Merge branch 'na/no-http-test-in-the-middle' into maint
[git/git.git] / Documentation / RelNotes / 1.9.0.txt
1 Git v1.9.0 Release Notes
2 ========================
4 Backward compatibility notes
5 ----------------------------
7 "git submodule foreach $cmd $args" used to treat "$cmd $args" the same
8 way "ssh" did, concatenating them into a single string and letting the
9 shell unquote. Careless users who forget to sufficiently quote $args
10 get their argument split at $IFS whitespaces by the shell, and got
11 unexpected results due to this. Starting from this release, the
12 command line is passed directly to the shell, if it has an argument.
14 Read-only support for experimental loose-object format, in which users
15 could optionally choose to write their loose objects for a short
16 while between v1.4.3 and v1.5.3 era, has been dropped.
18 The meanings of the "--tags" option to "git fetch" has changed; the
19 command fetches tags _in addition to_ what is fetched by the same
20 command line without the option.
22 The way "git push $there $what" interprets the $what part given on the
23 command line, when it does not have a colon that explicitly tells us
24 what ref at the $there repository is to be updated, has been enhanced.
26 A handful of ancient commands that have long been deprecated are
27 finally gone (repo-config, tar-tree, lost-found, and peek-remote).
30 Backward compatibility notes (for Git 2.0.0)
31 --------------------------------------------
33 When "git push [$there]" does not say what to push, we have used the
34 traditional "matching" semantics so far (all your branches were sent
35 to the remote as long as there already are branches of the same name
36 over there). In Git 2.0, the default will change to the "simple"
37 semantics, which pushes:
39 - only the current branch to the branch with the same name, and only
40 when the current branch is set to integrate with that remote
41 branch, if you are pushing to the same remote as you fetch from; or
43 - only the current branch to the branch with the same name, if you
44 are pushing to a remote that is not where you usually fetch from.
46 Use the user preference configuration variable "push.default" to
47 change this. If you are an old-timer who is used to the "matching"
48 semantics, you can set the variable to "matching" to keep the
49 traditional behaviour. If you want to live in the future early, you
50 can set it to "simple" today without waiting for Git 2.0.
52 When "git add -u" (and "git add -A") is run inside a subdirectory and
53 does not specify which paths to add on the command line, it
54 will operate on the entire tree in Git 2.0 for consistency
55 with "git commit -a" and other commands. There will be no
56 mechanism to make plain "git add -u" behave like "git add -u .".
57 Current users of "git add -u" (without a pathspec) should start
58 training their fingers to explicitly say "git add -u ."
59 before Git 2.0 comes. A warning is issued when these commands are
60 run without a pathspec and when you have local changes outside the
61 current directory, because the behaviour in Git 2.0 will be different
62 from today's version in such a situation.
64 In Git 2.0, "git add <path>" will behave as "git add -A <path>", so
65 that "git add dir/" will notice paths you removed from the directory
66 and record the removal. Versions before Git 2.0, including this
67 release, will keep ignoring removals, but the users who rely on this
68 behaviour are encouraged to start using "git add --ignore-removal <path>"
69 now before 2.0 is released.
71 The default prefix for "git svn" will change in Git 2.0. For a long
72 time, "git svn" created its remote-tracking branches directly under
73 refs/remotes, but it will place them under refs/remotes/origin/ unless
74 it is told otherwise with its --prefix option.
77 Updates since v1.8.5
78 --------------------
80 Foreign interfaces, subsystems and ports.
82 * The HTTP transport, when talking GSS-Negotiate, uses "100
83 Continue" response to avoid having to rewind and resend a large
84 payload, which may not be always doable.
86 * Various bugfixes to remote-bzr and remote-hg (in contrib/).
88 * The build procedure is aware of MirBSD now.
90 * Various "git p4", "git svn" and "gitk" updates.
93 UI, Workflows & Features
95 * Fetching from a shallowly-cloned repository used to be forbidden,
96 primarily because the codepaths involved were not carefully vetted
97 and we did not bother supporting such usage. This release attempts
98 to allow object transfer out of a shallowly-cloned repository in a
99 more controlled way (i.e. the receiver becomes a shallow repository
100 with a truncated history).
102 * Just like we give a reasonable default for "less" via the LESS
103 environment variable, we now specify a reasonable default for "lv"
104 via the "LV" environment variable when spawning the pager.
106 * Two-level configuration variable names in "branch.*" and "remote.*"
107 hierarchies, whose variables are predominantly three-level, were
108 not completed by hitting a <TAB> in bash and zsh completions.
110 * Fetching a 'frotz' branch with "git fetch", while a 'frotz/nitfol'
111 remote-tracking branch from an earlier fetch was still there, would
112 error out, primarily because the command was not told that it is
113 allowed to lose any information on our side. "git fetch --prune"
114 now can be used to remove 'frotz/nitfol' to make room for fetching and
115 storing the 'frotz' remote-tracking branch.
117 * "diff.orderfile=<file>" configuration variable can be used to
118 pretend as if the "-O<file>" option were given from the command
119 line of "git diff", etc.
121 * The negative pathspec syntax allows "git log -- . ':!dir'" to tell
122 us "I am interested in everything but 'dir' directory".
124 * "git difftool" shows how many different paths there are in total,
125 and how many of them have been shown so far, to indicate progress.
127 * "git push origin master" used to push our 'master' branch to update
128 the 'master' branch at the 'origin' repository. This has been
129 enhanced to use the same ref mapping "git push origin" would use to
130 determine what ref at the 'origin' to be updated with our 'master'.
131 For example, with this configuration
133 [remote "origin"]
134 push = refs/heads/*:refs/review/*
136 that would cause "git push origin" to push out our local branches
137 to corresponding refs under refs/review/ hierarchy at 'origin',
138 "git push origin master" would update 'refs/review/master' over
139 there. Alternatively, if push.default is set to 'upstream' and our
140 'master' is set to integrate with 'topic' from the 'origin' branch,
141 running "git push origin" while on our 'master' would update their
142 'topic' branch, and running "git push origin master" while on any
143 of our branches does the same.
145 * "gitweb" learned to treat ref hierarchies other than refs/heads as
146 if they are additional branch namespaces (e.g. refs/changes/ in
147 Gerrit).
149 * "git for-each-ref --format=..." learned a few formatting directives;
150 e.g. "%(color:red)%(HEAD)%(color:reset) %(refname:short) %(subject)".
152 * The command string given to "git submodule foreach" is passed
153 directly to the shell, without being eval'ed. This is a backward
154 incompatible change that may break existing users.
156 * "git log" and friends learned the "--exclude=<glob>" option, to
157 allow people to say "list history of all branches except those that
158 match this pattern" with "git log --exclude='*/*' --branches".
160 * "git rev-parse --parseopt" learned a new "--stuck-long" option to
161 help scripts parse options with an optional parameter.
163 * The "--tags" option to "git fetch" no longer tells the command to
164 fetch _only_ the tags. It instead fetches tags _in addition to_
165 what are fetched by the same command line without the option.
168 Performance, Internal Implementation, etc.
170 * When parsing a 40-hex string into the object name, the string is
171 checked to see if it can be interpreted as a ref so that a warning
172 can be given for ambiguity. The code kicked in even when the
173 core.warnambiguousrefs is set to false to squelch this warning, in
174 which case the cycles spent to look at the ref namespace were an
175 expensive no-op, as the result was discarded without being used.
177 * The naming convention of the packfiles has been updated; it used to
178 be based on the enumeration of names of the objects that are
179 contained in the pack, but now it also depends on how the packed
180 result is represented---packing the same set of objects using
181 different settings (or delta order) would produce a pack with
182 different name.
184 * "git diff --no-index" mode used to unnecessarily attempt to read
185 the index when there is one.
187 * The deprecated parse-options macro OPT_BOOLEAN has been removed;
188 use OPT_BOOL or OPT_COUNTUP in new code.
190 * A few duplicate implementations of prefix/suffix string comparison
191 functions have been unified to starts_with() and ends_with().
193 * The new PERLLIB_EXTRA makefile variable can be used to specify
194 additional directories Perl modules (e.g. the ones necessary to run
195 git-svn) are installed on the platform when building.
197 * "git merge-base" learned the "--fork-point" mode, that implements
198 the same logic used in "git pull --rebase" to find a suitable fork
199 point out of the reflog entries for the remote-tracking branch the
200 work has been based on. "git rebase" has the same logic that can be
201 triggered with the "--fork-point" option.
203 * A third-party "receive-pack" (the responder to "git push") can
204 advertise the "no-thin" capability to tell "git push" not to use
205 the thin-pack optimization. Our receive-pack has always been
206 capable of accepting and fattening a thin-pack, and will continue
207 not to ask "git push" to use a non-thin pack.
210 Also contains various documentation updates and code clean-ups.
213 Fixes since v1.8.5
214 ------------------
216 Unless otherwise noted, all the fixes since v1.8.5 in the maintenance
217 track are contained in this release (see the maintenance releases' notes
218 for details).
220 * The pathspec matching code, while comparing two trees (e.g. "git
221 diff A B -- path1 path2") was too aggressive and failed to match
222 some paths when multiple pathspecs were involved.
224 * "git repack --max-pack-size=8g" stopped being parsed correctly when
225 the command was reimplemented in C.
227 * An earlier update in v1.8.4.x to "git rev-list --objects" with
228 negative ref had a performance regression.
229 (merge 200abe7 jk/mark-edges-uninteresting later to maint).
231 * A recent update to "git send-email" broke platforms where
232 /etc/ssl/certs/ directory exists but cannot be used as SSL_ca_path
233 (e.g. Fedora rawhide).
235 * A handful of bugs around interpreting $branch@{upstream} notation
236 and its lookalike, when $branch part has interesting characters,
237 e.g. "@", and ":", have been fixed.
239 * "git clone" would fail to clone from a repository that has a ref
240 directly under "refs/", e.g. "refs/stash", because different
241 validation paths do different things on such a refname. Loosen the
242 client side's validation to allow such a ref.
244 * "git log --left-right A...B" lost the "leftness" of commits
245 reachable from A when A is a tag as a side effect of a recent
246 bugfix. This is a regression in 1.8.4.x series.
248 * documentations to "git pull" hinted there is an "-m" option because
249 it incorrectly shared the documentation with "git merge".
251 * "git diff A B submod" and "git diff A B submod/" ought to have done
252 the same for a submodule "submod", but didn't.
254 * "git clone $origin foo\bar\baz" on Windows failed to create the
255 leading directories (i.e. a moral-equivalent of "mkdir -p").
257 * "submodule.*.update=checkout", when propagated from .gitmodules to
258 .git/config, turned into a "submodule.*.update=none", which did not
259 make much sense.
260 (merge efa8fd7 fp/submodule-checkout-mode later to maint).
262 * The implementation of 'git stash $cmd "stash@{...}"' did not quote
263 the stash argument properly and left it split at IFS whitespace.
265 * The "--[no-]informative-errors" options to "git daemon" were parsed
266 a bit too loosely, allowing any other string after these option
267 names.
269 * There is no reason to have a hardcoded upper limit for the number of
270 parents of an octopus merge, created via the graft mechanism, but
271 there was.
273 * The basic test used to leave unnecessary trash directories in the
274 t/ directory.
275 (merge 738a8be jk/test-framework-updates later to maint).
277 * "git merge-base --octopus" used to leave cleaning up suboptimal
278 result to the caller, but now it does the clean-up itself.
280 * A "gc" process running as a different user should be able to stop a
281 new "gc" process from starting, but it didn't.
283 * An earlier "clean-up" introduced an unnecessary memory leak.
285 * "git add -A" (no other arguments) in a totally empty working tree
286 used to emit an error.
288 * "git log --decorate" did not handle a tag pointed by another tag
289 nicely.
291 * When we figure out how many file descriptors to allocate for
292 keeping packfiles open, a system with non-working getrlimit() could
293 cause us to die(), but because we make this call only to get a
294 rough estimate of how many are available and we do not even attempt
295 to use up all available file descriptors ourselves, it is nicer to
296 fall back to a reasonable low value rather than dying.
298 * read_sha1_file(), that is the workhorse to read the contents given
299 an object name, honoured object replacements, but there was no
300 corresponding mechanism to sha1_object_info() that was used to
301 obtain the metainfo (e.g. type & size) about the object. This led
302 callers to weird inconsistencies.
303 (merge 663a856 cc/replace-object-info later to maint).
305 * "git cat-file --batch=", an admittedly useless command, did not
306 behave very well.
308 * "git rev-parse <revs> -- <paths>" did not implement the usual
309 disambiguation rules the commands in the "git log" family used in
310 the same way.
312 * "git mv A B/", when B does not exist as a directory, should error
313 out, but it didn't.
315 * A workaround to an old bug in glibc prior to glibc 2.17 has been
316 retired; this would remove a side effect of the workaround that
317 corrupts system error messages in non-C locales.
319 * SSL-related options were not passed correctly to underlying socket
320 layer in "git send-email".
322 * "git commit -v" appends the patch to the log message before
323 editing, and then removes the patch when the editor returned
324 control. However, the patch was not stripped correctly when the
325 first modified path was a submodule.
327 * "git fetch --depth=0" was a no-op, and was silently ignored.
328 Diagnose it as an error.
330 * Remote repository URLs expressed in scp-style host:path notation are
331 parsed more carefully (e.g. "foo/bar:baz" is local, "[::1]:/~user" asks
332 to connect to user's home directory on host at address ::1.
334 * "git diff -- ':(icase)makefile'" was unnecessarily rejected at the
335 command line parser.
337 * "git cat-file --batch-check=ok" did not check the existence of
338 the named object.
340 * "git am --abort" sometimes complained about not being able to write
341 a tree with an 0{40} object in it.
343 * Two processes creating loose objects at the same time could have
344 failed unnecessarily when the name of their new objects started
345 with the same byte value, due to a race condition.