mv-leakfix later to maint).
 484    (merge 69885ab015 en/t3031-title-fix later to maint).
 485    (merge 8578037bed nd/config-blame-sort later to maint).
 486    (merge 8ad169c4ba hn/config-in-code-comment later to maint).
 487    (merge b7446fcfdf ar/t4150-am-scissors-test-fix later to maint).