| blob | 93165bd2c33e4bf9b25feceadd64f9c363c5e686 |
1 Git 2.12 Release Notes
2 ======================
4 Backward compatibility notes.
6 * Use of an empty string that is used for 'everything matches' is
7 still warned and Git asks users to use a more explicit '.' for that
8 instead. The hope is that existing users will not mind this
9 change, and eventually the warning can be turned into a hard error,
10 upgrading the deprecation into removal of this (mis)feature. That
11 is not scheduled to happen in the upcoming release (yet).
13 * The historical argument order "git merge <msg> HEAD <commit>..."
14 has been deprecated for quite some time, and will be removed in a
15 future release.
17 * An ancient script "git relink" has been removed.
20 Updates since v2.11
21 -------------------
23 UI, Workflows & Features
25 * Various updates to "git p4".
27 * "git p4" didn't interact with the internal of .git directory
28 correctly in the modern "git-worktree"-enabled world.
30 * "git branch --list" and friends learned "--ignore-case" option to
31 optionally sort branches and tags case insensitively.
33 * In addition to %(subject), %(body), "log --pretty=format:..."
34 learned a new placeholder %(trailers).
36 * "git rebase" learned "--quit" option, which allows a user to
37 remove the metadata left by an earlier "git rebase" that was
38 manually aborted without using "git rebase --abort".
40 * "git clone --reference $there --recurse-submodules $super" has been
41 taught to guess repositories usable as references for submodules of
42 $super that are embedded in $there while making a clone of the
43 superproject borrow objects from $there; extend the mechanism to
44 also allow submodules of these submodules to borrow repositories
45 embedded in these clones of the submodules embedded in the clone of
46 the superproject.
48 * Porcelain scripts written in Perl are getting internationalized.
50 * "git merge --continue" has been added as a synonym to "git commit"
51 to conclude a merge that has stopped due to conflicts.
53 * Finer-grained control of what protocols are allowed for transports
54 during clone/fetch/push have been enabled via a new configuration
55 mechanism.
57 * "git shortlog" learned "--committer" option to group commits by
58 committer, instead of author.
60 * GitLFS integration with "git p4" has been updated.
62 * The isatty() emulation for Windows has been updated to eradicate
63 the previous hack that depended on internals of (older) MSVC
64 runtime.
66 * Some platforms no longer understand "latin-1" that is still seen in
67 the wild in e-mail headers; replace them with "iso-8859-1" that is
68 more widely known when conversion fails from/to it.
70 * "git grep" has been taught to optionally recurse into submodules.
72 * "git rm" used to refuse to remove a submodule when it has its own
73 git repository embedded in its working tree. It learned to move
74 the repository away to $GIT_DIR/modules/ of the superproject
75 instead, and allow the submodule to be deleted (as long as there
76 will be no loss of local modifications, that is).
78 * A recent updates to "git p4" was not usable for older p4 but it
79 could be made to work with minimum changes. Do so.
81 * "git diff" learned diff.interHunkContext configuration variable
82 that gives the default value for its --inter-hunk-context option.
84 * The prereleaseSuffix feature of version comparison that is used in
85 "git tag -l" did not correctly when two or more prereleases for the
86 same release were present (e.g. when 2.0, 2.0-beta1, and 2.0-beta2
87 are there and the code needs to compare 2.0-beta1 and 2.0-beta2).
89 * "git submodule push" learned "--recurse-submodules=only option to
90 push submodules out without pushing the top-level superproject.
92 * "git tag" and "git verify-tag" learned to put GPG verification
93 status in their "--format=<placeholders>" output format.
95 * An ancient repository conversion tool left in contrib/ has been
96 removed.
98 * "git show-ref HEAD" used with "--verify" because the user is not
99 interested in seeing refs/remotes/origin/HEAD, and used with
100 "--head" because the user does not want HEAD to be filtered out,
101 i.e. "git show-ref --head --verify HEAD", did not work as expected.
103 * "git submodule add" used to be confused and refused to add a
104 locally created repository; users can now use "--force" option
105 to add them.
106 (merge 619acfc78c sb/submodule-add-force later to maint).
108 * Some people feel the default set of colors used by "git log --graph"
109 rather limiting. A mechanism to customize the set of colors has
110 been introduced.
112 * "git read-tree" and its underlying unpack_trees() machinery learned
113 to report problematic paths prefixed with the --super-prefix option.
115 * When a submodule "A", which has another submodule "B" nested within
116 it, is "absorbed" into the top-level superproject, the inner
117 submodule "B" used to be left in a strange state. The logic to
118 adjust the .git pointers in these submodules has been corrected.
120 * The user can specify a custom update method that is run when
121 "submodule update" updates an already checked out submodule. This
122 was ignored when checking the submodule out for the first time and
123 we instead always just checked out the commit that is bound to the
124 path in the superproject's index.
126 * The command line completion (in contrib/) learned that
127 "git diff --submodule=" can take "diff" as a recently added option.
129 * The "core.logAllRefUpdates" that used to be boolean has been
130 enhanced to take 'always' as well, to record ref updates to refs
131 other than the ones that are expected to be updated (i.e. branches,
132 remote-tracking branches and notes).
134 * Comes with more command line completion (in contrib/) for recently
135 introduced options.
138 Performance, Internal Implementation, Development Support etc.
140 * Commands that operate on a log message and add lines to the trailer
141 blocks, such as "format-patch -s", "cherry-pick (-x|-s)", and
142 "commit -s", have been taught to use the logic of and share the
143 code with "git interpret-trailer".
145 * The default Travis-CI configuration specifies newer P4 and GitLFS.
147 * The "fast hash" that had disastrous performance issues in some
148 corner cases has been retired from the internal diff.
150 * The character width table has been updated to match Unicode 9.0
152 * Update the procedure to generate "tags" for developer support.
154 * The codeflow of setting NOATIME and CLOEXEC on file descriptors Git
155 opens has been simplified.
157 * "git diff" and its family had two experimental heuristics to shift
158 the contents of a hunk to make the patch easier to read. One of
159 them turns out to be better than the other, so leave only the
160 "--indent-heuristic" option and remove the other one.
162 * A new submodule helper "git submodule embedgitdirs" to make it
163 easier to move embedded .git/ directory for submodules in a
164 superproject to .git/modules/ (and point the latter with the former
165 that is turned into a "gitdir:" file) has been added.
167 * "git push \\server\share\dir" has recently regressed and then
168 fixed. A test has retroactively been added for this breakage.
170 * Build updates for Cygwin.
172 * The implementation of "real_path()" was to go there with chdir(2)
173 and call getcwd(3), but this obviously wouldn't be usable in a
174 threaded environment. Rewrite it to manually resolve relative
175 paths including symbolic links in path components.
177 * Adjust documentation to help AsciiDoctor render better while not
178 breaking the rendering done by AsciiDoc.
180 * The sequencer machinery has been further enhanced so that a later
181 set of patches can start using it to reimplement "rebase -i".
183 * Update the definition of the MacOSX test environment used by
184 TravisCI.
186 * Rewrite a scripted porcelain "git difftool" in C.
188 * "make -C t failed" will now run only the tests that failed in the
189 previous run. This is usable only when prove is not use, and gives
190 a useless error message when run after "make clean", but otherwise
191 is serviceable.
193 * "uchar [40]" to "struct object_id" conversion continues.
196 Also contains various documentation updates and code clean-ups.
198 Fixes since v2.10
199 -----------------
201 Unless otherwise noted, all the fixes since v2.9 in the maintenance
202 track are contained in this release (see the maintenance releases'
203 notes for details).
205 * We often decide if a session is interactive by checking if the
206 standard I/O streams are connected to a TTY, but isatty() that
207 comes with Windows incorrectly returned true if it is used on NUL
208 (i.e. an equivalent to /dev/null). This has been fixed.
210 * "git svn" did not work well with path components that are "0", and
211 some configuration variable it uses were not documented.
213 * "git rev-parse --symbolic" failed with a more recent notation like
214 "HEAD^-1" and "HEAD^!".
216 * An empty directory in a working tree that can simply be nuked used
217 to interfere while merging or cherry-picking a change to create a
218 submodule directory there, which has been fixed..
220 * The code in "git push" to compute if any commit being pushed in the
221 superproject binds a commit in a submodule that hasn't been pushed
222 out was overly inefficient, making it unusable even for a small
223 project that does not have any submodule but have a reasonable
224 number of refs.
226 * "git push --dry-run --recurse-submodule=on-demand" wasn't
227 "--dry-run" in the submodules.
229 * The output from "git worktree list" was made in readdir() order,
230 and was unstable.
232 * mergetool.<tool>.trustExitCode configuration variable did not apply
233 to built-in tools, but now it does.
235 * "git p4" LFS support was broken when LFS stores an empty blob.
237 * A corner case in merge-recursive regression that crept in
238 during 2.10 development cycle has been fixed.
240 * Transport with dumb http can be fooled into following foreign URLs
241 that the end user does not intend to, especially with the server
242 side redirects and http-alternates mechanism, which can lead to
243 security issues. Tighten the redirection and make it more obvious
244 to the end user when it happens.
246 * Update the error messages from the dumb-http client when it fails
247 to obtain loose objects; we used to give sensible error message
248 only upon 404 but we now forbid unexpected redirects that needs to
249 be reported with something sensible.
251 * When diff.renames configuration is on (and with Git 2.9 and later,
252 it is enabled by default, which made it worse), "git stash"
253 misbehaved if a file is removed and another file with a very
254 similar content is added.
256 * "git diff --no-index" did not take "--no-abbrev" option.
258 * "git difftool --dir-diff" had a minor regression when started from
259 a subdirectory, which has been fixed.
261 * "git commit --allow-empty --only" (no pathspec) with dirty index
262 ought to be an acceptable way to create a new commit that does not
263 change any paths, but it was forbidden, perhaps because nobody
264 needed it so far.
266 * Git 2.11 had a minor regression in "merge --ff-only" that competed
267 with another process that simultanously attempted to update the
268 index. We used to explain what went wrong with an error message,
269 but the new code silently failed. The error message has been
270 resurrected.
272 * A pathname that begins with "//" or "\\" on Windows is special but
273 path normalization logic was unaware of it.
275 * "git pull --rebase", when there is no new commits on our side since
276 we forked from the upstream, should be able to fast-forward without
277 invoking "git rebase", but it didn't.
279 * The way to specify hotkeys to "xxdiff" that is used by "git
280 mergetool" has been modernized to match recent versions of xxdiff.
282 * Unlike "git am --abort", "git cherry-pick --abort" moved HEAD back
283 to where cherry-pick started while picking multiple changes, when
284 the cherry-pick stopped to ask for help from the user, and the user
285 did "git reset --hard" to a different commit in order to re-attempt
286 the operation.
288 * Code cleanup in shallow boundary computation.
290 * A recent update to receive-pack to make it easier to drop garbage
291 objects made it clear that GIT_ALTERNATE_OBJECT_DIRECTORIES cannot
292 have a pathname with a colon in it (no surprise!), and this in turn
293 made it impossible to push into a repository at such a path. This
294 has been fixed by introducing a quoting mechanism used when
295 appending such a path to the colon-separated list.
297 * The function usage_msg_opt() has been updated to say "fatal:"
298 before the custom message programs give, when they want to die
299 with a message about wrong command line options followed by the
300 standard usage string.
302 * "git index-pack --stdin" needs an access to an existing repository,
303 but "git index-pack file.pack" to generate an .idx file that
304 corresponds to a packfile does not.
306 * Fix for NDEBUG builds.
308 * A lazy "git push" without refspec did not internally use a fully
309 specified refspec to perform 'current', 'simple', or 'upstream'
310 push, causing unnecessary "ambiguous ref" errors.
312 * "git p4" misbehaved when swapping a directory and a symbolic link.
314 * Even though an fix was attempted in Git 2.9.3 days, but running
315 "git difftool --dir-diff" from a subdirectory never worked. This
316 has been fixed.
318 * "git p4" that tracks multile p4 paths imported a single changelist
319 that touches files in these multiple paths as one commit, followed
320 by many empty commits. This has been fixed.
322 * A potential but unlikely buffer overflow in Windows port has been
323 fixed.
325 * When the http server gives an incomplete response to a smart-http
326 rpc call, it could lead to client waiting for a full response that
327 will never come. Teach the client side to notice this condition
328 and abort the transfer.
330 * Compression setting for producing packfiles were spread across
331 three codepaths, one of which did not honor any configuration.
332 Unify these so that all of them honor core.compression and
333 pack.compression variables the same way.
335 * "git fast-import" sometimes mishandled while rebalancing notes
336 tree, which has been fixed.
338 * Recent update to the default abbreviation length that auto-scales
339 lacked documentation update, which has been corrected.
341 * Leakage of lockfiles in the config subsystem has been fixed.
343 * It is natural that "git gc --auto" may not attempt to pack
344 everything into a single pack, and there is no point in warning
345 when the user has configured the system to use the pack bitmap,
346 leading to disabling further "gc".
348 * "git archive" did not read the standard configuration files, and
349 failed to notice a file that is marked as binary via the userdiff
350 driver configuration.
352 * "git blame --porcelain" misidentified the "previous" <commit, path>
353 pair (aka "source") when contents came from two or more files.
355 * "git rebase -i" with a recent update started showing an incorrect
356 count when squashing more than 10 commits.
358 * "git <cmd> @{push}" on a detached HEAD used to segfault; it has
359 been corrected to error out with a message.
361 * Running "git add a/b" when "a" is a submodule correctly errored
362 out, but without a meaningful error message.
363 (merge 2d81c48fa7 sb/pathspec-errors later to maint).
365 * Typing ^C to pager, which usually does not kill it, killed Git and
366 took the pager down as a collateral damage in certain process-tree
367 structure. This has been fixed.
369 * "git mergetool" without any pathspec on the command line that is
370 run from a subdirectory became no-op in Git v2.11 by mistake, which
371 has been fixed.
373 * Retire long unused/unmaintained gitview from the contrib/ area.
374 (merge 3120925c25 sb/remove-gitview later to maint).
376 * Tighten a test to avoid mistaking an extended ERE regexp engine as
377 a PRE regexp engine.
379 * An error message with an ASCII control character like '\r' in it
380 can alter the message to hide its early part, which is problematic
381 when a remote side gives such an error message that the local side
382 will relay with a "remote: " prefix.
383 (merge f290089879 jk/vreport-sanitize later to maint).
385 * "git fsck" inspects loose objects more carefully now.
386 (merge cce044df7f jk/loose-object-fsck later to maint).
388 * A crashing bug introduced in v2.11 timeframe has been found (it is
389 triggerable only in fast-import) and fixed.
390 (merge abd5a00268 jk/clear-delta-base-cache-fix later to maint).
392 * With an anticipatory tweak for remotes defined in ~/.gitconfig
393 (e.g. "remote.origin.prune" set to true, even though there may or
394 may not actually be "origin" remote defined in a particular Git
395 repository), "git remote rename" and other commands misinterpreted
396 and behaved as if such a non-existing remote actually existed.
397 (merge e459b073fb js/remote-rename-with-half-configured-remote later to maint).
399 * A few codepaths had to rely on a global variable when sorting
400 elements of an array because sort(3) API does not allow extra data
401 to be passed to the comparison function. Use qsort_s() when
402 natively available, and a fallback implementation of it when not,
403 to eliminate the need, which is a prerequisite for making the
404 codepath reentrant.
406 * "git fsck --connectivity-check" was not working at all.
407 (merge a2b22854bd jk/fsck-connectivity-check-fix later to maint).
409 * After starting "git rebase -i", which first opens the user's editor
410 to edit the series of patches to apply, but before saving the
411 contents of that file, "git status" failed to show the current
412 state (i.e. you are in an interactive rebase session, but you have
413 applied no steps yet) correctly.
414 (merge df9ded4984 js/status-pre-rebase-i later to maint).
416 * Test tweak for FreeBSD where /usr/bin/unzip is unsuitable to run
417 our tests but /usr/local/bin/unzip is usable.
418 (merge d98b2c5fce js/unzip-in-usr-bin-workaround later to maint).
420 * "git p4" did not work well with multiple git-p4.mapUser entries on
421 Windows.
422 (merge c3c2b05776 gv/mingw-p4-mapuser later to maint).
424 * "git help" enumerates executable files in $PATH; the implementation
425 of "is this file executable?" on Windows has been optimized.
426 (merge c755015f79 hv/mingw-help-is-executable later to maint).
428 * Test tweaks for those who have default ACL in their git source tree
429 that interfere with the umask test.
430 (merge d549d21307 mm/reset-facl-before-umask-test later to maint).
432 * Names of the various hook scripts must be spelled exactly, but on
433 Windows, an .exe binary must be named with .exe suffix; notice
434 $GIT_DIR/hooks/<hookname>.exe as a valid <hookname> hook.
435 (merge 235be51fbe js/mingw-hooks-with-exe-suffix later to maint).
437 * Asciidoctor, an alternative reimplementation of AsciiDoc, still
438 needs some changes to work with documents meant to be formatted
439 with AsciiDoc. "make USE_ASCIIDOCTOR=YesPlease" to use it out of
440 the box to document our pages is getting closer to reality.
442 * Correct command line completion (in contrib/) on "git svn"
443 (merge 2cbad17642 ew/complete-svn-authorship-options later to maint).
445 * Incorrect usage help message for "git worktree prune" has been fixed.
446 (merge 2488dcab22 ps/worktree-prune-help-fix later to maint).
448 * Adjust a perf test to new world order where commands that do
449 require a repository are really strict about having a repository.
450 (merge c86000c1a7 rs/p5302-create-repositories-before-tests later to maint).
452 * "git log --graph" did not work well with "--name-only", even though
453 other forms of "diff" output were handled correctly.
454 (merge f5022b5fed jk/log-graph-name-only later to maint).
456 * The push-options given via the "--push-options" option were not
457 passed through to external remote helpers such as "smart HTTP" that
458 are invoked via the transport helper.
460 * The documentation explained what "git stash" does to the working
461 tree (after stashing away the local changes) in terms of "reset
462 --hard", which was exposing an unnecessary implementation detail.
463 (merge 20a7e06172 tg/stash-doc-cleanup later to maint).
465 * When "git p4" imports changelist that removes paths, it failed to
466 convert pathnames when the p4 used encoding different from the one
467 used on the Git side. This has been corrected.
468 (merge a8b05162e8 ls/p4-path-encoding later to maint).
470 * A new coccinelle rule that catches a check of !pointer before the
471 pointer is free(3)d, which most likely is a bug.
472 (merge ec6cd14c7a rs/cocci-check-free-only-null later to maint).
474 * "ls-files" run with pathspec has been micro-optimized to avoid
475 having to memmove(3) unnecessary bytes.
476 (merge 96f6d3f61a rs/ls-files-partial-optim later to maint).
478 * A hotfix for a topic already in 'master'.
479 (merge a4d92d579f js/mingw-isatty later to maint).
481 * Other minor doc, test and build updates and code cleanups.
482 (merge f2627d9b19 sb/submodule-config-cleanup later to maint).
483 (merge 384f1a167b sb/unpack-trees-cleanup later to maint).
484 (merge 874444b704 rh/diff-orderfile-doc later to maint).
485 (merge eafd5d9483 cw/doc-sign-off later to maint).
486 (merge 0aaad415bc rs/absolute-pathdup later to maint).
487 (merge 4432dd6b5b rs/receive-pack-cleanup later to maint).
488 (merge 540a398e9c sg/mailmap-self later to maint).
489 (merge 209df269a6 nd/rev-list-all-includes-HEAD-doc later to maint).
490 (merge 941b9c5270 sb/doc-unify-bottom later to maint).
491 (merge 2aaf37b62c jk/doc-remote-helpers-markup-fix later to maint).
492 (merge e91461b332 jk/doc-submodule-markup-fix later to maint).
493 (merge 8ab9740d9f dp/submodule-doc-markup-fix later to maint).
494 (merge 0838cbc22f jk/tempfile-ferror-fclose-confusion later to maint).