1 Git v1.8.4 Release Notes
2 ========================
4 Backward compatibility notes (for Git 2.0)
5 ------------------------------------------
7 When "git push [$there]" does not say what to push, we have used the
8 traditional "matching" semantics so far (all your branches were sent
9 to the remote as long as there already are branches of the same name
10 over there). In Git 2.0, the default will change to the "simple"
11 semantics that pushes:
13 - only the current branch to the branch with the same name, and only
14 when the current branch is set to integrate with that remote
15 branch, if you are pushing to the same remote as you fetch from; or
17 - only the current branch to the branch with the same name, if you
18 are pushing to a remote that is not where you usually fetch from.
20 Use the user preference configuration variable "push.default" to
21 change this. If you are an old-timer who is used to the "matching"
22 semantics, you can set the variable to "matching" to keep the
23 traditional behaviour. If you want to live in the future early, you
24 can set it to "simple" today without waiting for Git 2.0.
26 When "git add -u" (and "git add -A") is run inside a subdirectory and
27 does not specify which paths to add on the command line, it
28 will operate on the entire tree in Git 2.0 for consistency
29 with "git commit -a" and other commands. There will be no
30 mechanism to make plain "git add -u" behave like "git add -u .".
31 Current users of "git add -u" (without a pathspec) should start
32 training their fingers to explicitly say "git add -u ."
33 before Git 2.0 comes. A warning is issued when these commands are
34 run without a pathspec and when you have local changes outside the
35 current directory, because the behaviour in Git 2.0 will be different
36 from today's version in such a situation.
38 In Git 2.0, "git add <path>" will behave as "git add -A <path>", so
39 that "git add dir/" will notice paths you removed from the directory
40 and record the removal. Versions before Git 2.0, including this
41 release, will keep ignoring removals, but the users who rely on this
42 behaviour are encouraged to start using "git add --ignore-removal <path>"
43 now before 2.0 is released.
49 Foreign interfaces, subsystems and ports.
51 * Cygwin port has been updated for more recent Cygwin 1.7.
53 * "git rebase -i" now honors --strategy and -X options.
55 * Git-gui has been updated to its 0.18.0 version.
57 * MediaWiki remote helper (in contrib/) has been updated to use the
58 credential helper interface from Git.pm.
60 * Update build for Cygwin 1.[57]. Torsten Bögershausen reports that
61 this is fine with Cygwin 1.7 (cf. <51A606A0.5060101@web.de>) so let's try moving it
64 * The credential helper to talk to keychain on OS X (in contrib/) has
65 been updated to kick in not just when talking http/https but also
68 * Remote transport helper has been updated to report errors and
69 maintain ref hierarchy used to keep track of its own state better.
71 * With "export" remote-helper protocol, (1) a push that tries to
72 update a remote ref whose name is different from the pushing side
73 does not work yet, and (2) the helper may not know how to do
74 --dry-run; these problematic cases are disabled for now.
76 * git-remote-hg/bzr (in contrib/) updates.
78 * git-remote-mw (in contrib/) hints users to check the certificate,
79 when https:// connection failed.
81 * git-remote-mw (in contrib/) adds a command to allow previewing the
82 contents locally before pushing it out, when working with a
86 UI, Workflows & Features
88 * Sample "post-receive-email" hook script got an enhanced replacement
89 "multimail" (in contrib/).
91 * Also in contrib/ is a new "contacts" script that runs "git blame"
92 to find out the people who may be interested in a set of changes.
94 * "git clean" command learned an interactive mode.
96 * The "--head" option to "git show-ref" was only to add "HEAD" to the
97 list of candidate refs to be filtered by the usual rules
98 (e.g. "--heads" that only show refs under refs/heads). The meaning
99 of the option has been changed to always show "HEAD" regardless of
100 what filtering will be applied to any other ref.
102 This is a backward incompatible change and might cause breakages to
103 people's existing scripts.
105 * "git show -s" was less discoverable than it should have been. It
106 now has a natural synonym "git show --no-patch".
108 * "git check-mailmap" is a new command that lets you map usernames
109 and e-mail addresses through the mailmap mechanism, just like many
110 built-in commands do.
112 * "git name-rev" learned to name an annotated tag object back to its
113 tagname; "git name-rev $(git rev-parse v1.0.0)" gives "tags/v1.0.0",
116 * "git cat-file --batch-check=<format>" is added, primarily to allow
117 on-disk footprint of objects in packfiles (often they are a lot
118 smaller than their true size, when expressed as deltas) to be
121 * "git rebase [-i]" used to leave just "rebase" as its reflog messages
122 for some operations. They have been reworded to be more informative.
124 * In addition to the choice from "rebase, merge, or checkout-detach",
125 "submodule update" can allow a custom command to be used in to
126 update the working tree of submodules via the "submodule.*.update"
127 configuration variable.
129 * "git submodule update" can optionally clone the submodule
130 repositories shallowly.
132 * "git format-patch" learned "--from[=whom]" option, which sets the
133 "From: " header to the specified person (or the person who runs the
134 command, if "=whom" part is missing) and move the original author
135 information to an in-body From: header as necessary.
137 * The configuration variable "merge.ff" was cleary a tri-state to
138 choose one from "favor fast-forward when possible", "always create
139 a merge even when the history could fast-forward" and "do not
140 create any merge, only update when the history fast-forwards", but
141 the command line parser did not implement the usual convention of
142 "last one wins, and command line overrides the configuration"
145 * "gitweb" learned to optionally place extra links that point at the
146 levels higher than the Gitweb pages themselves in the breadcrumbs,
147 so that it can be used as part of a larger installation.
149 * "git log --format=" now honors i18n.logoutputencoding configuration
152 * The "push.default=simple" mode of "git push" has been updated to
153 behave like "current" without requiring a remote tracking
154 information, when you push to a remote that is different from where
155 you fetch from (i.e. a triangular workflow).
157 * Having multiple "fixup!" on a line in the rebase instruction sheet
158 did not work very well with "git rebase -i --autosquash".
160 * "git log" learned the "--author-date-order" option, with which the
161 output is topologically sorted and commits in parallel histories
162 are shown intermixed together based on the author timestamp.
164 * Various subcommands of "git submodule" refused to run from anywhere
165 other than the top of the working tree of the superproject, but
166 they have been taught to let you run from a subdirectory.
168 * "git diff" learned a mode that ignores hunks whose change consists
169 only of additions and removals of blank lines, which is the same as
170 "diff -B" (ignore blank lines) of GNU diff.
172 * "git rm" gives a single message followed by list of paths to report
173 multiple paths that cannot be removed.
175 * "git rebase" can be told with ":/look for this string" syntax commits
176 to replay the changes onto and where the work to be replayed begins.
178 * Many tutorials teach users to set "color.ui" to "auto" as the first
179 thing after you set "user.name/email" to introduce yourselves to
180 Git. Now the variable defaults to "auto".
182 * On Cygwin, "cygstart" is now recognised as a possible way to start
183 a web browser (used in "help -w" and "instaweb" among others).
185 * "git status" learned status.branch and status.short configuration
186 variables to use --branch and --short options by default (override
187 with --no-branch and --no-short options from the command line).
189 * "git cmd <name>", when <name> happens to be a 40-hex string,
190 directly uses the 40-hex string as an object name, even if a ref
191 "refs/<some hierarchy>/<name>" exists. This disambiguation order
192 is unlikely to change, but we should warn about the ambiguity just
193 like we warn when more than one refs/ hierarchies share the same
196 * "git rebase" learned "--[no-]autostash" option to save local
197 changes instead of refusing to run (to which people's normal
198 response was to stash them and re-run). This introduced a corner
199 case breakage to "git am --abort" but it has been fixed.
201 * "check-ignore" (new feature since 1.8.2) has been updated to work
202 more like "check-attr" over bidi-pipes.
204 * "git describe" learned "--first-parent" option to limit its closest
205 tagged commit search to the first-parent chain.
207 * "git merge foo" that might have meant "git merge origin/foo" is
208 diagnosed with a more informative error message.
210 * "git log -L<line>,<range>:<filename>" has been added. This may
211 still have leaks and rough edges, though.
213 * We used the approxidate() parser for "--expire=<timestamp>" options
214 of various commands, but it is better to treat --expire=all and
215 --expire=now a bit more specially than using the current timestamp.
216 "git gc" and "git reflog" have been updated with a new parsing
217 function for expiry dates.
219 * Updates to completion (both bash and zsh) helpers.
221 * The behaviour of the "--chain-reply-to" option of "git send-email"
222 have changed at 1.7.0, and we added a warning/advice message to
223 help users adjust to the new behaviour back then, but we kept it
224 around for too long. The message has finally been removed.
226 * "git fetch origin master" unlike "git fetch origin" or "git fetch"
227 did not update "refs/remotes/origin/master"; this was an early
228 design decision to keep the update of remote tracking branches
229 predictable, but in practice it turns out that people find it more
230 convenient to opportunistically update them whenever we have a
231 chance, and we have been updating them when we run "git push" which
232 already breaks the original "predictability" anyway.
234 * The configuration variable core.checkstat was advertised in the
235 documentation but the code expected core.statinfo instead.
236 For now, we accept both core.checkstat and core.statinfo, but the
237 latter will be removed in the longer term.
240 Performance, Internal Implementation, etc.
242 * On Cygwin, we used to use our own lstat(2) emulation that is
243 allegedly faster than the platform one in codepaths where some of
244 the information it returns did not matter, but it started to bite
245 us in a few codepaths where the trick it uses to cheat does show
246 breakages. This emulation has been removed and we use the native
247 lstat(2) emulation supplied by Cygwin now.
249 * The function attributes extensions are used to catch mistakes in
250 use of our own variadic functions that use NULL sentinel at the end
251 (i.e. like execl(3)) and format strings (i.e. like printf(3)).
253 * The code to allow configuration data to be read from in-tree blob
254 objects is in. This may help working in a bare repository and
257 * Fetching between repositories with many refs employed O(n^2)
258 algorithm to match up the common objects, which has been corrected.
260 * The original way to specify remote repository using .git/branches/
261 used to have a nifty feature. The code to support the feature was
262 still in a function but the caller was changed not to call it 5
263 years ago, breaking that feature and leaving the supporting code
264 unreachable. The dead code has been removed.
266 * "git pack-refs" that races with new ref creation or deletion have
267 been susceptible to lossage of refs under right conditions, which
268 has been tightened up.
270 * We read loose and packed references in two steps, but after
271 deciding to read a loose ref but before actually opening it to read
272 it, another process racing with us can unlink it, which would cause
273 us to barf. The codepath has been updated to retry when such a
274 race is detected, instead of outright failing.
276 * Uses of the platform fnmatch(3) function (many places in the code,
277 matching pathspec, .gitignore and .gitattributes to name a few)
278 have been replaced with wildmatch, allowing "foo/**/bar" that would
279 match foo/bar, foo/a/bar, foo/a/b/bar, etc.
281 * Memory ownership and lifetime rules for what for-each-ref feeds to
282 its callbacks have been clarified (in short, "you do not own it, so
283 make a copy if you want to keep it").
285 * The revision traversal logic to improve culling of irrelevant
286 parents while traversing a mergy history has been updated.
288 * Some leaks in unpack-trees (used in merge, cherry-pick and other
289 codepaths) have been plugged.
291 * The codepath to read from marks files in fast-import/export did not
292 have to accept anything but 40-hex representation of the object
293 name. Further, fast-export did not need full in-core object
294 representation to have parsed wen reading from them. These
295 codepaths have been optimized by taking advantage of these access
298 * Object lookup logic, when the object hashtable starts to become
299 crowded, has been optimized.
301 * When TEST_OUTPUT_DIRECTORY setting is used, it was handled somewhat
302 inconsistently between the test framework and t/Makefile, and logic
303 to summarize the results looked at a wrong place.
305 * "git clone" uses a lighter-weight implementation when making sure
306 that the history behind refs are complete.
308 * Many warnings from sparse source checker in compat/ area has been
311 * The code to reading and updating packed-refs file has been updated,
312 correcting corner case bugs.
315 Also contains various documentation updates and code clean-ups.
321 Unless otherwise noted, all the fixes since v1.8.3 in the maintenance
322 track are contained in this release (see release notes to them for
325 * Newer Net::SMTP::SSL module does not want the user programs to use
326 the default behaviour to let server certificate go without
327 verification, so by default enable the verification with a
328 mechanism to turn it off if needed.
329 (merge 35035bb rr/send-email-ssl-verify later to maint).
331 * When "git" is spawned in such a way that any of the low 3 file
332 descriptors is closed, our first open() may yield file descriptor 2,
333 and writing error message to it would screw things up in a big way.
334 (merge a11c396 tr/protect-low-3-fds later to maint).
336 * The mailmap mechanism unnecessarily downcased the e-mail addresses
337 in the output, and also ignored the human name when it is a single
339 (merge bd23794 jc/mailmap-case-insensitivity later to maint).
341 * In two places we did not check return value (expected to be a file
342 descriptor) correctly.
343 (merge a77f106 tr/fd-gotcha-fixes later to maint).
345 * Logic to auto-detect character encodings in the commit log message
346 did not reject overlong and invalid UTF-8 characters.
347 (merge 81050ac bc/commit-invalid-utf8 later to maint).
349 * Pass port number as a separate argument when "send-email" initializes
350 Net::SMTP, instead of as a part of the hostname, i.e. host:port.
351 This allows GSSAPI codepath to match with the hostname given.
352 (merge 1a741bf bc/send-email-use-port-as-separate-param later to maint).
354 * "git diff" refused to even show difference when core.safecrlf is
355 set to true (i.e. error out) and there are offending lines in the
357 (merge 5430bb2 jc/maint-diff-core-safecrlf later to maint).
359 * A test that should have failed but didn't revealed a bug that needs
361 (merge 94d75d1 jc/t1512-fix later to maint).
363 * An overlong path to a .git directory may have overflown the
364 temporary path buffer used to create a name for lockfiles.
365 (merge 2fbd4f9 mh/maint-lockfile-overflow later to maint).
367 * Invocations of "git checkout" used internally by "git rebase" were
368 counted as "checkout", and affected later "git checkout -" to the
369 the user to an unexpected place.
370 (merge 3bed291 rr/rebase-checkout-reflog later to maint).
372 * The configuration variable column.ui was poorly documented.
373 (merge 5e62cc1 rr/column-doc later to maint).
375 * "git name-rev --refs=tags/v*" were forbidden, which was a bit
376 inconvenient (you had to give a pattern to match refs fully, like
377 --refs=refs/tags/v*).
378 (merge 98c5c4a nk/name-rev-abbreviated-refs later to maint).
380 * "git apply" parsed patches that add new files, generated by
381 programs other than Git, incorrectly. This is an old breakage in
382 v1.7.11 and will need to be merged down to the maintenance tracks.
384 * Older cURL wanted piece of memory we call it with to be stable, but
385 we updated the auth material after handing it to a call.
387 * "git pull" into nothing trashed "local changes" that were in the
388 index, and this avoids it.
390 * Many "git submodule" operations do not work on a submodule at a
391 path whose name is not in ASCII.
393 * "cherry-pick" had a small leak in an error codepath.
395 * Logic used by git-send-email to suppress cc mishandled names like
396 "A U. Thor" <author@example.xz>, where the human readable part
397 needs to be quoted (the user input may not have the double quotes
398 around the name, and comparison was done between quoted and
399 unquoted strings). It also mishandled names that need RFC2047
402 * Call to discard_cache/discard_index (used when we use different
403 contents of the index in-core, in many operations like commit,
404 apply, and merge) used to leak memory that held the array of index
405 entries, which has been plugged.
406 (merge a0fc4db rs/discard-index-discard-array later to maint).
408 * "gitweb" forgot to clear a global variable $search_regexp upon each
409 request, mistakenly carrying over the previous search to a new one
410 when used as a persistent CGI.
412 * The wildmatch engine did not honor WM_CASEFOLD option correctly.
414 * "git log -c --follow $path" segfaulted upon hitting the commit that
415 renamed the $path being followed.
417 * When a reflog notation is used for implicit "current branch", we
418 did not say which branch and worse said "branch ''".
420 * "difftool --dir-diff" did not copy back changes made by the
421 end-user in the diff tool backend to the working tree in some
424 * "git push $there HEAD:branch" did not resolve HEAD early enough, so
425 it was easy to flip it around while push is still going on and push
426 out a branch that the user did not originally intended when the
429 * The bash prompt code (in contrib/) displayed the name of the branch
430 being rebased when "rebase -i/-m/-p" modes are in use, but not the
431 plain vanilla "rebase".
433 * Handling of negative exclude pattern for directories "!dir" was
434 broken in the update to v1.8.3.
436 * zsh prompt script that borrowed from bash prompt script did not
437 work due to slight differences in array variable notation between
440 * An entry for "file://" scheme in the enumeration of URL types Git
441 can take in the HTML documentation was made into a clickable link
444 * "git push --[no-]verify" was not documented.
446 * Stop installing the git-remote-testpy script that is only used for
449 * "git commit --allow-empty-message -m ''" should not start an
452 * "git merge @{-1}~22" was rewritten to "git merge frotz@{1}~22"
453 incorrectly when your previous branch was "frotz" (it should be
454 rewritten to "git merge frotz~22" instead).
456 * "git diff -c -p" was not showing a deleted line from a hunk when
457 another hunk immediately begins where the earlier one ends.
459 * "git log --ancestry-path A...B" did not work as expected, as it did
460 not pay attention to the fact that the merge base between A and B
461 was the bottom of the range being specified.
463 * Mac OS X does not like to write(2) more than INT_MAX number of
464 bytes; work it around by chopping write(2) into smaller pieces.
466 * Newer MacOS X encourages the programs to compile and link with
467 their CommonCrypto, not with OpenSSL.
469 * "git clone foo/bar:baz" cannot be a request to clone from a remote
470 over git-over-ssh specified in the scp style. This case is now
471 detected and clones from a local repository at "foo/bar:baz".
473 * When $HOME is misconfigured to point at an unreadable directory, we
474 used to complain and die. Loosen the check.
476 * "git subtree" (in contrib/) had one codepath with loose error
477 checks to lose data at the remote side.
479 * "git fetch" into a shallow repository from a repository that does
480 not know about the shallow boundary commits (e.g. a different fork
481 from the repository the current shallow repository was cloned from)
482 did not work correctly.
484 * "git checkout foo" DWIMs the intended "upstream" and turns it into
485 "git checkout -t -b foo remotes/origin/foo". This codepath has been
486 updated to correctly take existing remote definitions into account.
projects / git / blob