4 Backward compatibility notes
5 ----------------------------
10 UI, Workflows & Features
12 * "git pull --rebase --verify-signature" learned to warn the user
13 that "--verify-signature" is a no-op when rebasing.
15 * An upstream project can make a recommendation to shallowly clone
16 some submodules in the .gitmodules file it ships.
18 * "git worktree add" learned that '-' can be used as a short-hand for
19 "@{-1}", the previous branch.
21 * Update the funcname definition to support css files.
23 * The completion script (in contrib/) learned to complete "git
26 * Messages that are generated by auto gc during "git push" on the
27 receiving end are now passed back to the sending end in such a way
28 that they are shown with "remote: " prefix to avoid confusing the
31 * "git add -i/-p" learned to honor diff.compactionHeuristic
32 experimental knob, so that the user can work on the same hunk split
35 * "upload-pack" allows a custom "git pack-objects" replacement when
36 responding to "fetch/clone" via the uploadpack.packObjectsHook.
37 (merge b738396 jk/upload-pack-hook later to maint).
39 * Teach format-patch and mailsplit (hence "am") how a line that
40 happens to begin with "From " in the e-mail message is quoted with
41 ">", so that these lines can be restored to their original shape.
42 (merge d9925d1 ew/mboxrd-format-am later to maint).
44 * "git repack" learned the "--keep-unreachable" option, which sends
45 loose unreachable objects to a pack instead of leaving them loose.
46 This helps heuristics based on the number of loose objects
48 (merge e26a8c4 jk/repack-keep-unreachable later to maint).
50 * "log --graph --format=" learned that "%>|(N)" specifies the width
51 relative to the terminal's left edge, not relative to the area to
52 draw text that is to the right of the ancestry-graph section. It
53 also now accepts negative N that means the column limit is relative
56 * A careless invocation of "git send-email directory/" after editing
57 0001-change.patch with an editor often ends up sending both
58 0001-change.patch and its backup file, 0001-change.patch~, causing
59 embarrassment and a minor confusion. Detect such an input and
60 offer to skip the backup files when sending the patches out.
61 (merge 531220b jc/send-email-skip-backup later to maint).
63 * "git submodule update" that drives many "git clone" could
64 eventually hit flaky servers/network conditions on one of the
65 submodules; the command learned to retry the attempt.
67 * The output coloring scheme learned two new attributes, italic and
68 strike, in addition to existing bold, reverse, etc.
70 * "git log" learns log.showSignature configuration variable, and a
71 command line option "--no-show-signature" to countermand it.
72 (merge fce04c3 mj/log-show-signature-conf later to maint).
74 * More markings of messages for i18n, with updates to various tests
75 to pass GETTEXT_POISON tests.
77 * "git archive" learned to handle files that are larger than 8GB and
78 commits far in the future than expressible by the traditional US-TAR
80 (merge 5caeeb8 jk/big-and-future-archive-tar later to maint).
82 * A new configuration variable core.sshCommand has been added to
83 specify what value for GIT_SSH_COMMAND to use per repository.
85 * "git worktree prune" protected worktrees that are marked as
86 "locked" by creating a file in a known location. "git worktree"
87 command learned a dedicated command pair to create and remove such
88 a file, so that the users do not have to do this with editor.
90 * A handful of "git svn" updates.
92 * "git push" learned to accept and pass extra options to the
93 receiving end so that hooks can read and react to them.
95 * "git status" learned to suggest "merge --abort" during a conflicted
96 merge, just like it already suggests "rebase --abort" during a
99 * "git jump" script (in contrib/) has been updated a bit.
100 (merge a91e692 jk/git-jump later to maint).
102 * "git push" and "git clone" learned to give better progress meters
103 to the end user who is waiting on the terminal.
105 * An entry "git log --decorate" for the tip of the current branch is
106 shown as "HEAD -> name" (where "name" is the name of the branch);
107 the arrow is now painted in the same color as "HEAD", not in the
110 * "git format-patch" learned format.from configuration variable to
111 specify the default settings for its "--from" option.
113 * "git am -3" calls "git merge-recursive" when it needs to fall back
114 to a three-way merge; this call has been turned into an internal
115 subroutine call instead of spawning a separate subprocess.
117 * The command line completion scripts (in contrib/) now knows about
118 "git branch --delete/--move [--remote]".
119 (merge 2703c22 vs/completion-branch-fully-spelled-d-m-r later to maint).
121 * "git rev-parse --git-path hooks/<hook>" learned to take
122 core.hooksPath configuration variable (introduced during 2.9 cycle)
124 (merge 9445b49 ab/hooks later to maint).
126 * "git log --show-signature" and other commands that display the
127 verification status of PGP signature now shows the longer key-id,
128 as 32-bit key-id is so last century.
131 Performance, Internal Implementation, Development Support etc.
133 * "git fast-import" learned the same performance trick to avoid
134 creating too small a packfile as "git fetch" and "git push" have,
135 using *.unpackLimit configuration.
137 * When "git daemon" is run without --[init-]timeout specified, a
138 connection from a client that silently goes offline can hang around
139 for a long time, wasting resources. The socket-level KEEPALIVE has
140 been enabled to allow the OS to notice such failed connections.
142 * "git upload-pack" command has been updated to use the parse-options
145 * The "git apply" standalone program is being libified; the first
146 step to move many state variables into a structure that can be
147 explicitly (re)initialized to make the machinery callable more
148 than once has been merged.
150 * HTTP transport gained an option to produce more detailed debugging
152 (merge 73e57aa ep/http-curl-trace later to maint).
154 * Instead of taking advantage of the fact that a struct string_list
155 that is allocated with all NULs happens to be the INIT_NODUP kind,
156 the users of string_list structures are taught to initialize them
157 explicitly as such, to document their behaviour better.
158 (merge 2721ce2 jk/string-list-static-init later to maint).
160 * HTTPd tests learned to show the server error log to help diagnosing
162 (merge 44f243d nd/test-lib-httpd-show-error-log-in-verbose later to maint).
164 * The ownership rule for the piece of memory that hold references to
165 be fetched in "git fetch" was screwy, which has been cleaned up.
167 * "git bisect" makes an internal call to "git diff-tree" when
168 bisection finds the culprit, but this call did not initialize the
169 data structure to pass to the diff-tree API correctly.
171 * Further preparatory clean-up for "worktree" feature continues.
172 (merge 0409e0b nd/worktree-cleanup-post-head-protection later to maint).
174 * Formats of the various data (and how to validate them) where we use
175 GPG signature have been documented.
177 * A new run-command API function pipe_command() is introduced to
178 sanely feed data to the standard input while capturing data from
179 the standard output and the standard error of an external process,
180 which is cumbersome to hand-roll correctly without deadlocking.
182 The codepath to sign data in a prepared buffer with GPG has been
183 updated to use this API to read from the status-fd to check for
184 errors (instead of relying on GPG's exit status).
185 (merge efee955 jk/gpg-interface-cleanup later to maint).
187 * Allow t/perf framework to use the features from the most recent
188 version of Git even when testing an older installed version.
190 * The commands in the "log/diff" family have had an FILE* pointer in the
191 data structure they pass around for a long time, but some codepaths
192 used to always write to the standard output. As a preparatory step
193 to make "git format-patch" available to the internal callers, these
194 codepaths have been updated to consistently write into that FILE*
197 * Conversion from unsigned char sha1[20] to struct object_id
200 * Improve the look of the way "git fetch" reports what happened to
201 each ref that was fetched.
203 * The .c/.h sources are marked as such in our .gitattributes file so
204 that "git diff -W" and friends would work better.
206 * Code clean-up to avoid using a variable string that compilers may
207 feel untrustable as printf-style format given to write_file()
210 * "git p4" used a location outside $GIT_DIR/refs/ to place its
211 temporary branches, which has been moved to refs/git-p4-tmp/.
213 * Existing autoconf generated test for the need to link with pthread
214 library did not check all the functions from pthread libraries;
215 recent FreeBSD has some functions in libc but not others, and we
216 mistakenly thought linking with libc is enough when it is not.
218 * When "git fsck" reports a broken link (e.g. a tree object contains
219 a blob that does not exist), both containing object and the object
220 that is referred to were reported with their 40-hex object names.
221 The command learned the "--name-objects" option to show the path to
222 the containing object from existing refs (e.g. "HEAD~24^2:file.txt").
224 * Allow http daemon tests in Travis CI tests.
226 * Makefile assumed that -lrt is always available on platforms that
227 want to use clock_gettime() and CLOCK_MONOTONIC, which is not a
228 case for recent Mac OS X. The necessary symbols are often found in
229 libc on many modern systems and having -lrt on the command line, as
230 long as the library exists, had no effect, but when the platform
231 removes librt.a that is a different matter--having -lrt will break
234 This change could be seen as a regression for those who do need to
235 specify -lrt, as they now specifically ask for NEEDS_LIBRT when
236 building. Hopefully they are in the minority these days.
238 * Further preparatory work on the refs API before the pluggable
239 backend series can land.
241 * Error handling in the codepaths that updates refs has been
244 * The API to iterate over all the refs (i.e. for_each_ref(), etc.)
247 * The handling of the "text=auto" attribute has been corrected.
248 $ echo "* text=auto eol=crlf" >.gitattributes
249 used to have the same effect as
250 $ echo "* text eol=crlf" >.gitattributes
251 i.e. declaring all files are text (ignoring "auto"). The
252 combination has been fixed to be equivalent to doing
253 $ git config core.autocrlf true
255 * A few tests that specifically target "git rebase -i" have been
258 * Dumb http transport on the client side has been optimized.
259 (merge ecba195 ew/http-walker later to maint).
261 * Users of the parse_options_concat() API function need to allocate
262 extra slots in advance and fill them with OPT_END() when they want
263 to decide the set of supported options dynamically, which makes the
264 code error-prone and hard to read. This has been corrected by tweaking
265 the API to allocate and return a new copy of "struct option" array.
267 * "git fetch" exchanges batched have/ack messages between the sender
268 and the receiver, initially doubling every time and then falling
269 back to enlarge the window size linearly. The "smart http"
270 transport, being an half-duplex protocol, outgrows the preset limit
271 too quickly and becomes inefficient when interacting with a large
272 repository. The internal mechanism learned to grow the window size
273 more aggressively when working with the "smart http" transport.
275 * Tests for "git svn" have been taught to reuse the lib-httpd test
276 infrastructure when testing the subversion integration that
277 interacts with subversion repositories served over the http://
279 (merge a8a5d25 ew/git-svn-http-tests later to maint).
281 * "git pack-objects" has a few options that tell it not to pack
282 objects found in certain packfiles, which require it to scan .idx
283 files of all available packs. The codepaths involved in these
284 operations have been optimized for a common case of not having any
285 non-local pack and/or any .kept pack.
287 * The t3700 test about "add --chmod=-x" have been made a bit more
288 robust and generally cleaned up.
289 (merge 766cdc4 ib/t3700-add-chmod-x-updates later to maint).
291 * The build procedure learned PAGER_ENV knob that lists what default
292 environment variable settings to export for popular pagers. This
293 mechanism is used to tweak the default settings to MORE on FreeBSD.
294 (merge 995bc22 ew/build-time-pager-tweaks later to maint).
296 * The http-backend (the server-side component of smart-http
297 transport) used to trickle the HTTP header one at a time. Now
298 these write(2)s are batched.
299 (merge b36045c ew/http-backend-batch-headers later to maint).
301 * When "git rebase" tries to compare set of changes on the updated
302 upstream and our own branch, it computes patch-id for all of these
303 changes and attempts to find matches. This has been optimized by
304 lazily computing the full patch-id (which is expensive) to be
305 compared only for changes that touch the same set of paths.
306 (merge b3dfeeb kw/patch-ids-optim later to maint).
308 * A handful of tests that were broken under gettext-poison build have
312 Also contains various documentation updates and code clean-ups.
318 Unless otherwise noted, all the fixes since v2.8 in the maintenance
319 track are contained in this release (see the maintenance releases'
322 * The commands in `git log` family take %C(auto) in a custom format
323 string. This unconditionally turned the color on, ignoring
324 --no-color or with --color=auto when the output is not connected to
325 a tty; this was corrected to make the format truly behave as
328 * "git rev-list --count" whose walk-length is limited with "-n"
329 option did not work well with the counting optimized to look at the
332 * "git show -W" (extend hunks to cover the entire function, delimited
333 by lines that match the "funcname" pattern) used to show the entire
334 file when a change added an entire function at the end of the file,
335 which has been fixed.
337 * The documentation set has been updated so that literal commands,
338 configuration variables and environment variables are consistently
339 typeset in fixed-width font and bold in manpages.
341 * "git svn propset" subcommand that was added in 2.3 days is
344 * The documentation tries to consistently spell "GPG"; when
345 referring to the specific program name, "gpg" is used.
347 * "git reflog" stopped upon seeing an entry that denotes a branch
348 creation event (aka "unborn"), which made it appear as if the
349 reflog was truncated.
351 * The git-prompt scriptlet (in contrib/) was not friendly with those
352 who uses "set -u", which has been fixed.
354 * compat/regex code did not cleanly compile.
356 * A codepath that used alloca(3) to place an unbounded amount of data
357 on the stack has been updated to avoid doing so.
359 * "git update-index --add --chmod=+x file" may be usable as an escape
360 hatch, but not a friendly thing to force for people who do need to
361 use it regularly. "git add --chmod=+x file" can be used instead.
363 * Build improvements for gnome-keyring (in contrib/)
365 * "git status" used to say "working directory" when it meant "working
368 * Comments about misbehaving FreeBSD shells have been clarified with
369 the version number (9.x and before are broken, newer ones are OK).
371 * "git cherry-pick A" worked on an unborn branch, but "git
372 cherry-pick A..B" didn't.
374 * Fix an unintended regression in v2.9 that breaks "clone --depth"
375 that recurses down to submodules by forcing the submodules to also
376 be cloned shallowly, which many server instances that host upstream
377 of the submodules are not prepared for.
379 * Fix unnecessarily waste in the idiomatic use of ': ${VAR=default}'
380 to set the default value, without enclosing it in double quotes.
382 * Some platform-specific code had non-ANSI strict declarations of C
383 functions that do not take any parameters, which has been
386 * The internal code used to show local timezone offset is not
387 prepared to handle timestamps beyond year 2100, and gave a
388 bogus offset value to the caller. Use a more benign looking
389 +0000 instead and let "git log" going in such a case, instead
392 * One among four invocations of readlink(1) in our test suite has
393 been rewritten so that the test can run on systems without the
394 command (others are in valgrind test framework and t9802).
396 * t/perf needs /usr/bin/time with GNU extension; the invocation of it
397 is updated to "gtime" on Darwin.
399 * A bug, which caused "git p4" while running under verbose mode to
400 report paths that are omitted due to branch prefix incorrectly, has
401 been fixed; the command said "Ignoring file outside of prefix" for
402 paths that are _inside_.
404 * The top level documentation "git help git" still pointed at the
405 documentation set hosted at now-defunct google-code repository.
406 Update it to point to https://git.github.io/htmldocs/git.html
409 * A helper function that takes the contents of a commit object and
410 finds its subject line did not ignore leading blank lines, as is
411 commonly done by other codepaths. Make it ignore leading blank
414 * For a long time, we carried an in-code comment that said our
415 colored output would work only when we use fprintf/fputs on
416 Windows, which no longer is the case for the past few years.
418 * "gc.autoPackLimit" when set to 1 should not trigger a repacking
419 when there is only one pack, but the code counted poorly and did
422 * Add a test to specify the desired behaviour that currently is not
423 available in "git rebase -Xsubtree=...".
425 * More mark-up updates to typeset strings that are expected to
426 literally typed by the end user in fixed-width font.
428 * "git commit --amend --allow-empty-message -S" for a commit without
429 any message body could have misidentified where the header of the
432 * "git rebase -i --autostash" did not restore the auto-stashed change
433 when the operation was aborted.
435 * Git does not know what the contents in the index should be for a
436 path added with "git add -N" yet, so "git grep --cached" should not
437 show hits (or show lack of hits, with -L) in such a path, but that
438 logic does not apply to "git grep", i.e. searching in the working
439 tree files. But we did so by mistake, which has been corrected.
441 * "git blame -M" missed a single line that was moved within the file.
443 * Fix recently introduced codepaths that are involved in parallel
444 submodule operations, which gave up on reading too early, and
445 could have wasted CPU while attempting to write under a corner
448 * "git grep -i" has been taught to fold case in non-ascii locales
451 * A test that unconditionally used "mktemp" learned that the command
452 is not necessarily available everywhere.
454 * There are certain house-keeping tasks that need to be performed at
455 the very beginning of any Git program, and programs that are not
456 built-in commands had to do them exactly the same way as "git"
457 potty does. It was easy to make mistakes in one-off standalone
458 programs (like test helpers). A common "main()" function that
459 calls cmd_main() of individual program has been introduced to
460 make it harder to make mistakes.
461 (merge de61ceb jk/common-main later to maint).
463 * The test framework learned a new helper test_match_signal to
464 check an exit code from getting killed by an expected signal.
466 * General code clean-up around a helper function to write a
467 single-liner to a file.
468 (merge 7eb6e10 jk/write-file later to maint).
470 * One part of "git am" had an oddball helper function that called
471 stuff from outside "his" as opposed to calling what we have "ours",
472 which was not gender-neutral and also inconsistent with the rest of
473 the system where outside stuff is usuall called "theirs" in
476 * "git blame file" allowed the lineage of lines in the uncommitted,
477 unadded contents of "file" to be inspected, but it refused when
478 "file" did not appear in the current commit. When "file" was
479 created by renaming an existing file (but the change has not been
480 committed), this restriction was unnecessarily tight.
482 * "git add -N dir/file && git write-tree" produced an incorrect tree
483 when there are other paths in the same directory that sorts after
486 * "git fetch http://user:pass@host/repo..." scrubbed the userinfo
487 part, but "git push" didn't.
489 * "git merge" with renormalization did not work well with
490 merge-recursive, due to "safer crlf" conversion kicking in when it
492 (merge 1335d76 jc/renormalize-merge-kill-safer-crlf later to maint).
494 * The use of strbuf in "git rm" to build filename to remove was a bit
495 suboptimal, which has been fixed.
497 * An age old bug that caused "git diff --ignore-space-at-eol"
498 misbehave has been fixed.
500 * "git notes merge" had a code to see if a path exists (and fails if
501 it does) and then open the path for writing (when it doesn't).
502 Replace it with open with O_EXCL.
504 * "git pack-objects" and "git index-pack" mostly operate with off_t
505 when talking about the offset of objects in a packfile, but there
506 were a handful of places that used "unsigned long" to hold that
507 value, leading to an unintended truncation.
509 * Recent update to "git daemon" tries to enable the socket-level
510 KEEPALIVE, but when it is spawned via inetd, the standard input
511 file descriptor may not necessarily be connected to a socket.
512 Suppress an ENOTSOCK error from setsockopt().
514 * Recent FreeBSD stopped making perl available at /usr/bin/perl;
515 switch the default the built-in path to /usr/local/bin/perl on not
516 too ancient FreeBSD releases.
518 * "git commit --help" said "--no-verify" is only about skipping the
519 pre-commit hook, and failed to say that it also skipped the
522 * "git merge" in Git v2.9 was taught to forbid merging an unrelated
523 lines of history by default, but that is exactly the kind of thing
524 the "--rejoin" mode of "git subtree" (in contrib/) wants to do.
525 "git subtree" has been taught to use the "--allow-unrelated-histories"
526 option to override the default.
528 * The build procedure for "git persistent-https" helper (in contrib/)
529 has been updated so that it can be built with more recent versions
532 * There is an optimization used in "git diff $treeA $treeB" to borrow
533 an already checked-out copy in the working tree when it is known to
534 be the same as the blob being compared, expecting that open/mmap of
535 such a file is faster than reading it from the object store, which
536 involves inflating and applying delta. This however kicked in even
537 when the checked-out copy needs to go through the convert-to-git
538 conversion (including the clean filter), which defeats the whole
539 point of the optimization. The optimization has been disabled when
540 the conversion is necessary.
542 * "git -c grep.patternType=extended log --basic-regexp" misbehaved
543 because the internal API to access the grep machinery was not
546 * Windows port was failing some tests in t4130, due to the lack of
547 inum in the returned values by its lstat(2) emulation.
549 * The reflog output format is documented better, and a new format
550 --date=unix to report the seconds-since-epoch (without timezone)
552 (merge 442f6fd jk/reflog-date later to maint).
554 * "git difftool <paths>..." started in a subdirectory failed to
555 interpret the paths relative to that directory, which has been
558 * The characters in the label shown for tags/refs for commits in
559 "gitweb" output are now properly escaped for proper HTML output.
561 * FreeBSD can lie when asked mtime of a directory, which made the
562 untracked cache code to fall back to a slow-path, which in turn
563 caused tests in t7063 to fail because it wanted to verify the
564 behaviour of the fast-path.
566 * Squelch compiler warnings for netmalloc (in compat/) library.
568 * A small memory leak in the command line parsing of "git blame"
571 * The API documentation for hashmap was unclear if hashmap_entry
572 can be safely discarded without any other consideration. State
573 that it is safe to do so.
575 * Not-so-recent rewrite of "git am" that started making internal
576 calls into the commit machinery had an unintended regression, in
577 that no matter how many seconds it took to apply many patches, the
578 resulting committer timestamp for the resulting commits were all
581 * "git push --force-with-lease" already had enough logic to allow
582 ensuring that such a push results in creation of a ref (i.e. the
583 receiving end did not have another push from sideways that would be
584 discarded by our force-pushing), but didn't expose this possibility
585 to the users. It does so now.
586 (merge 9eed4f3 jk/push-force-with-lease-creation later to maint).
588 * The mechanism to limit the pack window memory size, when packing is
589 done using multiple threads (which is the default), is per-thread,
590 but this was not documented clearly.
591 (merge 954176c ms/document-pack-window-memory-is-per-thread later to maint).
593 * "import-tars" fast-import script (in contrib/) used to ignore a
594 hardlink target and replaced it with an empty file, which has been
595 corrected to record the same blob as the other file the hardlink is
597 (merge 04e0869 js/import-tars-hardlinks later to maint).
599 * "git mv dir non-existing-dir/" did not work in some environments
600 the same way as existing mainstream platforms. The code now moves
601 "dir" to "non-existing-dir", without relying on rename("A", "B/")
602 that strips the trailing slash of '/'.
603 (merge 189d035 js/mv-dir-to-new-directory later to maint).
605 * The "t/" hierarchy is prone to get an unusual pathname; "make test"
606 has been taught to make sure they do not contain paths that cannot
607 be checked out on Windows (and the mechanism can be reusable to
608 catch pathnames that are not portable to other platforms as need
610 (merge c2cafd3 js/test-lint-pathname later to maint).
612 * When "git merge-recursive" works on history with many criss-cross
613 merges in "verbose" mode, the names the command assigns to the
614 virtual merge bases could have overwritten each other by unintended
615 reuse of the same piece of memory.
616 (merge 5447a76 rs/pull-signed-tag later to maint).
618 * "git checkout --detach <branch>" used to give the same advice
619 message as that is issued when "git checkout <tag>" (or anything
620 that is not a branch name) is given, but asking with "--detach" is
621 an explicit enough sign that the user knows what is going on. The
622 advice message has been squelched in this case.
623 (merge 779b88a sb/checkout-explit-detach-no-advice later to maint).
625 * "git difftool" by default ignores the error exit from the backend
626 commands it spawns, because often they signal that they found
627 differences by exiting with a non-zero status code just like "diff"
628 does; the exit status codes 126 and above however are special in
629 that they are used to signal that the command is not executable,
630 does not exist, or killed by a signal. "git difftool" has been
631 taught to notice these exit status codes.
632 (merge 45a4f5d jk/difftool-command-not-found later to maint).
634 * Other minor clean-ups and documentation updates
635 (merge 02a8cfa rs/merge-add-strategies-simplification later to maint).
636 (merge af4941d rs/merge-recursive-string-list-init later to maint).
637 (merge 1eb47f1 rs/use-strbuf-add-unique-abbrev later to maint).
638 (merge ddd0bfa jk/tighten-alloc later to maint).
639 (merge ecf30b2 rs/mailinfo-lib later to maint).
640 (merge 0eb75ce sg/reflog-past-root later to maint).