1 GIT v1.5.0 Release Notes
2 ========================
7 This section is for people who are upgrading from ancient
8 versions of git. Although all of the changes in this section
9 happened before the current v1.4.4 release, they are summarized
10 here in the v1.5.0 release notes for people who skipped earlier
13 As of git v1.5.0 there are some optional features that changes
14 the repository to allow data to be stored and transferred more
15 efficiently. These features are not enabled by default, as they
16 will make the repository unusable with older versions of git.
17 Specifically, the available options are:
19 - There is a configuration variable core.legacyheaders that
20 changes the format of loose objects so that they are more
21 efficient to pack and to send out of the repository over git
22 native protocol, since v1.4.2. However, loose objects
23 written in the new format cannot be read by git older than
24 that version; people fetching from your repository using
25 older clients over dumb transports (e.g. http) using older
26 versions of git will also be affected.
28 To let git use the new loose object format, you have to
29 set core.legacyheaders to false.
31 - Since v1.4.3, configuration repack.usedeltabaseoffset allows
32 packfile to be created in more space efficient format, which
33 cannot be read by git older than that version.
35 To let git use the new format for packfiles, you have to
36 set repack.usedeltabaseoffset to true.
38 The above two new features are not enabled by default and you
39 have to explicitly ask for them, because they make repositories
40 unreadable by older versions of git, and in v1.5.0 we still do
41 not enable them by default for the same reason. We will change
42 this default probably 1 year after 1.4.2's release, when it is
43 reasonable to expect everybody to have new enough version of
46 - 'git pack-refs' appeared in v1.4.4; this command allows tags
47 to be accessed much more efficiently than the traditional
48 'one-file-per-tag' format. Older git-native clients can
49 still fetch from a repository that packed and pruned refs
50 (the server side needs to run the up-to-date version of git),
51 but older dumb transports cannot. Packing of refs is done by
52 an explicit user action, either by use of "git pack-refs
53 --prune" command or by use of "git gc" command.
55 - 'git -p' to paginate anything -- many commands do pagination
56 by default on a tty. Introduced between v1.4.1 and v1.4.2;
57 this may surprise old timers.
59 - 'git archive' superseded 'git tar-tree' in v1.4.3;
61 - 'git cvsserver' was new invention in v1.3.0;
63 - 'git repo-config', 'git grep', 'git rebase' and 'gitk' were
64 seriously enhanced during v1.4.0 timeperiod.
66 - 'gitweb' became part of git.git during v1.4.0 timeperiod and
67 seriously modified since then.
69 - reflog is an v1.4.0 invention. This allows you to name a
70 revision that a branch used to be at (e.g. "git diff
71 master@{yesterday} master" allows you to see changes since
72 yesterday's tip of the branch).
75 Updates in v1.5.0 since v1.4.4 series
76 -------------------------------------
80 - git-add is to add contents to the index (aka "staging area"
81 for the next commit), whether the file the contents happen to
82 be is an existing one or a newly created one.
84 - git-add without any argument does not add everything
85 anymore. Use 'git-add .' instead. Also you can add
86 otherwise ignored files with an -f option.
88 - git-add tries to be more friendly to users by offering an
89 interactive mode ("git-add -i").
91 - git-commit <path> used to refuse to commit if <path> was
92 different between HEAD and the index (i.e. update-index was
93 used on it earlier). This check was removed.
95 - git-rm is much saner and safer. It is used to remove paths
96 from both the index file and the working tree, and makes sure
97 you are not losing any local modification before doing so.
99 - git-reset <tree> <paths>... can be used to revert index
100 entries for selected paths.
102 - git-update-index is much less visible. Many suggestions to
103 use the command in git output and documentation have now been
104 replaced by simpler commands such as "git add" or "git rm".
107 * Repository layout and objects transfer
109 - The data for origin repository is stored in the configuration
110 file $GIT_DIR/config, not in $GIT_DIR/remotes/, for newly
111 created clones. The latter is still supported and there is
112 no need to convert your existing repository if you are
113 already comfortable with your workflow with the layout.
115 - git-clone always uses what is known as "separate remote"
116 layout for a newly created repository with a working tree.
118 A repository with the separate remote layout starts with only
119 one default branch, 'master', to be used for your own
120 development. Unlike the traditional layout that copied all
121 the upstream branches into your branch namespace (while
122 renaming their 'master' to your 'origin'), the new layout
123 puts upstream branches into local "remote-tracking branches"
124 with their own namespace. These can be referenced with names
125 such as "origin/$upstream_branch_name" and are stored in
126 .git/refs/remotes rather than .git/refs/heads where normal
129 This layout keeps your own branch namespace less cluttered,
130 avoids name collision with your upstream, makes it possible
131 to automatically track new branches created at the remote
132 after you clone from it, and makes it easier to interact with
133 more than one remote repository (you can use "git remote" to
134 add other repositories to track). There might be some
137 * 'git branch' does not show the remote tracking branches.
138 It only lists your own branches. Use '-r' option to view
139 the tracking branches.
141 * If you are forking off of a branch obtained from the
142 upstream, you would have done something like 'git branch
143 my-next next', because traditional layout dropped the
144 tracking branch 'next' into your own branch namespace.
145 With the separate remote layout, you say 'git branch next
146 origin/next', which allows you to use the matching name
147 'next' for your own branch. It also allows you to track a
148 remote other than 'origin' (i.e. where you initially cloned
149 from) and fork off of a branch from there the same way
150 (e.g. "git branch mingw j6t/master").
152 Repositories initialized with the traditional layout continue
155 - New branches that appear on the origin side after a clone is
156 made are also tracked automatically. This is done with an
157 wildcard refspec "refs/heads/*:refs/remotes/origin/*", which
158 older git does not understand, so if you clone with 1.5.0,
159 you would need to downgrade remote.*.fetch in the
160 configuration file to specify each branch you are interested
161 in individually if you plan to fetch into the repository with
162 older versions of git (but why would you?).
164 - Similarly, wildcard refspec "refs/heads/*:refs/remotes/me/*"
165 can be given to "git-push" command to update the tracking
166 branches that is used to track the repository you are pushing
167 from on the remote side.
169 - git-branch and git-show-branch know remote tracking branches
170 (use the command line switch "-r" to list only tracked branches).
172 - git-push can now be used to delete a remote branch or a tag.
173 This requires the updated git on the remote side (use "git
174 push <remote> :refs/heads/<branch>" to delete "branch").
176 - git-push more aggressively keeps the transferred objects
177 packed. Earlier we recommended to monitor amount of loose
178 objects and repack regularly, but you should repack when you
179 accumulated too many small packs this way as well. Updated
180 git-count-objects helps you with this.
182 - git-fetch also more aggressively keeps the transferred objects
183 packed. This behavior of git-push and git-fetch can be
184 tweaked with a single configuration transfer.unpacklimit (but
185 usually there should not be any need for a user to tweak it).
187 - A new command, git-remote, can help you manage your remote
188 tracking branch definitions.
190 - You may need to specify explicit paths for upload-pack and/or
191 receive-pack due to your ssh daemon configuration on the
192 other end. This can now be done via remote.*.uploadpack and
193 remote.*.receivepack configuration.
198 - Certain commands change their behavior in a bare repository
199 (i.e. a repository without associated working tree). We use
200 a fairly conservative heuristic (if $GIT_DIR is ".git", or
201 ends with "/.git", the repository is not bare) to decide if a
202 repository is bare, but "core.bare" configuration variable
203 can be used to override the heuristic when it misidentifies
206 - git-fetch used to complain updating the current branch but
207 this is now allowed for a bare repository. So is the use of
208 'git-branch -f' to update the current branch.
210 - Porcelain-ish commands that require a working tree refuses to
211 work in a bare repository.
216 - Reflog records the history from the view point of the local
217 repository. In other words, regardless of the real history,
218 the reflog shows the history as seen by one particular
219 repository (this enables you to ask "what was the current
220 revision in _this_ repository, yesterday at 1pm?"). This
221 facility is enabled by default for repositories with working
222 trees, and can be accessed with the "branch@{time}" and
223 "branch@{Nth}" notation.
225 - "git show-branch" learned showing the reflog data with the
226 new -g option. "git log" has -g option to view reflog
227 entries in a more verbose manner.
229 - git-branch knows how to rename branches and moves existing
230 reflog data from the old branch to the new one.
232 - In addition to the reflog support in v1.4.4 series, HEAD
233 reference maintains its own log. "HEAD@{5.minutes.ago}"
234 means the commit you were at 5 minutes ago, which takes
235 branch switching into account. If you want to know where the
236 tip of your current branch was at 5 minutes ago, you need to
237 explicitly say its name (e.g. "master@{5.minutes.ago}") or
238 omit the refname altogether i.e. "@{5.minutes.ago}".
240 - The commits referred to by reflog entries are now protected
241 against pruning. The new command "git reflog expire" can be
242 used to truncate older reflog entries and entries that refer
243 to commits that have been pruned away previously with older
246 Existing repositories that have been using reflog may get
247 complaints from fsck-objects and may not be able to run
248 git-repack, if you had run git-prune from older git; please
249 run "git reflog expire --stale-fix --all" first to remove
250 reflog entries that refer to commits that are no longer in
251 the repository when that happens.
256 - We used to say "old commits are retrievable using reflog and
257 'master@{yesterday}' syntax as long as you haven't run
258 git-prune". We no longer have to say the latter half of the
259 above sentence, as git-prune does not remove things reachable
262 - There is a toplevel garbage collector script, 'git-gc', that
263 runs periodic cleanup functions, including 'git-repack -a -d',
264 'git-reflog expire', 'git-pack-refs --prune', and 'git-rerere
267 - The output from fsck ("fsck-objects" is called just "fsck"
268 now, but the old name continues to work) was needlessly
269 alarming in that it warned missing objects that are reachable
270 only from dangling objects. This has been corrected and the
271 output is much more useful.
276 - You can use 'git-checkout' to check out an arbitrary revision
277 or a tag as well, instead of named branches. This will
278 dissociate your HEAD from the branch you are currently on.
280 A typical use of this feature is to "look around". E.g.
282 $ git checkout v2.6.16
283 ... compile, test, etc.
284 $ git checkout v2.6.17
285 ... compile, test, etc.
287 - After detaching your HEAD, you can go back to an existing
288 branch with usual "git checkout $branch". Also you can
289 start a new branch using "git checkout -b $newbranch" to
290 start a new branch at that commit.
292 - You can even pull from other repositories, make merges and
293 commits while your HEAD is detached. Also you can use "git
294 reset" to jump to arbitrary commit, while still keeping your
297 Remember that a detached state is volatile, i.e. it will be forgotten
298 as soon as you move away from it with the checkout or reset command,
299 unless a branch is created from it as mentioned above. It is also
300 possible to rescue a lost detached state from the HEAD reflog.
305 - Repositories with hundreds of tags have been paying large
306 overhead, both in storage and in runtime, due to the
307 traditional one-ref-per-file format. A new command,
308 git-pack-refs, can be used to "pack" them in more efficient
309 representation (you can let git-gc do this for you).
311 - Clones and fetches over dumb transports are now aware of
312 packed refs and can download from repositories that use
318 - configuration related to color setting are consolidated under
319 color.* namespace (older diff.color.*, status.color.* are
322 - 'git-repo-config' command is accessible as 'git-config' now.
327 - git-describe uses better criteria to pick a base ref. It
328 used to pick the one with the newest timestamp, but now it
329 picks the one that is topologically the closest (that is,
330 among ancestors of commit C, the ref T that has the shortest
331 output from "git-rev-list T..C" is chosen).
333 - git-describe gives the number of commits since the base ref
334 between the refname and the hash suffix. E.g. the commit one
335 before v2.6.20-rc6 in the kernel repository is:
337 v2.6.20-rc5-306-ga21b069
339 which tells you that its object name begins with a21b069,
340 v2.6.20-rc5 is an ancestor of it (meaning, the commit
341 contains everything -rc5 has), and there are 306 commits
344 - git-describe with --abbrev=0 can be used to show only the
345 name of the base ref.
347 - git-blame learned a new option, --incremental, that tells it
348 to output the blames as they are assigned. A sample script
349 to use it is also included as contrib/blameview.
351 - git-blame starts annotating from the working tree by default.
354 * Less external dependency
356 - We no longer require the "merge" program from the RCS suite.
357 All 3-way file-level merges are now done internally.
359 - The original implementation of git-merge-recursive which was
360 in Python has been removed; we have a C implementation of it
363 - git-shortlog is no longer a Perl script. It no longer
364 requires output piped from git-log; it can accept revision
365 parameters directly on the command line.
370 - We have always encouraged the commit message to be encoded in
371 UTF-8, but the users are allowed to use legacy encoding as
372 appropriate for their projects. This will continue to be the
373 case. However, a non UTF-8 commit encoding _must_ be
374 explicitly set with i18n.commitencoding in the repository
375 where a commit is made; otherwise git-commit-tree will
376 complain if the log message does not look like a valid UTF-8
379 - The value of i18n.commitencoding in the originating
380 repository is recorded in the commit object on the "encoding"
381 header, if it is not UTF-8. git-log and friends notice this,
382 and reencodes the message to the log output encoding when
383 displaying, if they are different. The log output encoding
384 is determined by "git log --encoding=<encoding>",
385 i18n.logoutputencoding configuration, or i18n.commitencoding
386 configuration, in the decreasing order of preference, and
389 - Tools for e-mailed patch application now default to -u
390 behavior; i.e. it always re-codes from the e-mailed encoding
391 to the encoding specified with i18n.commitencoding. This
392 unfortunately forces projects that have happily been using a
393 legacy encoding without setting i18n.commitencoding to set
394 the configuration, but taken with other improvement, please
395 excuse us for this very minor one-time inconvenience.
400 - See the above I18n section.
402 - git-format-patch now enables --binary without being asked.
403 git-am does _not_ default to it, as sending binary patch via
404 e-mail is unusual and is harder to review than textual
405 patches and it is prudent to require the person who is
406 applying the patch to explicitly ask for it.
408 - The default suffix for git-format-patch output is now ".patch",
409 not ".txt". This can be changed with --suffix=.txt option,
410 or setting the config variable "format.suffix" to ".txt".
413 * Foreign SCM interfaces
415 - git-svn now requires the Perl SVN:: libraries, the
416 command-line backend was too slow and limited.
418 - the 'commit' subcommand of git-svn has been renamed to
419 'set-tree', and 'dcommit' is the recommended replacement for
422 - git fast-import backend.
427 - Quite a lot of documentation updates.
429 - Bash completion scripts have been updated heavily.
431 - Better error messages for often used Porcelainish commands.
433 - Git GUI. This is a simple Tk based graphical interface for
434 common Git operations.
439 - We used to assume that we can mmap the whole packfile while
440 in use, but with a large project this consumes huge virtual
441 memory space and truly huge ones would not fit in the
442 userland address space on 32-bit platforms. We now mmap huge
443 packfile in pieces to avoid this problem.
448 - There is a partial support for 'shallow' repositories that
449 keeps only recent history. A 'shallow clone' is created by
450 specifying how deep that truncated history should be
451 (e.g. "git clone --depth 5 git://some.where/repo.git").
453 Currently a shallow repository has number of limitations:
455 - Cloning and fetching _from_ a shallow clone are not
456 supported (nor tested -- so they might work by accident but
457 they are not expected to).
459 - Pushing from nor into a shallow clone are not expected to
462 - Merging inside a shallow repository would work as long as a
463 merge base is found in the recent history, but otherwise it
464 will be like merging unrelated histories and may result in
467 but this would be more than adequate for people who want to
468 look at near the tip of a big project with a deep history and
469 send patches in e-mail format.
projects / git / blob