4 Backward compatibility notes.
6 * An empty string used as a pathspec element has always meant
7 'everything matches', but it is too easy to write a script that
8 finds a path to remove in $path and run 'git rm "$paht"', which
9 ends up removing everything. This release starts warning about the
10 use of an empty string that is used for 'everything matches' and
11 asks users to use a more explicit '.' for that instead.
13 The hope is that existing users will not mind this change, and
14 eventually the warning can be turned into a hard error, upgrading
15 the deprecation into removal of this (mis)feature.
17 * The historical argument order "git merge <msg> HEAD <commit>..."
18 has been deprecated for quite some time, and will be removed in the
19 next release (not this one).
21 * The default abbreviation length, which has historically been 7, now
22 scales as the repository grows, using the approximate number of
23 objects in the reopsitory and a bit of math around the birthday
24 paradox. The logic suggests to use 12 hexdigits for the Linux
25 kernel, and 9 to 10 for Git itself.
31 UI, Workflows & Features
33 * Comes with new version of git-gui, now at its 0.21.0 tag.
35 * "git format-patch --cover-letter HEAD^" to format a single patch
36 with a separate cover letter now numbers the output as [PATCH 0/1]
37 and [PATCH 1/1] by default.
39 * An incoming "git push" that attempts to push too many bytes can now
40 be rejected by setting a new configuration variable at the receiving
43 * "git nosuchcommand --help" said "No manual entry for gitnosuchcommand",
44 which was not intuitive, given that "git nosuchcommand" said "git:
45 'nosuchcommand' is not a git command".
47 * "git clone --resurse-submodules --reference $path $URL" is a way to
48 reduce network transfer cost by borrowing objects in an existing
49 $path repository when cloning the superproject from $URL; it
50 learned to also peek into $path for presense of corresponding
51 repositories of submodules and borrow objects from there when able.
53 * The "git diff --submodule={short,log}" mechanism has been enhanced
54 to allow "--submodule=diff" to show the patch between the submodule
55 commits bound to the superproject.
57 * Even though "git hash-objects", which is a tool to take an
58 on-filesystem data stream and put it into the Git object store,
59 allowed to perform the "outside-world-to-Git" conversions (e.g.
60 end-of-line conversions and application of the clean-filter), and
61 it had the feature on by default from very early days, its reverse
62 operation "git cat-file", which takes an object from the Git object
63 store and externalize for the consumption by the outside world,
64 lacked an equivalent mechanism to run the "Git-to-outside-world"
65 conversion. The command learned the "--filters" option to do so.
67 * Output from "git diff" can be made easier to read by selecting
68 which lines are common and which lines are added/deleted
69 intelligently when the lines before and after the changed section
70 are the same. A command line option is added to help with the
71 experiment to find a good heuristics.
73 * In some projects, it is common to use "[RFC PATCH]" as the subject
74 prefix for a patch meant for discussion rather than application. A
75 new option "--rfc" was a short-hand for "--subject-prefix=RFC PATCH"
76 to help the participants of such projects.
78 * "git add --chmod=+x <pathspec>" added recently only toggled the
79 executable bit for paths that are either new or modified. This has
80 been corrected to flip the executable bit for all paths that match
83 * When "git format-patch --stdout" output is placed as an in-body
84 header and it uses the RFC2822 header folding, "git am" failed to
85 put the header line back into a single logical line. The
86 underlying "git mailinfo" was taught to handle this properly.
88 * "gitweb" can spawn "highlight" to show blob contents with
89 (programming) language-specific syntax highlighting, but only
90 when the language is known. "highlight" can however be told
91 to make the guess itself by giving it "--force" option, which
94 * "git gui" l10n to Portuguese.
96 * When given an abbreviated object name that is not (or more
97 realistically, "no longer") unique, we gave a fatal error
98 "ambiguous argument". This error is now accompanied by hints that
99 lists the objects that begins with the given prefix. During the
100 course of development of this new feature, numerous minor bugs were
101 uncovered and corrected, the most notable one of which is that we
102 gave "short SHA1 xxxx is ambiguous." twice without good reason.
104 * "git log rev^..rev" is an often-used revision range specification
105 to show what was done on a side branch merged at rev. This has
106 gained a short-hand "rev^-1". In general "rev^-$n" is the same as
107 "^rev^$n rev", i.e. what has happened on other branches while the
108 history leading to nth parent was looking the other way.
110 * In recent versions of cURL, GSSAPI credential delegation is
111 disabled by default due to CVE-2011-2192; introduce a configuration
112 to selectively allow enabling this.
113 (merge 26a7b23429 ps/http-gssapi-cred-delegation later to maint).
115 * "git mergetool" learned to honor "-O<orderfile>" to control the
116 order of paths to present to the end user.
118 * "git diff/log --ws-error-highlight=<kind>" lacked the corresponding
119 configuration variable to set it by default.
121 * "git ls-files" learned "--recurse-submodules" option that can be
122 used to get a listing of tracked files across submodules (i.e. this
123 only works with "--cached" option, not for listing untracked or
124 ignored files). This would be a useful tool to sit on the upstream
125 side of a pipe that is read with xargs to work on all working tree
126 files from the top-level superproject.
128 * A new credential helper that talks via "libsecret" with
129 implementations of XDG Secret Service API has been added to
132 * The GPG verification status shown in "%G?" pretty format specifier
133 was not rich enough to differentiate a signature made by an expired
134 key, a signature made by a revoked key, etc. New output letters
135 have been assigned to express them.
137 * In addition to purely abbreviated commit object names, "gitweb"
138 learned to turn "git describe" output (e.g. v2.9.3-599-g2376d31787)
139 into clickable links in its output.
141 * When new paths were added by "git add -N" to the index, it was
142 enough to circumvent the check by "git commit" to refrain from
143 making an empty commit without "--allow-empty". The same logic
144 prevented "git status" to show such a path as "new file" in the
145 "Changes not staged for commit" section.
148 Performance, Internal Implementation, Development Support etc.
150 * The delta-base-cache mechanism has been a key to the performance in
151 a repository with a tightly packed packfile, but it did not scale
152 well even with a larger value of core.deltaBaseCacheLimit.
154 * Enhance "git status --porcelain" output by collecting more data on
155 the state of the index and the working tree files, which may
156 further be used to teach git-prompt (in contrib/) to make fewer
159 * Extract a small helper out of the function that reads the authors
160 script file "git am" internally uses.
161 (merge a77598e jc/am-read-author-file later to maint).
163 * Lifts calls to exit(2) and die() higher in the callchain in
164 sequencer.c files so that more helper functions in it can be used
165 by callers that want to handle error conditions themselves.
167 * "git am" has been taught to make an internal call to "git apply"'s
168 innards without spawning the latter as a separate process.
170 * The ref-store abstraction was introduced to the refs API so that we
171 can plug in different backends to store references.
173 * The "unsigned char sha1[20]" to "struct object_id" conversion
174 continues. Notable changes in this round includes that ce->sha1,
175 i.e. the object name recorded in the cache_entry, turns into an
178 * JGit can show a fake ref "capabilities^{}" to "git fetch" when it
179 does not advertise any refs, but "git fetch" was not prepared to
180 see such an advertisement. When the other side disconnects without
181 giving any ref advertisement, we used to say "there may not be a
182 repository at that URL", but we may have seen other advertisement
183 like "shallow" and ".have" in which case we definitely know that a
184 repository is there. The code to detect this case has also been
187 * Some codepaths in "git pack-objects" were not ready to use an
188 existing pack bitmap; now they are and as the result they have
191 * The codepath in "git fsck" to detect malformed tree objects has
192 been updated not to die but keep going after detecting them.
194 * We call "qsort(array, nelem, sizeof(array[0]), fn)", and most of
195 the time third parameter is redundant. A new QSORT() macro lets us
198 * "git pack-objects" in a repository with many packfiles used to
199 spend a lot of time looking for/at objects in them; the accesses to
200 the packfiles are now optimized by checking the most-recently-used
202 (merge c9af708b1a jk/pack-objects-optim-mru later to maint).
204 * Codepaths involved in interacting alternate object store have
207 * In order for the receiving end of "git push" to inspect the
208 received history and decide to reject the push, the objects sent
209 from the sending end need to be made available to the hook and
210 the mechanism for the connectivity check, and this was done
211 traditionally by storing the objects in the receiving repository
212 and letting "git gc" to expire it. Instead, store the newly
213 received objects in a temporary area, and make them available by
214 reusing the alternate object store mechanism to them only while we
215 decide if we accept the check, and once we decide, either migrate
216 them to the repository or purge them immediately.
218 * The require_clean_work_tree() helper was recreated in C when "git
219 pull" was rewritten from shell; the helper is now made available to
220 other callers in preparation for upcoming "rebase -i" work.
222 * "git upload-pack" had its code cleaned-up and performance improved
223 by reducing use of timestamp-ordered commit-list, which was
224 replaced with a priority queue.
226 * "git diff --no-index" codepath has been updated not to try to peek
227 into .git/ directory that happens to be under the current
228 directory, when we know we are operating outside any repository.
230 * Update of the sequencer codebase to make it reusable to reimplement
231 "rebase -i" continues.
234 Also contains various documentation updates and code clean-ups.
240 Unless otherwise noted, all the fixes since v2.9 in the maintenance
241 track are contained in this release (see the maintenance releases'
244 * Clarify various ways to specify the "revision ranges" in the
247 * "diff-highlight" script (in contrib/) learned to work better with
248 "git log -p --graph" output.
250 * The test framework left the number of tests and success/failure
251 count in the t/test-results directory, keyed by the name of the
252 test script plus the process ID. The latter however turned out not
253 to serve any useful purpose. The process ID part of the filename
256 * Having a submodule whose ".git" repository is somehow corrupt
257 caused a few commands that recurse into submodules loop forever.
259 * "git symbolic-ref -d HEAD" happily removes the symbolic ref, but
260 the resulting repository becomes an invalid one. Teach the command
261 to forbid removal of HEAD.
263 * A test spawned a short-lived background process, which sometimes
264 prevented the test directory from getting removed at the end of the
265 script on some platforms.
267 * Update a few tests that used to use GIT_CURL_VERBOSE to use the
268 newer GIT_TRACE_CURL.
270 * "git pack-objects --include-tag" was taught that when we know that
271 we are sending an object C, we want a tag B that directly points at
272 C but also a tag A that points at the tag B. We used to miss the
273 intermediate tag B in some cases.
275 * Update Japanese translation for "git-gui".
277 * "git fetch http::/site/path" did not die correctly and segfaulted
280 * "git commit-tree" stopped reading commit.gpgsign configuration
281 variable that was meant for Porcelain "git commit" in Git 2.9; we
282 forgot to update "git gui" to look at the configuration to match
285 * "git add --chmod=+x" added recently lacked documentation, which has
288 * "git log --cherry-pick" used to include merge commits as candidates
289 to be matched up with other commits, resulting a lot of wasted time.
290 The patch-id generation logic has been updated to ignore merges to
293 * The http transport (with curl-multi option, which is the default
294 these days) failed to remove curl-easy handle from a curlm session,
295 which led to unnecessary API failures.
297 * There were numerous corner cases in which the configuration files
298 are read and used or not read at all depending on the directory a
299 Git command was run, leading to inconsistent behaviour. The code
300 to set-up repository access at the beginning of a Git process has
301 been updated to fix them.
302 (merge 4d0efa1 jk/setup-sequence-update later to maint).
304 * "git diff -W" output needs to extend the context backward to
305 include the header line of the current function and also forward to
306 include the body of the entire current function up to the header
307 line of the next one. This process may have to merge to adjacent
308 hunks, but the code forgot to do so in some cases.
310 * Performance tests done via "t/perf" did not use the same set of
311 build configuration if the user relied on autoconf generated
314 * "git format-patch --base=..." feature that was recently added
315 showed the base commit information after "-- " e-mail signature
316 line, which turned out to be inconvenient. The base information
317 has been moved above the signature line.
321 * Even when "git pull --rebase=preserve" (and the underlying "git
322 rebase --preserve") can complete without creating any new commit
323 (i.e. fast-forwards), it still insisted on having a usable ident
324 information (read: user.email is set correctly), which was less
325 than nice. As the underlying commands used inside "git rebase"
326 would fail with a more meaningful error message and advice text
327 when the bogus ident matters, this extra check was removed.
329 * "git gc --aggressive" used to limit the delta-chain length to 250,
330 which is way too deep for gaining additional space savings and is
331 detrimental for runtime performance. The limit has been reduced to
334 * Documentation for individual configuration variables to control use
335 of color (like `color.grep`) said that their default value is
336 'false', instead of saying their default is taken from `color.ui`.
337 When we updated the default value for color.ui from 'false' to
338 'auto' quite a while ago, all of them broke. This has been
341 * The pretty-format specifier "%C(auto)" used by the "log" family of
342 commands to enable coloring of the output is taught to also issue a
343 color-reset sequence to the output.
344 (merge 82b83da8d3 rs/c-auto-resets-attributes later to maint).
346 * A shell script example in check-ref-format documentation has been
349 * "git checkout <word>" does not follow the usual disambiguation
350 rules when the <word> can be both a rev and a path, to allow
351 checking out a branch 'foo' in a project that happens to have a
352 file 'foo' in the working tree without having to disambiguate.
353 This was poorly documented and the check was incorrect when the
354 command was run from a subdirectory.
356 * Some codepaths in "git diff" used regexec(3) on a buffer that was
357 mmap(2)ed, which may not have a terminating NUL, leading to a read
358 beyond the end of the mapped region. This was fixed by introducing
359 a regexec_buf() helper that takes a <ptr,len> pair with REG_STARTEND
361 (merge 842a516cb0 js/regexec-buf later to maint).
363 * The procedure to build Git on Mac OS X for Travis CI hardcoded the
364 internal directory structure we assumed HomeBrew uses, which was a
365 no-no. The procedure has been updated to ask HomeBrew things we
366 need to know to fix this.
368 * When "git rebase -i" is given a broken instruction, it told the
369 user to fix it with "--edit-todo", but didn't say what the step
370 after that was (i.e. "--continue").
372 * Documentation around tools to import from CVS was fairly outdated.
374 * "git clone --recurse-submodules" lost the progress eye-candy in
375 recent update, which has been corrected.
377 * A low-level function verify_packfile() was meant to show errors
378 that were detected without dying itself, but under some conditions
379 it didn't and died instead, which has been fixed.
381 * When "git fetch" tries to find where the history of the repository
382 it runs in has diverged from what the other side has, it has a
383 mechanism to avoid digging too deep into irrelevant side branches.
384 This however did not work well over the "smart-http" transport due
385 to a design bug, which has been fixed.
386 (merge 06b3d386e0 jt/fetch-pack-in-vain-count-with-stateless later to maint).
388 * In the codepath that comes up with the hostname to be used in an
389 e-mail when the user didn't tell us, we looked at ai_canonname
390 field in struct addrinfo without making sure it is not NULL first.
392 * "git worktree", even though it used the default_abbrev setting that
393 ought to be affected by core.abbrev configuration variable, ignored
394 the variable setting. The command has been taught to read the
395 default set of configuration variables to correct this.
397 * "git init" tried to record core.worktree in the repository's
398 'config' file when GIT_WORK_TREE environment variable was set and
399 it was different from where GIT_DIR appears as ".git" at its top,
400 but the logic was faulty when .git is a "gitdir:" file that points
401 at the real place, causing trouble in working trees that are
402 managed by "git worktree". This has been corrected.
404 * Codepaths that read from an on-disk loose object were too loose in
405 validating what they are reading is a proper object file and
406 sometimes read past the data they read from the disk, which has
407 been corrected. H/t to Gustavo Grieco for reporting.
409 * The original command line syntax for "git merge", which was "git
410 merge <msg> HEAD <parent>...", has been deprecated for quite some
411 time, and "git gui" was the last in-tree user of the syntax. This
412 is finally fixed, so that we can move forward with the deprecation.
414 * An author name, that spelled a backslash-quoted double quote in the
415 human readable part "My \"double quoted\" name", was not unquoted
416 correctly while applying a patch from a piece of e-mail.
418 * Doc update to clarify what "log -3 --reverse" does.
420 * Almost everybody uses DEFAULT_ABBREV to refer to the default
421 setting for the abbreviation, but "git blame" peeked into
422 underlying variable bypassing the macro for no good reason.
424 * The "graph" API used in "git log --graph" miscounted the number of
425 output columns consumed so far when drawing a padding line, which
426 has been fixed; this did not affect any existing code as nobody
427 tried to write anything after the padding on such a line, though.
429 * The code that parses the format parameter of for-each-ref command
430 has seen a micro-optimization.
432 * When we started cURL to talk to imap server when a new enough
433 version of cURL library is available, we forgot to explicitly add
434 imap(s):// before the destination. To some folks, that didn't work
435 and the library tried to make HTTP(s) requests instead.
436 (merge d2d07ab861 ak/curl-imap-send-explicit-scheme later to maint).
438 * The ./configure script generated from configure.ac was taught how
439 to detect support of SSL by libcurl better.
440 (merge 924b7eb1c9 dp/autoconf-curl-ssl later to maint).
442 * The command-line completion script (in contrib/) learned to
443 complete "git cmd ^mas<HT>" to complete the negative end of
444 reference to "git cmd ^master".
445 (merge 49416ad22a cp/completion-negative-refs later to maint).
447 * The existing "git fetch --depth=<n>" option was hard to use
448 correctly when making the history of an existing shallow clone
449 deeper. A new option, "--deepen=<n>", has been added to make this
450 easier to use. "git clone" also learned "--shallow-since=<date>"
451 and "--shallow-exclude=<tag>" options to make it easier to specify
452 "I am interested only in the recent N months worth of history" and
453 "Give me only the history since that version".
454 (merge cccf74e2da nd/shallow-deepen later to maint).
456 * It is a common mistake to say "git blame --reverse OLD path",
457 expecting that the command line is dwimmed as if asking how lines
458 in path in an old revision OLD have survived up to the current
460 (merge e1d09701a4 jc/blame-reverse later to maint).
462 * http.emptyauth configuration is a way to allow an empty username to
463 pass when attempting to authenticate using mechanisms like
464 Kerberos. We took an unspecified (NULL) username and sent ":"
465 (i.e. no username, no password) to CURLOPT_USERPWD, but did not do
466 the same when the username is explicitly set to an empty string.
467 (merge 5275c3081c dt/http-empty-auth later to maint).
469 * "git clone" of a local repository can be done at the filesystem
470 level, but the codepath did not check errors while copying and
471 adjusting the file that lists alternate object stores.
472 (merge 22d3b8de1b jk/clone-copy-alternates-fix later to maint).
474 * Documentation for "git commit" was updated to clarify that "commit
475 -p <paths>" adds to the current contents of the index to come up
477 (merge 7431596ab1 nd/commit-p-doc later to maint).
479 * A stray symbolic link in $GIT_DIR/refs/ directory could make name
480 resolution loop forever, which has been corrected.
481 (merge e8c42cb9ce jk/ref-symlink-loop later to maint).
483 * The "submodule.<name>.path" stored in .gitmodules is never copied
484 to .git/config and such a key in .git/config has no meaning, but
485 the documentation described it and submodule.<name>.url next to
486 each other as if both belong to .git/config. This has been fixed.
487 (merge 72710165c9 sb/submodule-config-doc-drop-path later to maint).
489 * In a worktree connected to a repository elsewhere, created via "git
490 worktree", "git checkout" attempts to protect users from confusion
491 by refusing to check out a branch that is already checked out in
492 another worktree. However, this also prevented checking out a
493 branch, which is designated as the primary branch of a bare
494 reopsitory, in a worktree that is connected to the bare
495 repository. The check has been corrected to allow it.
496 (merge 171c646f8c dk/worktree-dup-checkout-with-bare-is-ok later to maint).
498 * "git rebase" immediately after "git clone" failed to find the fork
499 point from the upstream.
500 (merge 4f21454b55 jk/merge-base-fork-point-without-reflog later to maint).
502 * When fetching from a remote that has many tags that are irrelevant
503 to branches we are following, we used to waste way too many cycles
504 when checking if the object pointed at by a tag (that we are not
505 going to fetch!) exists in our repository too carefully.
506 (merge 5827a03545 jk/fetch-quick-tag-following later to maint).
508 * Protect our code from over-eager compilers.
509 (merge 0ac52a38e8 jk/tighten-alloc later to maint).
511 * Recent git allows submodule.<name>.branch to use a special token
512 "." instead of the branch name; the documentation has been updated
514 (merge 15ef78008a bw/submodule-branch-dot-doc later to maint).
516 * A hot-fix for a test added by a recent topic that went to both
517 'master' and 'maint' already.
518 (merge 76e368c378 tg/add-chmod+x-fix later to maint).
520 * "git send-email" attempts to pick up valid e-mails from the
521 trailers, but people in real world write non-addresses there, like
522 "Cc: Stable <add@re.ss> # 4.8+", which broke the output depending
523 on the availability and vintage of Mail::Address perl module.
524 (merge dcfafc5214 mm/send-email-cc-cruft-after-address later to maint).
526 * The Travis CI configuration we ship ran the tests with --verbose
527 option but this risks non-TAP output that happens to be "ok" to be
528 misinterpreted as TAP signalling a test that passed. This resulted
529 in unnecessary failure. This has been corrected by introducing a
530 new mode to run our tests in the test harness to send the verbose
531 output separately to the log file.
532 (merge 614fe01521 jk/tap-verbose-fix later to maint).
534 * Some AsciiDoc formatter mishandles a displayed illustration with
535 tabs in it. Adjust a few of them in merge-base documentation to
537 (merge 6750f62699 po/fix-doc-merge-base-illustration later to maint).
539 * A minor regression fix for "git submodule" that was introduced
540 when more helper functions were reimplemented in C.
541 (merge 77b63ac31e sb/submodule-ignore-trailing-slash later to maint).
543 * The code that we have used for the past 10+ years to cycle
544 4-element ring buffers turns out to be not quite portable in
546 (merge bb84735c80 rs/ring-buffer-wraparound later to maint).
548 * "git daemon" used fixed-length buffers to turn URL to the
549 repository the client asked for into the server side directory
550 path, using snprintf() to avoid overflowing these buffers, but
551 allowed possibly truncated paths to the directory. This has been
552 tightened to reject such a request that causes overlong path to be
554 (merge 6bdb0083be jk/daemon-path-ok-check-truncation later to maint).
556 * Other minor doc, test and build updates and code cleanups.
557 (merge a94bb68397 rs/cocci later to maint).
558 (merge 641c900b2c js/reset-usage later to maint).
559 (merge 30cfe72d37 rs/pretty-format-color-doc-fix later to maint).
560 (merge d709f1fb9d jc/diff-unique-abbrev-comments later to maint).
561 (merge 13092a916d jc/cocci-xstrdup-or-null later to maint).
562 (merge 86009f32bb pb/test-parse-options-expect later to maint).
563 (merge 749a2279a4 yk/git-tag-remove-mention-of-old-layout-in-doc later to maint).
projects / git / blob