| blob | ad6496f826bdede34070d5e18c953d51063d8a4b |
1 Git v1.8.2 Release Notes
2 ========================
4 Backward compatibility notes
5 ----------------------------
7 In the upcoming major release (tentatively called 1.8.2), we will
8 change the 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 decendant 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".
26 Updates since v1.8.1
27 --------------------
29 UI, Workflows & Features
31 * Initial ports to QNX and z/OS UNIX System Services have started.
33 * Output from the tests is coloured using "green is okay, yellow is
34 questionable, red is bad and blue is informative" scheme.
36 * Mention of "GIT/Git/git" in the documentation have been updated to
37 be more uniform and consistent. The name of the system and the
38 concept it embodies is "Git"; the command the users type is "git".
39 All-caps "GIT" was merely a way to imitate "Git" typeset in small
40 caps in our ASCII text only documentation and to be avoided.
42 * The completion script (in contrib/completion) used to let the
43 default completer to suggest pathnames, which gave too many
44 irrelevant choices (e.g. "git add" would not want to add an
45 unmodified path). It learnt to use a more git-aware logic to
46 enumerate only relevant ones.
48 * In bare repositories, "git shortlog" and other commands now read
49 mailmap files from the tip of the history, to help running these
50 tools in server settings.
52 * Color specifiers, e.g. "%C(blue)Hello%C(reset)", used in the
53 "--format=" option of "git log" and friends can be disabled when
54 the output is not sent to a terminal by prefixing them with
55 "auto,", e.g. "%C(auto,blue)Hello%C(auto,reset)".
57 * Scripts can ask Git that wildcard patterns in pathspecs they give do
58 not have any significance, i.e. take them as literal strings.
60 * The patterns in .gitignore and .gitattributes files can have **/,
61 as a pattern that matches 0 or more levels of subdirectory.
62 E.g. "foo/**/bar" matches "bar" in "foo" itself or in a
63 subdirectory of "foo".
65 * When giving arguments without "--" disambiguation, object names
66 that come earlier on the command line must not be interpretable as
67 pathspecs and pathspecs that come later on the command line must
68 not be interpretable as object names. This disambiguation rule has
69 been tweaked so that ":/" (no other string before or after) is
70 always interpreted as a pathspec; "git cmd -- :/" is no longer
71 needed, you can just say "git cmd :/".
73 * Various "hint" lines Git gives when it asks the user to edit
74 messages in the editor are commented out with '#' by default. The
75 core.commentchar configuration variable can be used to customize
76 this '#' to a different character.
78 * "git add -u" and "git add -A" without pathspec issues warning to
79 make users aware that they are only operating on paths inside the
80 subdirectory they are in. Use ":/" (everything from the top) or
81 "." (everything from the $cwd) to disambiguate.
83 * "git blame" (and "git diff") learned the "--no-follow" option.
85 * "git branch" now rejects some nonsense combinations of command line
86 arguments (e.g. giving more than one branch name to rename) with
87 more case-specific error messages.
89 * "git check-ignore" command to help debugging .gitignore files has
90 been added.
92 * "git cherry-pick" can be used to replay a root commit to an unborn
93 branch.
95 * "git commit" can be told to use --cleanup=whitespace by setting the
96 configuration variable commit.cleanup to 'whitespace'.
98 * "git fetch --mirror" and fetch that uses other forms of refspec
99 with wildcard used to attempt to update a symbolic ref that match
100 the wildcard on the receiving end, which made little sense (the
101 real ref that is pointed at by the symbolic ref would be updated
102 anyway). Symbolic refs no longer are affected by such a fetch.
104 * "git format-patch" now detects more cases in which a whole branch
105 is being exported, and uses the description for the branch, when
106 asked to write a cover letter for the series.
108 * "git format-patch" learned "-v $count" option, and prepends a
109 string "v$count-" to the names of its output files, and also
110 automatically sets the subject prefix to "PATCH v$count". This
111 allows patches from rerolled series to be stored under different
112 names and makes it easier to reuse cover letter messsages.
114 * "git log" and friends can be told with --use-mailmap option to
115 rewrite the names and email addresses of people using the mailmap
116 mechanism.
118 * "git mergetool" and "git difftool" learned to list the available
119 tool backends in a more consistent manner.
121 * "git mergetool" is aware of TortoiseGitMerge now and uses it over
122 TortoiseMerge when available.
124 * "git push" now requires "-f" to update a tag, even if it is a
125 fast-forward, as tags are meant to be fixed points.
127 * Error messages from "git push" when it stops to prevent remote refs
128 from getting overwritten by mistake have been improved to explain
129 various situations separately.
131 * "git push" will stop without doing anything if the new "pre-push"
132 hook exists and exits with a failure.
134 * When "git rebase" fails to generate patches to be applied (e.g. due
135 to oom), it failed to detect the failure and instead behaved as if
136 there were nothing to do. A workaround to use a temporary file has
137 been applied, but we probably would want to revisit this later, as
138 it hurts the common case of not failing at all.
140 * Input and preconditions to "git reset" has been loosened where
141 appropriate. "git reset $fromtree Makefile" requires $fromtree to
142 be any tree (it used to require it to be a commit), for example.
143 "git reset" (without options or parameters) used to error out when
144 you do not have any commits in your history, but it now gives you
145 an empty index (to match non-existent commit you are not even on).
147 * "git submodule" started learning a new mode to integrate with the
148 tip of the remote branch (as opposed to integrating with the commit
149 recorded in the superproject's gitlink).
152 Foreign Interface
154 * "git fast-export" has been updated for its use in the context of
155 the remote helper interface.
157 * A new remote helper to interact with bzr has been added to contrib/.
159 * "git p4" got various bugfixes around its branch handling. It is
160 also made usable with Python 2.4/2.5. In addition, its various
161 portability issues for Cygwin have been addressed.
163 * The remote helper to interact with Hg in contrib/ has seen a few
164 fixes.
167 Performance, Internal Implementation, etc.
169 * "git fsck" has been taught to be pickier about entries in tree
170 objects that should not be there, e.g. ".", ".git", and "..".
172 * Matching paths with common forms of pathspecs that contain wildcard
173 characters has been optimized further.
175 * We stopped paying attention to $GIT_CONFIG environment that points
176 at a single configuration file from any command other than "git config"
177 quite a while ago, but "git clone" internally set, exported, and
178 then unexported the variable during its operation unnecessarily.
180 * "git reset" internals has been reworked and should be faster in
181 general. We tried to be careful not to break any behaviour but
182 there could be corner cases, especially when running the command
183 from a conflicted state, that we may have missed.
185 * The implementation of "imap-send" has been updated to reuse xml
186 quoting code from http-push codepath, and lost a lot of unused
187 code.
189 * There is a simple-minded checker for the test scripts in t/
190 directory to catch most common mistakes (it is not enabled by
191 default).
193 * You can build with USE_WILDMATCH=YesPlease to use a replacement
194 implementation of pattern matching logic used for pathname-like
195 things, e.g. refnames and paths in the repository. This new
196 implementation is not expected change the existing behaviour of Git
197 in this release, except for "git for-each-ref" where you can now
198 say "refs/**/master" and match with both refs/heads/master and
199 refs/remotes/origin/master. We plan to use this new implementation
200 in wider places (e.g. "git ls-files '**/Makefile' may find Makefile
201 at the top-level, and "git log '**/t*.sh'" may find commits that
202 touch a shell script whose name begins with "t" at any level) in
203 future versions of Git, but we are not there yet. By building with
204 USE_WILDMATCH, using the resulting Git daily and reporting when you
205 find breakages, you can help us get closer to that goal.
207 * Some reimplementations of Git do not write all the stat info back
208 to the index due to their implementation limitations (e.g. jgit).
209 A configuration option can tell Git to ignore changes to most of
210 the stat fields and only pay attention to mtime and size, which
211 these implementations can reliably update. This can be used to
212 avoid excessive revalidation of contents.
215 Also contains minor documentation updates and code clean-ups.
218 Fixes since v1.8.1
219 ------------------
221 Unless otherwise noted, all the fixes since v1.8.1 in the maintenance
222 track are contained in this release (see release notes to them for
223 details).
225 * An element on GIT_CEILING_DIRECTORIES list that does not name the
226 real path to a directory (i.e. a symbolic link) could have caused
227 the GIT_DIR discovery logic to escape the ceiling.
229 * When attempting to read the XDG-style $HOME/.config/git/config and
230 finding that $HOME/.config/git is a file, we gave a wrong error
231 message, instead of treating the case as "a custom config file does
232 not exist there" and moving on.
234 * The behaviour visible to the end users was confusing, when they
235 attempt to kill a process spawned in the editor that was in turn
236 launched by Git with SIGINT (or SIGQUIT), as Git would catch that
237 signal and die. We ignore these signals now.
238 (merge 0398fc34 pf/editor-ignore-sigint later to maint).
240 * A child process that was killed by a signal (e.g. SIGINT) was
241 reported in an inconsistent way depending on how the process was
242 spawned by us, with or without a shell in between.
244 * After failing to create a temporary file using mkstemp(), failing
245 pathname was not reported correctly on some platforms.
247 * We used to stuff "user@" and then append what we read from
248 /etc/mailname to come up with a default e-mail ident, but a bug
249 lost the "user@" part.
251 * The attribute mechanism didn't allow limiting attributes to be
252 applied to only a single directory itself with "path/" like the
253 exclude mechanism does. The initial implementation of this that
254 was merged to 'maint' and 1.8.1.2 was with a severe performance
255 degradations and needs to merge a fix-up topic.
257 * The smart HTTP clients forgot to verify the content-type that comes
258 back from the server side to make sure that the request is being
259 handled properly.
260 (merge 3443db5 sp/smart-http-content-type-check later to maint).
262 * "git am" did not parse datestamp correctly from Hg generated patch,
263 when it is run in a locale outside C (or en).
265 * "git apply" misbehaved when fixing whitespace breakages by removing
266 excess trailing blank lines.
268 * "git apply --summary" has been taught to make sure the similarity
269 value shown in its output is sensible, even when the input had a
270 bogus value.
271 (merge afcb6ac jk/apply-similaritly-parsing later to maint).
273 * A tar archive created by "git archive" recorded a directory in a
274 way that made NetBSD's implementation of "tar" sometimes unhappy.
276 * "git archive" did not record uncompressed size in the header when
277 streaming a zip archive, which confused some implementations of unzip.
279 * "git archive" did not parse configuration values in tar.* namespace
280 correctly.
281 (merge b3873c3 jk/config-parsing-cleanup later to maint).
283 * Attempt to "branch --edit-description" an existing branch, while
284 being on a detached HEAD, errored out.
286 * "git clean" showed what it was going to do, but sometimes end up
287 finding that it was not allowed to do so, which resulted in a
288 confusing output (e.g. after saying that it will remove an
289 untracked directory, it found an embedded git repository there
290 which it is not allowed to remove). It now performs the actions
291 and then reports the outcome more faithfully.
292 (merge f538a91 zk/clean-report-failure later to maint).
294 * When "git clone --separate-git-dir=$over_there" is interrupted, it
295 failed to remove the real location of the $GIT_DIR it created.
296 This was most visible when interrupting a submodule update.
298 * We used to have an arbitrary 32 limit for combined diff input,
299 resulting in incorrect number of leading colons shown when showing
300 the "--raw --cc" output.
301 (merge edbc00e jc/combine-diff-many-parents later to maint).
303 * "git fetch --depth" was broken in at least three ways. The
304 resulting history was deeper than specified by one commit, it was
305 unclear how to wipe the shallowness of the repository with the
306 command, and documentation was misleading.
307 (merge cfb70e1 nd/fetch-depth-is-broken later to maint).
309 * "git log --all -p" that walked refs/notes/textconv/ ref can later
310 try to use the textconv data incorrectly after it gets freed.
311 (merge be5c9fb jk/read-commit-buffer-data-after-free later to maint).
313 * We forgot to close the file descriptor reading from "gpg" output,
314 killing "git log --show-signature" on a long history.
316 * The way "git svn" asked for password using SSH_ASKPASS and
317 GIT_ASKPASS was not in line with the rest of the system.
319 * The --graph code fell into infinite loop when asked to do what the
320 code did not expect.
322 * http transport was wrong to ask for the username when the
323 authentication is done by certificate identity.
325 * "git pack-refs" that ran in parallel to another process that
326 created new refs had a nasty race.
328 * Rebasing the history of superproject with change in the submodule
329 has been broken since v1.7.12.
331 * After "git add -N" and then writing a tree object out of the
332 index, the cache-tree data structure got corrupted.
334 * "git clone" used to allow --bare and --separate-git-dir=$there
335 options at the same time, which was nonsensical.
336 (merge 95b63f1 nd/clone-no-separate-git-dir-with-bare later to maint).
338 * "git rebase --preserve-merges" lost empty merges in recent versions
339 of Git.
341 * "git merge --no-edit" computed who were involved in the work done
342 on the side branch, even though that information is to be discarded
343 without getting seen in the editor.
345 * "git merge" started calling prepare-commit-msg hook like "git
346 commit" does some time ago, but forgot to pay attention to the exit
347 status of the hook.
349 * A failure to push due to non-ff while on an unborn branch
350 dereferenced a NULL pointer when showing an error message.
352 * When users spell "cc:" in lowercase in the fake "header" in the
353 trailer part, "git send-email" failed to pick up the addresses from
354 there. As e-mail headers field names are case insensitive, this
355 script should follow suit and treat "cc:" and "Cc:" the same way.
357 * Output from "git status --ignored" showed an unexpected interaction
358 with "--untracked".
360 * "gitweb", when sorting by age to show repositories with new
361 activities first, used to sort repositories with absolutely
362 nothing in it early, which was not very useful.
364 * "gitweb"'s code to sanitize control characters before passing it to
365 "highlight" filter lost known-to-be-safe control characters by
366 mistake.
368 * "gitweb" pages served over HTTPS, when configured to show picon or
369 gravatar, referred to these external resources to be fetched via
370 HTTP, resulting in mixed contents warning in browsers.
371 (merge 5748558 ab/gitweb-use-same-scheme later to maint).
373 * When a line to be wrapped has a solid run of non space characters
374 whose length exactly is the wrap width, "git shortlog -w" failed
375 to add a newline after such a line.
377 * Command line completion leaked an unnecessary error message while
378 looking for possible matches with paths in <tree-ish>.
380 * Command line completion for "tcsh" emitted an unwanted space
381 after completing a single directory name.
383 * Command line completion code was inadvertently made incompatible with
384 older versions of bash by using a newer array notation.
386 * Some shells do not behave correctly when IFS is unset; work it
387 around by explicitly setting it to the default value.
389 * Some scripted programs written in Python did not get updated when
390 PYTHON_PATH changed.
391 (cherry-pick 96a4647fca54031974cd6ad1 later to maint).
393 * When autoconf is used, any build on a different commit always ran
394 "config.status --recheck" even when unnecessary.
396 * A fix was added to the build procedure to work around buggy
397 versions of ccache broke the auto-generation of dependencies, which
398 unfortunately is still relevant because some people use ancient
399 distros.
401 * The autoconf subsystem passed --mandir down to generated
402 config.mak.autogen but forgot to do the same for --htmldir.
403 (merge fc1c541 ct/autoconf-htmldir later to maint).
405 * We have been carrying a translated and long-unmaintained copy of an
406 old version of the tutorial; removed.
408 * t0050 had tests expecting failures from a bug that was fixed some
409 time ago.
411 * t4014, t9502 and t0200 tests had various portability issues that
412 broke on OpenBSD.
414 * t9020 and t3600 tests had various portability issues.
416 * t9200 runs "cvs init" on a directory that already exists, but a
417 platform can configure this fail for the current user (e.g. you
418 need to be in the cvsadmin group on NetBSD 6.0).
420 * t9020 and t9810 had a few non-portable shell script construct.
422 * Scripts to test bash completion was inherently flaky as it was
423 affected by whatever random things the user may have on $PATH.