4 Backward compatibility notes
5 ----------------------------
7 The end-user facing Porcelain level commands in the "git diff" and
8 "git log" family by default enable the rename detection; you can still
9 use "diff.renames" configuration variable to disable this.
11 Merging two branches that have no common ancestor with "git merge" is
12 by default forbidden now to prevent creating such an unusual merge by
15 The output formats of "git log" that indents the commit log message by
16 4 spaces now expands HT in the log message by default. You can use
17 the "--no-expand-tabs" option to disable this.
19 "git commit-tree" plumbing command required the user to always sign
20 its result when the user sets the commit.gpgsign configuration
21 variable, which was an ancient mistake, which this release corrects.
22 A script that drives commit-tree, if it relies on this mistake, now
23 needs to read commit.gpgsign and pass the -S option as necessary.
29 UI, Workflows & Features
31 * Comes with git-multimail 1.3.1 (in contrib/).
33 * The end-user facing commands like "git diff" and "git log"
34 now enable the rename detection by default.
36 * The credential.helper configuration variable is cumulative and
37 there is no good way to override it from the command line. As
38 a special case, giving an empty string as its value now serves
39 as the signal to clear the values specified in various files.
41 * A new "interactive.diffFilter" configuration can be used to
42 customize the diff shown in "git add -i" sessions.
44 * "git p4" now allows P4 author names to be mapped to Git author
47 * "git rebase -x" can be used without passing "-i" option.
49 * "git -c credential.<var>=<value> submodule" can now be used to
50 propagate configuration variables related to credential helper
51 down to the submodules.
53 * "git tag" can create an annotated tag without explicitly given an
54 "-a" (or "-s") option (i.e. when a tag message is given). A new
55 configuration variable, tag.forceSignAnnotated, can be used to tell
56 the command to create signed tag in such a situation.
58 * "git merge" used to allow merging two branches that have no common
59 base by default, which led to a brand new history of an existing
60 project created and then get pulled by an unsuspecting maintainer,
61 which allowed an unnecessary parallel history merged into the
62 existing project. The command has been taught not to allow this by
63 default, with an escape hatch "--allow-unrelated-histories" option
64 to be used in a rare event that merges histories of two projects
65 that started their lives independently.
67 * "git pull" has been taught to pass the "--allow-unrelated-histories"
68 option to underlying "git merge".
70 * "git apply -v" learned to report paths in the patch that were
71 skipped via --include/--exclude mechanism or being outside the
72 current working directory.
74 * Shell completion (in contrib/) updates.
76 * The commit object name reported when "rebase -i" stops has been
79 * "git worktree add" can be given "--no-checkout" option to only
80 create an empty worktree without checking out the files.
82 * "git mergetools" learned to drive ExamDiff.
84 * "git pull --rebase" learned "--[no-]autostash" option, so that
85 the rebase.autostash configuration variable set to true can be
86 overridden from the command line.
88 * When "git log" shows the log message indented by 4-spaces, the
89 remainder of a line after a HT does not align in the way the author
90 originally intended. The command now expands tabs by default to help
91 such a case, and allows the users to override it with a new option,
94 * "git send-email" now uses a more readable timestamps when
95 formulating a message ID.
97 * "git rerere" can encounter two or more files with the same conflict
98 signature that have to be resolved in different ways, but there was
99 no way to record these separate resolutions.
101 * "git p4" learned to record P4 jobs in Git commit that imports from
102 the history in Perforce.
104 * "git describe --contains" often made a hard-to-justify choice of
105 tag to name a given commit, because it tried to come up
106 with a name with smallest number of hops from a tag, causing an old
107 commit whose close descendant that is recently tagged were not
108 described with respect to an old tag but with a newer tag. It did
109 not help that its computation of "hop" count was further tweaked to
110 penalize being on a side branch of a merge. The logic has been
111 updated to favor using the tag with the oldest tagger date, which
112 is a lot easier to explain to the end users: "We describe a commit
113 in terms of the (chronologically) oldest tag that contains the
116 * "git clone" learned the "--shallow-submodules" option.
118 * HTTP transport clients learned to throw extra HTTP headers at the
119 server, specified via http.extraHeader configuration variable.
121 * The "--compaction-heuristic" option to "git diff" family of
122 commands enables a heuristic to make the patch output more readable
123 by using a blank line as a strong hint that the contents before and
124 after it belong to logically separate units. It is still
127 * A new configuration variable core.hooksPath allows customizing
128 where the hook directory is.
130 * An earlier addition of "sanitize_submodule_env" with 14111fc4 (git:
131 submodule honor -c credential.* from command line, 2016-02-29)
132 turned out to be a convoluted no-op; implement what it wanted to do
133 correctly, and stop filtering settings given via "git -c var=val".
135 * "git commit --dry-run" reported "No, no, you cannot commit." in one
136 case where "git commit" would have allowed you to commit, and this
137 improves it a little bit ("git commit --dry-run --short" still does
138 not give you the correct answer, for example). This is a stop-gap
139 measure in that "commit --short --dry-run" still gives an incorrect
142 * The experimental "multiple worktree" feature gains more safety to
143 forbid operations on a branch that is checked out or being actively
144 worked on elsewhere, by noticing that e.g. it is being rebased.
146 * "git format-patch" learned a new "--base" option to record what
147 (public, well-known) commit the original series was built on in
150 * "git commit" learned to pay attention to the "commit.verbose"
151 configuration variable and act as if the "--verbose" option
152 was given from the command line.
154 * Updated documentation gives hints to GMail users with two-factor
155 auth enabled that they need app-specific-password when using
158 * The manpage output of our documentation did not render well in
159 terminal; typeset literals in bold by default to make them stand
162 * The mark-up in the top-level README.md file has been updated to
163 typeset CLI command names differently from the body text.
166 Performance, Internal Implementation, Development Support etc.
168 * The embedded args argv-array in the child process is used to build
169 the command line to run pack-objects instead of using a separate
172 * A test for tags has been restructured so that more parts of it can
173 easily be run on a platform without a working GnuPG.
175 * The startup_info data, which records if we are working inside a
176 repository (among other things), are now uniformly available to Git
177 subcommand implementations, and Git avoids attempting to touch
178 references when we are not in a repository.
180 * The command line argument parser for "receive-pack" has been
181 rewritten to use parse-options.
183 * A major part of "git submodule update" has been ported to C to take
184 advantage of the recently added framework to run download tasks in
185 parallel. Other updates to "git submodule" that move pieces of
186 logic to C continues.
188 * Rename bunch of tests on "git clone" for better organization.
190 * The tests that involve running httpd leaked the system-wide
191 configuration in /etc/gitconfig to the tested environment.
193 * Build updates for MSVC.
195 * The repository set-up sequence has been streamlined (the biggest
196 change is that there is no longer git_config_early()), so that we
197 do not attempt to look into refs/* when we know we do not have a
200 * Code restructuring around the "refs" API to prepare for pluggable
203 * Sources to many test helper binaries and the generated helpers
204 have been moved to t/helper/ subdirectory to reduce clutter at the
205 top level of the tree.
207 * Unify internal logic between "git tag -v" and "git verify-tag"
208 commands by making one directly call into the other.
210 * "merge-recursive" strategy incorrectly checked if a path that is
211 involved in its internal merge exists in the working tree.
213 * The test scripts for "git p4" (but not "git p4" implementation
214 itself) has been updated so that they would work even on a system
215 where the installed version of Python is python 3.
217 * As nobody maintains our in-tree git.spec.in and distros use their
218 own spec file, we stopped pretending that we support "make rpm".
220 * Move from "unsigned char[20]" to "struct object_id" continues.
222 * The code for warning_errno/die_errno has been refactored and a new
223 error_errno() reporting helper is introduced.
224 (merge 1da045f nd/error-errno later to maint).
226 * Running tests with '-x' option to trace the individual command
227 executions is a useful way to debug test scripts, but some tests
228 that capture the standard error stream and check what the command
229 said can be broken with the trace output mixed in. When running
230 our tests under "bash", however, we can redirect the trace output
231 to another file descriptor to keep the standard error of programs
234 * t0040 had too many unnecessary repetitions in its test data. Teach
235 test-parse-options program so that a caller can tell what it
236 expects in its output, so that these repetitions can be cleaned up.
238 * Add perf test for "rebase -i".
240 * Common mistakes when writing gitlink: in our documentation are
241 found by "make check-docs".
243 * t9xxx series has been updated primarily for readability, while
244 fixing small bugs in it. A few scripted Porcelain commands have
245 also been updated to fix possible bugs around their use of
246 "test -z" and "test -n".
248 * CI test was taught to run git-svn tests.
250 * "git cat-file --batch-all" has been sped up, by taking advantage
251 of the fact that it does not have to read a list of objects, in two
254 * test updates to make it more readable and maintainable.
255 (merge e6273f4 es/t1500-modernize later to maint).
257 * "make DEVELOPER=1" worked as expected; setting DEVELOPER=1 in
259 (merge 51dd3e8 mm/makefile-developer-can-be-in-config-mak later to maint).
261 * The way how "submodule--helper list" signals unmatch error to its
262 callers has been updated.
264 * A bash-ism "local" has been removed from "git submodule" scripted
268 Also contains various documentation updates and code clean-ups.
274 Unless otherwise noted, all the fixes since v2.8 in the maintenance
275 track are contained in this release (see the maintenance releases'
278 * "git config --get-urlmatch", unlike other variants of the "git
279 config --get" family, did not signal error with its exit status
280 when there was no matching configuration.
282 * The "--local-env-vars" and "--resolve-git-dir" options of "git
283 rev-parse" failed to work outside a repository when the command's
284 option parsing was rewritten in 1.8.5 era.
286 * "git index-pack --keep[=<msg>] pack-$name.pack" simply did not work.
288 * Fetching of history by naming a commit object name directly didn't
289 work across remote-curl transport.
291 * A small memory leak in an error codepath has been plugged in xdiff
294 * strbuf_getwholeline() did not NUL-terminate the buffer on certain
295 corner cases in its error codepath.
297 * "git mergetool" did not work well with conflicts that both sides
300 * "git send-email" had trouble parsing alias file in mailrc format
301 when lines in it had trailing whitespaces on them.
303 * When "git merge --squash" stopped due to conflict, the concluding
304 "git commit" failed to read in the SQUASH_MSG that shows the log
305 messages from all the squashed commits.
307 * "git merge FETCH_HEAD" dereferenced NULL pointer when merging
308 nothing into an unborn history (which is arguably unusual usage,
309 which perhaps was the reason why nobody noticed it).
311 * When "git worktree" feature is in use, "git branch -d" allowed
312 deletion of a branch that is checked out in another worktree,
315 * When "git worktree" feature is in use, "git branch -m" renamed a
316 branch that is checked out in another worktree without adjusting
317 the HEAD symbolic ref for the worktree.
319 * "git diff -M" used to work better when two originally identical
320 files A and B got renamed to X/A and X/B by pairing A to X/A and B
321 to X/B, but this was broken in the 2.0 timeframe.
323 * "git send-pack --all <there>" was broken when its command line
324 option parsing was written in the 2.6 timeframe.
326 * "git format-patch --help" showed `-s` and `--no-patch` as if these
327 are valid options to the command. We already hide `--patch` option
328 from the documentation, because format-patch is about showing the
329 diff, and the documentation now hides these options as well.
331 * When running "git blame $path" with unnormalized data in the index
332 for the path, the data in the working tree was blamed, even though
333 "git add" would not have changed what is already in the index, due
334 to "safe crlf" that disables the line-end conversion. It has been
337 * A change back in version 2.7 to "git branch" broke display of a
338 symbolic ref in a non-standard place in the refs/ hierarchy (we
339 expect symbolic refs to appear in refs/remotes/*/HEAD to point at
340 the primary branch the remote has, and as .git/HEAD to point at the
341 branch we locally checked out).
343 * A partial rewrite of "git submodule" in the 2.7 timeframe changed
344 the way the gitdir: pointer in the submodules point at the real
345 repository location to use absolute paths by accident. This has
348 * "git commit" misbehaved in a few minor ways when an empty message
349 is given via -m '', all of which has been corrected.
351 * Support for CRAM-MD5 authentication method in "git imap-send" did
354 * Upcoming OpenSSL 1.1.0 will break compilation by updating a few API
355 elements we use in imap-send, which has been adjusted for the change.
357 * The socks5:// proxy support added back in 2.6.4 days was not aware
358 that socks5h:// proxies behave differently from socks5:// proxies.
360 * "git config" had a codepath that tried to pass a NULL to
361 printf("%s"), which nobody seems to have noticed.
363 * On Cygwin, object creation uses the "create a temporary and then
364 rename it to the final name" pattern, not "create a temporary,
365 hardlink it to the final name and then unlink the temporary"
368 This is necessary to use Git on Windows shared directories, and is
369 already enabled for the MinGW and plain Windows builds. It also
370 has been used in Cygwin packaged versions of Git for quite a while.
371 See https://lore.kernel.org/git/20160419091055.GF2345@dinwoodie.org/
373 * "merge-octopus" strategy did not ensure that the index is clean
376 * When "git merge" notices that the merge can be resolved purely at
377 the tree level (without having to merge blobs) and the resulting
378 tree happens to already exist in the object store, it forgot to
379 update the index, which left an inconsistent state that would
380 break later operations.
382 * "git submodule" reports the paths of submodules the command
383 recurses into, but these paths were incorrectly reported when
384 the command was not run from the root level of the superproject.
386 * The "user.useConfigOnly" configuration variable makes it an error
387 if users do not explicitly set user.name and user.email. However,
388 its check was not done early enough and allowed another error to
389 trigger, reporting that the default value we guessed from the
390 system setting was unusable. This was a suboptimal end-user
391 experience as we want the users to set user.name/user.email without
392 relying on the auto-detection at all.
394 * "git mv old new" did not adjust the path for a submodule that lives
395 as a subdirectory inside old/ directory correctly.
397 * "git replace -e" did not honour "core.editor" configuration.
399 * "git push" from a corrupt repository that attempts to push a large
400 number of refs deadlocked; the thread to relay rejection notices
401 for these ref updates blocked on writing them to the main thread,
402 after the main thread at the receiving end notices that the push
403 failed and decides not to read these notices and return a failure.
405 * mmap emulation on Windows has been optimized and work better without
406 consuming paging store when not needed.
408 * A question by "git send-email" to ask the identity of the sender
411 * UI consistency improvements for "git mergetool".
413 * "git rebase -m" could be asked to rebase an entire branch starting
414 from the root, but failed by assuming that there always is a parent
415 commit to the first commit on the branch.
417 * Fix a broken "p4 lfs" test.
419 * Recent update to Git LFS broke "git p4" by changing the output from
420 its "lfs pointer" subcommand.
422 * "git fetch" test t5510 was flaky while running a (forced) automagic
425 * Documentation updates to help contributors setting up Travis CI
426 test for their patches.
428 * Some multi-byte encoding can have a backslash byte as a later part
429 of one letter, which would confuse "highlight" filter used in
432 * "git commit-tree" plumbing command required the user to always sign
433 its result when the user sets the commit.gpgsign configuration
434 variable, which was an ancient mistake. Rework "git rebase" that
435 relied on this mistake so that it reads commit.gpgsign and pass (or
436 not pass) the -S option to "git commit-tree" to keep the end-user
437 expectation the same, while teaching "git commit-tree" to ignore
438 the configuration variable. This will stop requiring the users to
439 sign commit objects used internally as an implementation detail of
442 * "http.cookieFile" configuration variable clearly wants a pathname,
443 but we forgot to treat it as such by e.g. applying tilde expansion.
445 * Consolidate description of tilde-expansion that is done to
446 configuration variables that take pathname to a single place.
448 * Correct faulty recommendation to use "git submodule deinit ." when
449 de-initialising all submodules, which would result in a strange
450 error message in a pathological corner case.
452 * Many 'linkgit:<git documentation page>' references were broken,
453 which are all fixed with this.
455 * "git rerere" can get confused by conflict markers deliberately left
456 by the inner merge step, because they are indistinguishable from
457 the real conflict markers left by the outermost merge which are
458 what the end user and "rerere" need to look at. This was fixed by
459 making the conflict markers left by the inner merges a bit longer.
460 (merge 0f9fd5c jc/ll-merge-internal later to maint).
462 * CI test was taught to build documentation pages.
464 * "git fsck" learned to catch NUL byte in a commit object as
465 potential error and warn.
467 * Portability enhancement for "rebase -i" to help platforms whose
468 shell does not like "for i in <empty>" (which is not POSIX-kosher).
470 * On Windows, .git and optionally any files whose name starts with a
471 dot are now marked as hidden, with a core.hideDotFiles knob to
472 customize this behaviour.
474 * Documentation for "git merge --verify-signatures" has been updated
475 to clarify that the signature of only the commit at the tip is
476 verified. Also the phrasing used for signature and key validity is
477 adjusted to align with that used by OpenPGP.
479 * A couple of bugs around core.autocrlf have been fixed.
481 * Many commands normalize command line arguments from NFD to NFC
482 variant of UTF-8 on OSX, but commands in the "diff" family did
483 not, causing "git diff $path" to complain that no such path is
484 known to Git. They have been taught to do the normalization.
486 * "git difftool" learned to handle unmerged paths correctly in
489 * The "are we talking with TTY, doing an interactive session?"
490 detection has been updated to work better for "Git for Windows".
492 * We forgot to add "git log --decorate=auto" to documentation when we
493 added the feature back in v2.1.0 timeframe.
494 (merge 462cbb4 rj/log-decorate-auto later to maint).
496 * "git fast-import --export-marks" would overwrite the existing marks
497 file even when it makes a dump from its custom die routine.
498 Prevent it from doing so when we have an import-marks file but
499 haven't finished reading it.
500 (merge f4beed6 fc/fast-import-broken-marks-file later to maint).
502 * "git rebase -i", after it fails to auto-resolve the conflict, had
503 an unnecessary call to "git rerere" from its very early days, which
504 was spotted recently; the call has been removed.
505 (merge 7063693 js/rebase-i-dedup-call-to-rerere later to maint).
507 * Other minor clean-ups and documentation updates
508 (merge cd82b7a pa/cherry-pick-doc-typo later to maint).
509 (merge 2bb73ae rs/patch-id-use-skip-prefix later to maint).
510 (merge aa20cbc rs/apply-name-terminate later to maint).
511 (merge fe17fc0 jc/t2300-setup later to maint).
512 (merge e256eec jk/shell-portability later to maint).
projects / git / blob