7 UI, Workflows & Features
9 * "git format-patch --cover-letter HEAD^" to format a single patch
10 with a separate cover letter now numbers the output as [PATCH 0/1]
11 and [PATCH 1/1] by default.
13 * An incoming "git push" that attempts to push too many bytes can now
14 be rejected by setting a new configuration variable at the receiving
17 * "git nosuchcommand --help" said "No manual entry for gitnosuchcommand",
18 which was not intuitive, given that "git nosuchcommand" said "git:
19 'nosuchcommand' is not a git command".
21 * "git clone --resurse-submodules --reference $path $URL" is a way to
22 reduce network transfer cost by borrowing objects in an existing
23 $path repository when cloning the superproject from $URL; it
24 learned to also peek into $path for presense of corresponding
25 repositories of submodules and borrow objects from there when able.
27 * The "git diff --submodule={short,log}" mechanism has been enhanced
28 to allow "--submodule=diff" to show the patch between the submodule
29 commits bound to the superproject.
31 * Even though "git hash-objects", which is a tool to take an
32 on-filesystem data stream and put it into the Git object store,
33 allowed to perform the "outside-world-to-Git" conversions (e.g.
34 end-of-line conversions and application of the clean-filter), and
35 it had the feature on by default from very early days, its reverse
36 operation "git cat-file", which takes an object from the Git object
37 store and externalize for the consumption by the outside world,
38 lacked an equivalent mechanism to run the "Git-to-outside-world"
39 conversion. The command learned the "--filters" option to do so.
41 * Output from "git diff" can be made easier to read by selecting
42 which lines are common and which lines are added/deleted
43 intelligently when the lines before and after the changed section
44 are the same. A command line option is added to help with the
45 experiment to find a good heuristics.
47 * In some projects, it is common to use "[RFC PATCH]" as the subject
48 prefix for a patch meant for discussion rather than application. A
49 new option "--rfc" was a short-hand for "--subject-prefix=RFC PATCH"
50 to help the participants of such projects.
52 * "git add --chmod=+x <pathspec>" added recently only toggled the
53 executable bit for paths that are either new or modified. This has
54 been corrected to flip the executable bit for all paths that match
57 * When "git format-patch --stdout" output is placed as an in-body
58 header and it uses the RFC2822 header folding, "git am" failed to
59 put the header line back into a single logical line. The
60 underlying "git mailinfo" was taught to handle this properly.
62 * "gitweb" can spawn "highlight" to show blob contents with
63 (programming) language-specific syntax highlighting, but only
64 when the language is known. "highlight" can however be told
65 to make the guess itself by giving it "--force" option, which
68 * "git gui" l10n to Portuguese.
70 * When given an abbreviated object name that is not (or more
71 realistically, "no longer") unique, we gave a fatal error
72 "ambiguous argument". This error is now accompanied by hints that
73 lists the objects that begins with the given prefix. During the
74 course of development of this new feature, numerous minor bugs were
75 uncovered and corrected, the most notable one of which is that we
76 gave "short SHA1 xxxx is ambiguous." twice without good reason.
78 * "git log rev^..rev" is an often-used revision range specification
79 to show what was done on a side branch merged at rev. This has
80 gained a short-hand "rev^-1". In general "rev^-$n" is the same as
81 "^rev^$n rev", i.e. what has happened on other branches while the
82 history leading to nth parent was looking the other way.
84 * In recent versions of cURL, GSSAPI credential delegation is
85 disabled by default due to CVE-2011-2192; introduce a configuration
86 to selectively allow enabling this.
87 (merge 26a7b23429 ps/http-gssapi-cred-delegation later to maint).
89 * "git mergetool" learned to honor "-O<orderfile>" to control the
90 order of paths to present to the end user.
93 Performance, Internal Implementation, Development Support etc.
95 * The delta-base-cache mechanism has been a key to the performance in
96 a repository with a tightly packed packfile, but it did not scale
97 well even with a larger value of core.deltaBaseCacheLimit.
99 * Enhance "git status --porcelain" output by collecting more data on
100 the state of the index and the working tree files, which may
101 further be used to teach git-prompt (in contrib/) to make fewer
104 * Extract a small helper out of the function that reads the authors
105 script file "git am" internally uses.
106 (merge a77598e jc/am-read-author-file later to maint).
108 * Lifts calls to exit(2) and die() higher in the callchain in
109 sequencer.c files so that more helper functions in it can be used
110 by callers that want to handle error conditions themselves.
112 * "git am" has been taught to make an internal call to "git apply"'s
113 innards without spawning the latter as a separate process.
115 * The ref-store abstraction was introduced to the refs API so that we
116 can plug in different backends to store references.
118 * The "unsigned char sha1[20]" to "struct object_id" conversion
119 continues. Notable changes in this round includes that ce->sha1,
120 i.e. the object name recorded in the cache_entry, turns into an
123 * JGit can show a fake ref "capabilities^{}" to "git fetch" when it
124 does not advertise any refs, but "git fetch" was not prepared to
125 see such an advertisement. When the other side disconnects without
126 giving any ref advertisement, we used to say "there may not be a
127 repository at that URL", but we may have seen other advertisement
128 like "shallow" and ".have" in which case we definitely know that a
129 repository is there. The code to detect this case has also been
132 * Some codepaths in "git pack-objects" were not ready to use an
133 existing pack bitmap; now they are and as the result they have
136 * The codepath in "git fsck" to detect malformed tree objects has
137 been updated not to die but keep going after detecting them.
139 * We call "qsort(array, nelem, sizeof(array[0]), fn)", and most of
140 the time third parameter is redundant. A new QSORT() macro lets us
143 * "git pack-objects" in a repository with many packfiles used to
144 spend a lot of time looking for/at objects in them; the accesses to
145 the packfiles are now optimized by checking the most-recently-used
147 (merge c9af708b1a jk/pack-objects-optim-mru later to maint).
149 * Codepaths involved in interacting alternate object store have
152 * In order for the receiving end of "git push" to inspect the
153 received history and decide to reject the push, the objects sent
154 from the sending end need to be made available to the hook and
155 the mechanism for the connectivity check, and this was done
156 traditionally by storing the objects in the receiving repository
157 and letting "git gc" to expire it. Instead, store the newly
158 received objects in a temporary area, and make them available by
159 reusing the alternate object store mechanism to them only while we
160 decide if we accept the check, and once we decide, either migrate
161 them to the repository or purge them immediately.
164 Also contains various documentation updates and code clean-ups.
170 Unless otherwise noted, all the fixes since v2.9 in the maintenance
171 track are contained in this release (see the maintenance releases'
174 * Clarify various ways to specify the "revision ranges" in the
177 * "diff-highlight" script (in contrib/) learned to work better with
178 "git log -p --graph" output.
180 * The test framework left the number of tests and success/failure
181 count in the t/test-results directory, keyed by the name of the
182 test script plus the process ID. The latter however turned out not
183 to serve any useful purpose. The process ID part of the filename
186 * Having a submodule whose ".git" repository is somehow corrupt
187 caused a few commands that recurse into submodules loop forever.
189 * "git symbolic-ref -d HEAD" happily removes the symbolic ref, but
190 the resulting repository becomes an invalid one. Teach the command
191 to forbid removal of HEAD.
193 * A test spawned a short-lived background process, which sometimes
194 prevented the test directory from getting removed at the end of the
195 script on some platforms.
197 * Update a few tests that used to use GIT_CURL_VERBOSE to use the
198 newer GIT_TRACE_CURL.
200 * "git pack-objects --include-tag" was taught that when we know that
201 we are sending an object C, we want a tag B that directly points at
202 C but also a tag A that points at the tag B. We used to miss the
203 intermediate tag B in some cases.
205 * Update Japanese translation for "git-gui".
207 * "git fetch http::/site/path" did not die correctly and segfaulted
210 * "git commit-tree" stopped reading commit.gpgsign configuration
211 variable that was meant for Porcelain "git commit" in Git 2.9; we
212 forgot to update "git gui" to look at the configuration to match
215 * "git add --chmod=+x" added recently lacked documentation, which has
218 * "git log --cherry-pick" used to include merge commits as candidates
219 to be matched up with other commits, resulting a lot of wasted time.
220 The patch-id generation logic has been updated to ignore merges to
223 * The http transport (with curl-multi option, which is the default
224 these days) failed to remove curl-easy handle from a curlm session,
225 which led to unnecessary API failures.
227 * There were numerous corner cases in which the configuration files
228 are read and used or not read at all depending on the directory a
229 Git command was run, leading to inconsistent behaviour. The code
230 to set-up repository access at the beginning of a Git process has
231 been updated to fix them.
232 (merge 4d0efa1 jk/setup-sequence-update later to maint).
234 * "git diff -W" output needs to extend the context backward to
235 include the header line of the current function and also forward to
236 include the body of the entire current function up to the header
237 line of the next one. This process may have to merge to adjacent
238 hunks, but the code forgot to do so in some cases.
240 * Performance tests done via "t/perf" did not use the same set of
241 build configuration if the user relied on autoconf generated
244 * "git format-patch --base=..." feature that was recently added
245 showed the base commit information after "-- " e-mail signature
246 line, which turned out to be inconvenient. The base information
247 has been moved above the signature line.
251 * Even when "git pull --rebase=preserve" (and the underlying "git
252 rebase --preserve") can complete without creating any new commit
253 (i.e. fast-forwards), it still insisted on having a usable ident
254 information (read: user.email is set correctly), which was less
255 than nice. As the underlying commands used inside "git rebase"
256 would fail with a more meaningful error message and advice text
257 when the bogus ident matters, this extra check was removed.
259 * "git gc --aggressive" used to limit the delta-chain length to 250,
260 which is way too deep for gaining additional space savings and is
261 detrimental for runtime performance. The limit has been reduced to
264 * Documentation for individual configuration variables to control use
265 of color (like `color.grep`) said that their default value is
266 'false', instead of saying their default is taken from `color.ui`.
267 When we updated the default value for color.ui from 'false' to
268 'auto' quite a while ago, all of them broke. This has been
271 * The pretty-format specifier "%C(auto)" used by the "log" family of
272 commands to enable coloring of the output is taught to also issue a
273 color-reset sequence to the output.
274 (merge 82b83da8d3 rs/c-auto-resets-attributes later to maint).
276 * A shell script example in check-ref-format documentation has been
279 * "git checkout <word>" does not follow the usual disambiguation
280 rules when the <word> can be both a rev and a path, to allow
281 checking out a branch 'foo' in a project that happens to have a
282 file 'foo' in the working tree without having to disambiguate.
283 This was poorly documented and the check was incorrect when the
284 command was run from a subdirectory.
286 * Some codepaths in "git diff" used regexec(3) on a buffer that was
287 mmap(2)ed, which may not have a terminating NUL, leading to a read
288 beyond the end of the mapped region. This was fixed by introducing
289 a regexec_buf() helper that takes a <ptr,len> pair with REG_STARTEND
291 (merge 842a516cb0 js/regexec-buf later to maint).
293 * The procedure to build Git on Mac OS X for Travis CI hardcoded the
294 internal directory structure we assumed HomeBrew uses, which was a
295 no-no. The procedure has been updated to ask HomeBrew things we
296 need to know to fix this.
298 * When "git rebase -i" is given a broken instruction, it told the
299 user to fix it with "--edit-todo", but didn't say what the step
300 after that was (i.e. "--continue").
302 * Documentation around tools to import from CVS was fairly outdated.
304 * "git clone --recurse-submodules" lost the progress eye-candy in
305 recent update, which has been corrected.
307 * A low-level function verify_packfile() was meant to show errors
308 that were detected without dying itself, but under some conditions
309 it didn't and died instead, which has been fixed.
311 * When "git fetch" tries to find where the history of the repository
312 it runs in has diverged from what the other side has, it has a
313 mechanism to avoid digging too deep into irrelevant side branches.
314 This however did not work well over the "smart-http" transport due
315 to a design bug, which has been fixed.
316 (merge 06b3d386e0 jt/fetch-pack-in-vain-count-with-stateless later to maint).
318 * In the codepath that comes up with the hostname to be used in an
319 e-mail when the user didn't tell us, we looked at ai_canonname
320 field in struct addrinfo without making sure it is not NULL first.
322 * "git worktree", even though it used the default_abbrev setting that
323 ought to be affected by core.abbrev configuration variable, ignored
324 the variable setting. The command has been taught to read the
325 default set of configuration variables to correct this.
327 * "git init" tried to record core.worktree in the repository's
328 'config' file when GIT_WORK_TREE environment variable was set and
329 it was different from where GIT_DIR appears as ".git" at its top,
330 but the logic was faulty when .git is a "gitdir:" file that points
331 at the real place, causing trouble in working trees that are
332 managed by "git worktree". This has been corrected.
334 * Codepaths that read from an on-disk loose object were too loose in
335 validating what they are reading is a proper object file and
336 sometimes read past the data they read from the disk, which has
337 been corrected. H/t to Gustavo Grieco for reporting.
339 * The original command line syntax for "git merge", which was "git
340 merge <msg> HEAD <parent>...", has been deprecated for quite some
341 time, and "git gui" was the last in-tree user of the syntax. This
342 is finally fixed, so that we can move forward with the deprecation.
344 * An author name, that spelled a backslash-quoted double quote in the
345 human readable part "My \"double quoted\" name", was not unquoted
346 correctly while applying a patch from a piece of e-mail.
348 * Doc update to clarify what "log -3 --reverse" does.
350 * Almost everybody uses DEFAULT_ABBREV to refer to the default
351 setting for the abbreviation, but "git blame" peeked into
352 underlying variable bypassing the macro for no good reason.
354 * The "graph" API used in "git log --graph" miscounted the number of
355 output columns consumed so far when drawing a padding line, which
356 has been fixed; this did not affect any existing code as nobody
357 tried to write anything after the padding on such a line, though.
359 * The code that parses the format parameter of for-each-ref command
360 has seen a micro-optimization.
362 * When we started cURL to talk to imap server when a new enough
363 version of cURL library is available, we forgot to explicitly add
364 imap(s):// before the destination. To some folks, that didn't work
365 and the library tried to make HTTP(s) requests instead.
366 (merge d2d07ab861 ak/curl-imap-send-explicit-scheme later to maint).
368 * The ./configure script generated from configure.ac was taught how
369 to detect support of SSL by libcurl better.
370 (merge 924b7eb1c9 dp/autoconf-curl-ssl later to maint).
372 * The command-line completion script (in contrib/) learned to
373 complete "git cmd ^mas<HT>" to complete the negative end of
374 reference to "git cmd ^master".
375 (merge 49416ad22a cp/completion-negative-refs later to maint).
377 * The existing "git fetch --depth=<n>" option was hard to use
378 correctly when making the history of an existing shallow clone
379 deeper. A new option, "--deepen=<n>", has been added to make this
380 easier to use. "git clone" also learned "--shallow-since=<date>"
381 and "--shallow-exclude=<tag>" options to make it easier to specify
382 "I am interested only in the recent N months worth of history" and
383 "Give me only the history since that version".
384 (merge cccf74e2da nd/shallow-deepen later to maint).
386 * It is a common mistake to say "git blame --reverse OLD path",
387 expecting that the command line is dwimmed as if asking how lines
388 in path in an old revision OLD have survived up to the current
390 (merge e1d09701a4 jc/blame-reverse later to maint).
392 * http.emptyauth configuration is a way to allow an empty username to
393 pass when attempting to authenticate using mechanisms like
394 Kerberos. We took an unspecified (NULL) username and sent ":"
395 (i.e. no username, no password) to CURLOPT_USERPWD, but did not do
396 the same when the username is explicitly set to an empty string.
397 (merge 5275c3081c dt/http-empty-auth later to maint).
399 * "git clone" of a local repository can be done at the filesystem
400 level, but the codepath did not check errors while copying and
401 adjusting the file that lists alternate object stores.
402 (merge 22d3b8de1b jk/clone-copy-alternates-fix later to maint).
404 * Documentation for "git commit" was updated to clarify that "commit
405 -p <paths>" adds to the current contents of the index to come up
407 (merge 7431596ab1 nd/commit-p-doc later to maint).
409 * A stray symbolic link in $GIT_DIR/refs/ directory could make name
410 resolution loop forever, which has been corrected.
411 (merge e8c42cb9ce jk/ref-symlink-loop later to maint).
413 * The "submodule.<name>.path" stored in .gitmodules is never copied
414 to .git/config and such a key in .git/config has no meaning, but
415 the documentation described it and submodule.<name>.url next to
416 each other as if both belong to .git/config. This has been fixed.
417 (merge 72710165c9 sb/submodule-config-doc-drop-path later to maint).
419 * Other minor doc, test and build updates and code cleanups.
420 (merge a94bb68397 rs/cocci later to maint).
421 (merge 641c900b2c js/reset-usage later to maint).
422 (merge 30cfe72d37 rs/pretty-format-color-doc-fix later to maint).
projects / git / blob