Git 1.8.2-rc2
[git/git.git] / Documentation / RelNotes / 1.8.2.txt
1 Git v1.8.2 Release Notes
2 ========================
4 Backward compatibility notes
5 ----------------------------
7 In the next major release Git 2.0 (not *this* one), we will change the
8 behavior of the "git push" command.
10 When "git push [$there]" does not say what to push, we have used the
11 traditional "matching" semantics so far (all your branches were sent
12 to the remote as long as there already are branches of the same name
13 over there). We will use the "simple" semantics that pushes the
14 current branch to the branch with the same name, only when the current
15 branch is set to integrate with that remote branch. There is a user
16 preference configuration variable "push.default" to change this.
18 "git push $there tag v1.2.3" used to allow replacing a tag v1.2.3
19 that already exists in the repository $there, if the rewritten tag
20 you are pushing points at a commit that is a descendant of a commit
21 that the old tag v1.2.3 points at. This was found to be error prone
22 and starting with this release, any attempt to update an existing
23 ref under refs/tags/ hierarchy will fail, without "--force".
25 When "git add -u" and "git add -A", that does not specify what paths
26 to add on the command line, is run from inside a subdirectory, the
27 scope of the operation has always been limited to the subdirectory.
28 Many users found this counter-intuitive, given that "git commit -a"
29 and other commands operate on the entire tree regardless of where you
30 are. In this release, these commands give warning in such a case and
31 encourage the user to say "git add -u/-A ." instead when restricting
32 the scope to the current directory.
34 At Git 2.0 (not *this* one), we plan to change these commands without
35 pathspec to operate on the entire tree. Forming a habit to type "."
36 when you mean to limit the command to the current working directory
37 will protect you against the planned future change, and that is the
38 whole point of the new message (there will be no configuration
39 variable to squelch this warning---it goes against the "habit forming"
40 objective).
43 Updates since v1.8.1
44 --------------------
46 UI, Workflows & Features
48 * Initial ports to QNX and z/OS UNIX System Services have started.
50 * Output from the tests is coloured using "green is okay, yellow is
51 questionable, red is bad and blue is informative" scheme.
53 * Mention of "GIT/Git/git" in the documentation have been updated to
54 be more uniform and consistent. The name of the system and the
55 concept it embodies is "Git"; the command the users type is "git".
56 All-caps "GIT" was merely a way to imitate "Git" typeset in small
57 caps in our ASCII text only documentation and to be avoided.
59 * The completion script (in contrib/completion) used to let the
60 default completer to suggest pathnames, which gave too many
61 irrelevant choices (e.g. "git add" would not want to add an
62 unmodified path). It learnt to use a more git-aware logic to
63 enumerate only relevant ones.
65 * In bare repositories, "git shortlog" and other commands now read
66 mailmap files from the tip of the history, to help running these
67 tools in server settings.
69 * Color specifiers, e.g. "%C(blue)Hello%C(reset)", used in the
70 "--format=" option of "git log" and friends can be disabled when
71 the output is not sent to a terminal by prefixing them with
72 "auto,", e.g. "%C(auto,blue)Hello%C(auto,reset)".
74 * Scripts can ask Git that wildcard patterns in pathspecs they give do
75 not have any significance, i.e. take them as literal strings.
77 * The patterns in .gitignore and .gitattributes files can have **/,
78 as a pattern that matches 0 or more levels of subdirectory.
79 E.g. "foo/**/bar" matches "bar" in "foo" itself or in a
80 subdirectory of "foo".
82 * When giving arguments without "--" disambiguation, object names
83 that come earlier on the command line must not be interpretable as
84 pathspecs and pathspecs that come later on the command line must
85 not be interpretable as object names. This disambiguation rule has
86 been tweaked so that ":/" (no other string before or after) is
87 always interpreted as a pathspec; "git cmd -- :/" is no longer
88 needed, you can just say "git cmd :/".
90 * Various "hint" lines Git gives when it asks the user to edit
91 messages in the editor are commented out with '#' by default. The
92 core.commentchar configuration variable can be used to customize
93 this '#' to a different character.
95 * "git add -u" and "git add -A" without pathspec issues warning to
96 make users aware that they are only operating on paths inside the
97 subdirectory they are in. Use ":/" (everything from the top) or
98 "." (everything from the $cwd) to disambiguate.
100 * "git blame" (and "git diff") learned the "--no-follow" option.
102 * "git branch" now rejects some nonsense combinations of command line
103 arguments (e.g. giving more than one branch name to rename) with
104 more case-specific error messages.
106 * "git check-ignore" command to help debugging .gitignore files has
107 been added.
109 * "git cherry-pick" can be used to replay a root commit to an unborn
110 branch.
112 * "git commit" can be told to use --cleanup=whitespace by setting the
113 configuration variable commit.cleanup to 'whitespace'.
115 * "git diff" and other Porcelain commands can be told to use a
116 non-standard algorithm by setting diff.algorithm configuration
117 variable.
119 * "git fetch --mirror" and fetch that uses other forms of refspec
120 with wildcard used to attempt to update a symbolic ref that match
121 the wildcard on the receiving end, which made little sense (the
122 real ref that is pointed at by the symbolic ref would be updated
123 anyway). Symbolic refs no longer are affected by such a fetch.
125 * "git format-patch" now detects more cases in which a whole branch
126 is being exported, and uses the description for the branch, when
127 asked to write a cover letter for the series.
129 * "git format-patch" learned "-v $count" option, and prepends a
130 string "v$count-" to the names of its output files, and also
131 automatically sets the subject prefix to "PATCH v$count". This
132 allows patches from rerolled series to be stored under different
133 names and makes it easier to reuse cover letter messages.
135 * "git log" and friends can be told with --use-mailmap option to
136 rewrite the names and email addresses of people using the mailmap
137 mechanism.
139 * "git log --cc --graph" now shows the combined diff output with the
140 ancestry graph.
142 * "git log --grep=<pattern>" honors i18n.logoutputencoding to look
143 for the pattern after fixing the log message to the specified
144 encoding.
146 * "git mergetool" and "git difftool" learned to list the available
147 tool backends in a more consistent manner.
149 * "git mergetool" is aware of TortoiseGitMerge now and uses it over
150 TortoiseMerge when available.
152 * "git push" now requires "-f" to update a tag, even if it is a
153 fast-forward, as tags are meant to be fixed points.
155 * Error messages from "git push" when it stops to prevent remote refs
156 from getting overwritten by mistake have been improved to explain
157 various situations separately.
159 * "git push" will stop without doing anything if the new "pre-push"
160 hook exists and exits with a failure.
162 * When "git rebase" fails to generate patches to be applied (e.g. due
163 to oom), it failed to detect the failure and instead behaved as if
164 there were nothing to do. A workaround to use a temporary file has
165 been applied, but we probably would want to revisit this later, as
166 it hurts the common case of not failing at all.
168 * Input and preconditions to "git reset" has been loosened where
169 appropriate. "git reset $fromtree Makefile" requires $fromtree to
170 be any tree (it used to require it to be a commit), for example.
171 "git reset" (without options or parameters) used to error out when
172 you do not have any commits in your history, but it now gives you
173 an empty index (to match non-existent commit you are not even on).
175 * "git status" says what branch is being bisected or rebased when
176 able, not just "bisecting" or "rebasing".
178 * "git submodule" started learning a new mode to integrate with the
179 tip of the remote branch (as opposed to integrating with the commit
180 recorded in the superproject's gitlink).
182 * "git upload-pack" which implements the service "ls-remote" and
183 "fetch" talk to can be told to hide ref hierarchies the server
184 side internally uses (and that clients have no business learning
185 about) with transfer.hiderefs configuration.
188 Foreign Interface
190 * "git fast-export" has been updated for its use in the context of
191 the remote helper interface.
193 * A new remote helper to interact with bzr has been added to contrib/.
195 * "git p4" got various bugfixes around its branch handling. It is
196 also made usable with Python 2.4/2.5. In addition, its various
197 portability issues for Cygwin have been addressed.
199 * The remote helper to interact with Hg in contrib/ has seen a few
200 fixes.
203 Performance, Internal Implementation, etc.
205 * "git fsck" has been taught to be pickier about entries in tree
206 objects that should not be there, e.g. ".", ".git", and "..".
208 * Matching paths with common forms of pathspecs that contain wildcard
209 characters has been optimized further.
211 * We stopped paying attention to $GIT_CONFIG environment that points
212 at a single configuration file from any command other than "git config"
213 quite a while ago, but "git clone" internally set, exported, and
214 then unexported the variable during its operation unnecessarily.
216 * "git reset" internals has been reworked and should be faster in
217 general. We tried to be careful not to break any behaviour but
218 there could be corner cases, especially when running the command
219 from a conflicted state, that we may have missed.
221 * The implementation of "imap-send" has been updated to reuse xml
222 quoting code from http-push codepath, and lost a lot of unused
223 code.
225 * There is a simple-minded checker for the test scripts in t/
226 directory to catch most common mistakes (it is not enabled by
227 default).
229 * You can build with USE_WILDMATCH=YesPlease to use a replacement
230 implementation of pattern matching logic used for pathname-like
231 things, e.g. refnames and paths in the repository. This new
232 implementation is not expected change the existing behaviour of Git
233 in this release, except for "git for-each-ref" where you can now
234 say "refs/**/master" and match with both refs/heads/master and
235 refs/remotes/origin/master. We plan to use this new implementation
236 in wider places (e.g. "git ls-files '**/Makefile' may find Makefile
237 at the top-level, and "git log '**/t*.sh'" may find commits that
238 touch a shell script whose name begins with "t" at any level) in
239 future versions of Git, but we are not there yet. By building with
240 USE_WILDMATCH, using the resulting Git daily and reporting when you
241 find breakages, you can help us get closer to that goal.
243 * Some reimplementations of Git do not write all the stat info back
244 to the index due to their implementation limitations (e.g. jgit).
245 A configuration option can tell Git to ignore changes to most of
246 the stat fields and only pay attention to mtime and size, which
247 these implementations can reliably update. This can be used to
248 avoid excessive revalidation of contents.
250 * Some platforms ship with old version of expat where xmlparse.h
251 needs to be included instead of expat.h; the build procedure has
252 been taught about this.
254 * "make clean" on platforms that cannot compute header dependencies
255 on the fly did not work with implementations of "rm" that do not
256 like an empty argument list.
258 Also contains minor documentation updates and code clean-ups.
261 Fixes since v1.8.1
262 ------------------
264 Unless otherwise noted, all the fixes since v1.8.1 in the maintenance
265 track are contained in this release (see release notes to them for
266 details).
268 * An element on GIT_CEILING_DIRECTORIES list that does not name the
269 real path to a directory (i.e. a symbolic link) could have caused
270 the GIT_DIR discovery logic to escape the ceiling.
272 * When attempting to read the XDG-style $HOME/.config/git/config and
273 finding that $HOME/.config/git is a file, we gave a wrong error
274 message, instead of treating the case as "a custom config file does
275 not exist there" and moving on.
277 * The behaviour visible to the end users was confusing, when they
278 attempt to kill a process spawned in the editor that was in turn
279 launched by Git with SIGINT (or SIGQUIT), as Git would catch that
280 signal and die. We ignore these signals now.
281 (merge 0398fc34 pf/editor-ignore-sigint later to maint).
283 * A child process that was killed by a signal (e.g. SIGINT) was
284 reported in an inconsistent way depending on how the process was
285 spawned by us, with or without a shell in between.
287 * After failing to create a temporary file using mkstemp(), failing
288 pathname was not reported correctly on some platforms.
290 * We used to stuff "user@" and then append what we read from
291 /etc/mailname to come up with a default e-mail ident, but a bug
292 lost the "user@" part.
294 * The attribute mechanism didn't allow limiting attributes to be
295 applied to only a single directory itself with "path/" like the
296 exclude mechanism does. The initial implementation of this that
297 was merged to 'maint' and was with a severe performance
298 degradations and needs to merge a fix-up topic.
300 * The smart HTTP clients forgot to verify the content-type that comes
301 back from the server side to make sure that the request is being
302 handled properly.
304 * "git am" did not parse datestamp correctly from Hg generated patch,
305 when it is run in a locale outside C (or en).
307 * "git apply" misbehaved when fixing whitespace breakages by removing
308 excess trailing blank lines.
310 * "git apply --summary" has been taught to make sure the similarity
311 value shown in its output is sensible, even when the input had a
312 bogus value.
314 * A tar archive created by "git archive" recorded a directory in a
315 way that made NetBSD's implementation of "tar" sometimes unhappy.
317 * "git archive" did not record uncompressed size in the header when
318 streaming a zip archive, which confused some implementations of unzip.
320 * "git archive" did not parse configuration values in tar.* namespace
321 correctly.
322 (merge b3873c3 jk/config-parsing-cleanup later to maint).
324 * Attempt to "branch --edit-description" an existing branch, while
325 being on a detached HEAD, errored out.
327 * "git clean" showed what it was going to do, but sometimes end up
328 finding that it was not allowed to do so, which resulted in a
329 confusing output (e.g. after saying that it will remove an
330 untracked directory, it found an embedded git repository there
331 which it is not allowed to remove). It now performs the actions
332 and then reports the outcome more faithfully.
334 * When "git clone --separate-git-dir=$over_there" is interrupted, it
335 failed to remove the real location of the $GIT_DIR it created.
336 This was most visible when interrupting a submodule update.
338 * "git cvsimport" mishandled timestamps at DST boundary.
340 * We used to have an arbitrary 32 limit for combined diff input,
341 resulting in incorrect number of leading colons shown when showing
342 the "--raw --cc" output.
344 * "git fetch --depth" was broken in at least three ways. The
345 resulting history was deeper than specified by one commit, it was
346 unclear how to wipe the shallowness of the repository with the
347 command, and documentation was misleading.
348 (merge cfb70e1 nd/fetch-depth-is-broken later to maint).
350 * "git log --all -p" that walked refs/notes/textconv/ ref can later
351 try to use the textconv data incorrectly after it gets freed.
353 * We forgot to close the file descriptor reading from "gpg" output,
354 killing "git log --show-signature" on a long history.
356 * The way "git svn" asked for password using SSH_ASKPASS and
357 GIT_ASKPASS was not in line with the rest of the system.
359 * The --graph code fell into infinite loop when asked to do what the
360 code did not expect.
362 * http transport was wrong to ask for the username when the
363 authentication is done by certificate identity.
365 * "git pack-refs" that ran in parallel to another process that
366 created new refs had a nasty race.
368 * Rebasing the history of superproject with change in the submodule
369 has been broken since v1.7.12.
371 * After "git add -N" and then writing a tree object out of the
372 index, the cache-tree data structure got corrupted.
374 * "git clone" used to allow --bare and --separate-git-dir=$there
375 options at the same time, which was nonsensical.
377 * "git rebase --preserve-merges" lost empty merges in recent versions
378 of Git.
380 * "git merge --no-edit" computed who were involved in the work done
381 on the side branch, even though that information is to be discarded
382 without getting seen in the editor.
384 * "git merge" started calling prepare-commit-msg hook like "git
385 commit" does some time ago, but forgot to pay attention to the exit
386 status of the hook.
388 * A failure to push due to non-ff while on an unborn branch
389 dereferenced a NULL pointer when showing an error message.
391 * When users spell "cc:" in lowercase in the fake "header" in the
392 trailer part, "git send-email" failed to pick up the addresses from
393 there. As e-mail headers field names are case insensitive, this
394 script should follow suit and treat "cc:" and "Cc:" the same way.
396 * Output from "git status --ignored" showed an unexpected interaction
397 with "--untracked".
399 * "gitweb", when sorting by age to show repositories with new
400 activities first, used to sort repositories with absolutely
401 nothing in it early, which was not very useful.
403 * "gitweb"'s code to sanitize control characters before passing it to
404 "highlight" filter lost known-to-be-safe control characters by
405 mistake.
407 * "gitweb" pages served over HTTPS, when configured to show picon or
408 gravatar, referred to these external resources to be fetched via
409 HTTP, resulting in mixed contents warning in browsers.
411 * When a line to be wrapped has a solid run of non space characters
412 whose length exactly is the wrap width, "git shortlog -w" failed
413 to add a newline after such a line.
415 * Command line completion leaked an unnecessary error message while
416 looking for possible matches with paths in <tree-ish>.
418 * Command line completion for "tcsh" emitted an unwanted space
419 after completing a single directory name.
421 * Command line completion code was inadvertently made incompatible with
422 older versions of bash by using a newer array notation.
424 * "git push" was taught to refuse updating the branch that is
425 currently checked out long time ago, but the user manual was left
426 stale.
427 (merge 50995ed wk/man-deny-current-branch-is-default-these-days later to maint).
429 * Some shells do not behave correctly when IFS is unset; work it
430 around by explicitly setting it to the default value.
432 * Some scripted programs written in Python did not get updated when
433 PYTHON_PATH changed.
434 (cherry-pick 96a4647fca54031974cd6ad1 later to maint).
436 * When autoconf is used, any build on a different commit always ran
437 "config.status --recheck" even when unnecessary.
439 * A fix was added to the build procedure to work around buggy
440 versions of ccache broke the auto-generation of dependencies, which
441 unfortunately is still relevant because some people use ancient
442 distros.
444 * The autoconf subsystem passed --mandir down to generated
445 config.mak.autogen but forgot to do the same for --htmldir.
446 (merge 55d9bf0 ct/autoconf-htmldir later to maint).
448 * A change made on v1.8.1.x maintenance track had a nasty regression
449 to break the build when autoconf is used.
450 (merge 7f1b697 jn/less-reconfigure later to maint).
452 * We have been carrying a translated and long-unmaintained copy of an
453 old version of the tutorial; removed.
455 * t0050 had tests expecting failures from a bug that was fixed some
456 time ago.
458 * t4014, t9502 and t0200 tests had various portability issues that
459 broke on OpenBSD.
461 * t9020 and t3600 tests had various portability issues.
463 * t9200 runs "cvs init" on a directory that already exists, but a
464 platform can configure this fail for the current user (e.g. you
465 need to be in the cvsadmin group on NetBSD 6.0).
467 * t9020 and t9810 had a few non-portable shell script construct.
469 * Scripts to test bash completion was inherently flaky as it was
470 affected by whatever random things the user may have on $PATH.
472 * An element on GIT_CEILING_DIRECTORIES could be a "logical" pathname
473 that uses a symbolic link to point at somewhere else (e.g. /home/me
474 that points at /net/host/export/home/me, and the latter directory
475 is automounted). Earlier when Git saw such a pathname e.g. /home/me
476 on this environment variable, the "ceiling" mechanism did not take
477 effect. With this release (the fix has also been merged to the
478 v1.8.1.x maintenance series), elements on GIT_CEILING_DIRECTORIES
479 are by default checked for such aliasing coming from symbolic
480 links. As this needs to actually resolve symbolic links for each
481 element on the GIT_CEILING_DIRECTORIES, you can disable this
482 mechanism for some elements by listing them after an empty element
483 on the GIT_CEILING_DIRECTORIES. e.g. Setting /home/me::/home/him to
484 GIT_CEILING_DIRECTORIES makes Git resolve symbolic links in
485 /home/me when checking if the current directory is under /home/me,
486 but does not do so for /home/him.
487 (merge 7ec30aa mh/maint-ceil-absolute later to maint).