| blob | 5ec39bc300998a1d3add97026a46030cbc4b1926 |
1 Git 2.8 Release Notes
2 =====================
4 Backward compatibility note
5 ---------------------------
7 The rsync:// transport has been removed.
10 Updates since v2.7
11 ------------------
13 UI, Workflows & Features
15 * It turns out "git clone" over rsync transport has been broken when
16 the source repository has packed references for a long time, and
17 nobody noticed nor complained about it.
19 * "push" learned that its "--delete" option can be shortened to
20 "-d", just like "branch --delete" and "branch -d" are the same
21 thing.
23 * "git blame" learned to produce the progress eye-candy when it takes
24 too much time before emitting the first line of the result.
26 * "git grep" can now be configured (or told from the command line)
27 how many threads to use when searching in the working tree files.
29 * Some "git notes" operations, e.g. "git log --notes=<note>", should
30 be able to read notes from any tree-ish that is shaped like a notes
31 tree, but the notes infrastructure required that the argument must
32 be a ref under refs/notes/. Loosen it to require a valid ref only
33 when the operation would update the notes (in which case we must
34 have a place to store the updated notes tree, iow, a ref).
36 * "git grep" by default does not fall back to its "--no-index"
37 behaviour outside a directory under Git's control (otherwise the
38 user may by mistake end up running a huge recursive search); with a
39 new configuration (set in $HOME/.gitconfig--by definition this
40 cannot be set in the config file per project), this safety can be
41 disabled.
43 * "git pull --rebase" has been extended to allow invoking
44 "rebase -i".
46 * "git p4" learned to cope with the type of a file getting changed.
48 * "git format-patch" learned to notice format.outputDirectory
49 configuration variable. This allows "-o <dir>" option to be
50 omitted on the command line if you always use the same directory in
51 your workflow.
53 * "interpret-trailers" has been taught to optionally update a file in
54 place, instead of always writing the result to the standard output.
56 * Many commands that read files that are expected to contain text
57 that is generated (or can be edited) by the end user to control
58 their behaviour (e.g. "git grep -f <filename>") have been updated
59 to be more tolerant to lines that are terminated with CRLF (they
60 used to treat such a line to contain payload that ends with CR,
61 which is usually not what the users expect).
63 * "git notes merge" used to limit the source of the merged notes tree
64 to somewhere under refs/notes/ hierarchy, which was too limiting
65 when inventing a workflow to exchange notes with remote
66 repositories using remote-tracking notes trees (located in e.g.
67 refs/remote-notes/ or somesuch).
69 * "git ls-files" learned a new "--eol" option to help diagnose
70 end-of-line problems.
72 * "ls-remote" learned an option to show which branch the remote
73 repository advertises as its primary by pointing its HEAD at.
75 * New http.proxyAuthMethod configuration variable can be used to
76 specify what authentication method to use, as a way to work around
77 proxies that do not give error response expected by libcurl when
78 CURLAUTH_ANY is used. Also, the codepath for proxy authentication
79 has been taught to use credential API to store the authentication
80 material in user's keyrings.
82 * Update the untracked cache subsystem and change its primary UI from
83 "git update-index" to "git config".
85 * There were a few "now I am doing this thing" progress messages in
86 the TCP connection code that can be triggered by setting a verbose
87 option internally in the code, but "git fetch -v" and friends never
88 passed the verbose option down to that codepath.
90 * Clean/smudge filters defined in a configuration file of lower
91 precedence can now be overridden to be a pass-through no-op by
92 setting the variable to an empty string.
94 * A new "<branch>^{/!-<pattern>}" notation can be used to name a
95 commit that is reachable from <branch> that does not match the
96 given <pattern>.
98 * The "user.useConfigOnly" configuration variable can be used to
99 force the user to always set user.email & user.name configuration
100 variables, serving as a reminder for those who work on multiple
101 projects and do not want to put these in their $HOME/.gitconfig.
103 * "git fetch" and friends that make network connections can now be
104 told to only use ipv4 (or ipv6).
106 * Some authentication methods do not need username or password, but
107 libcurl needs some hint that it needs to perform authentication.
108 Supplying an empty username and password string is a valid way to
109 do so, but you can set the http.[<url>.]emptyAuth configuration
110 variable to achieve the same, if you find it cleaner.
112 * You can now set http.[<url>.]pinnedpubkey to specify the pinned
113 public key when building with recent enough versions of libcURL.
115 * The configuration system has been taught to phrase where it found a
116 bad configuration variable in a better way in its error messages.
117 "git config" learnt a new "--show-origin" option to indicate where
118 the values come from.
120 * The "credential-cache" daemon process used to run in whatever
121 directory it happened to start in, but this made umount(2)ing the
122 filesystem that houses the repository harder; now the process
123 chdir()s to the directory that house its own socket on startup.
125 * When "git submodule update" did not result in fetching the commit
126 object in the submodule that is referenced by the superproject, the
127 command learned to retry another fetch, specifically asking for
128 that commit that may not be connected to the refs it usually
129 fetches.
131 * "git merge-recursive" learned "--no-renames" option to disable its
132 rename detection logic.
134 * Across the transition at around Git version 2.0, the user used to
135 get a pretty loud warning when running "git push" without setting
136 push.default configuration variable. We no longer warn because the
137 transition was completed a long time ago.
139 * README has been renamed to README.md and its contents got tweaked
140 slightly to make it easier on the eyes.
143 Performance, Internal Implementation, Development Support etc.
145 * Add a framework to spawn a group of processes in parallel, and use
146 it to run "git fetch --recurse-submodules" in parallel.
148 * A slight update to the Makefile to mark ".PHONY" targets as such
149 correctly.
151 * In-core storage of the reverse index for .pack files (which lets
152 you go from a pack offset to an object name) has been streamlined.
154 * d95138e6 (setup: set env $GIT_WORK_TREE when work tree is set, like
155 $GIT_DIR, 2015-06-26) attempted to work around a glitch in alias
156 handling by overwriting GIT_WORK_TREE environment variable to
157 affect subprocesses when set_git_work_tree() gets called, which
158 resulted in a rather unpleasant regression to "clone" and "init".
159 Try to address the same issue by always restoring the environment
160 and respawning the real underlying command when handling alias.
162 * The low-level code that is used to create symbolic references has
163 been updated to share more code with the code that deals with
164 normal references.
166 * strbuf_getline() and friends have been redefined to make it easier
167 to identify which callsite of (new) strbuf_getline_lf() should
168 allow and silently ignore carriage-return at the end of the line to
169 help users on DOSsy systems.
171 * "git shortlog" used to accumulate various pieces of information
172 regardless of what was asked to be shown in the final output. It
173 has been optimized by noticing what need not to be collected
174 (e.g. there is no need to collect the log messages when showing
175 only the number of changes).
177 * "git checkout $branch" (and other operations that share the same
178 underlying machinery) has been optimized.
180 * Automated tests in Travis CI environment has been optimized by
181 persisting runtime statistics of previous "prove" run, executing
182 tests that take longer before other ones; this reduces the total
183 wallclock time.
185 * Test scripts have been updated to remove assumptions that are not
186 portable between Git for POSIX and Git for Windows, or to skip ones
187 with expectations that are not satisfiable on Git for Windows.
189 * Some calls to strcpy(3) triggers a false warning from static
190 analysers that are less intelligent than humans, and reducing the
191 number of these false hits helps us notice real issues. A few
192 calls to strcpy(3) in a couple of protrams that are already safe
193 has been rewritten to avoid false warnings.
195 * The "name_path" API was an attempt to reduce the need to construct
196 the full path out of a series of path components while walking a
197 tree hierarchy, but over time made less efficient because the path
198 needs to be flattened, e.g. to be compared with another path that
199 is already flat. The API has been removed and its users have been
200 rewritten to simplify the overall code complexity.
202 * Help those who debug http(s) part of the system.
203 (merge 0054045 sp/remote-curl-ssl-strerror later to maint).
205 * The internal API to interact with "remote.*" configuration
206 variables has been streamlined.
208 * The ref-filter's format-parsing code has been refactored, in
209 preparation for "branch --format" and friends.
211 * Traditionally, the tests that try commands that work on the
212 contents in the working tree were named with "worktree" in their
213 filenames, but with the recent addition of "git worktree"
214 subcommand, whose tests are also named similarly, it has become
215 harder to tell them apart. The traditional tests have been renamed
216 to use "work-tree" instead in an attempt to differentiate them.
217 (merge 5549029 mg/work-tree-tests later to maint).
219 * Many codepaths forget to check return value from git_config_set();
220 the function is made to die() to make sure we do not proceed when
221 setting a configuration variable failed.
222 (merge 3d18064 ps/config-error later to maint).
224 * Handling of errors while writing into our internal asynchronous
225 process has been made more robust, which reduces flakiness in our
226 tests.
227 (merge 43f3afc jk/epipe-in-async later to maint).
229 * There is a new DEVELOPER knob that enables many compiler warning
230 options in the Makefile.
232 * The way the test scripts configure the Apache web server has been
233 updated to work also for Apache 2.4 running on RedHat derived
234 distros.
236 * Out of maintenance gcc on OSX 10.6 fails to compile the code in
237 'master'; work it around by using clang by default on the platform.
239 * The "name_path" API was an attempt to reduce the need to construct
240 the full path out of a series of path components while walking a
241 tree hierarchy, but over time made less efficient because the path
242 needs to be flattened, e.g. to be compared with another path that
243 is already flat, in many cases. The API has been removed and its
244 users have been rewritten to simplify the overall code complexity.
245 This incidentally also closes some heap-corruption holes.
247 * Recent versions of GNU grep is pickier than before to decide if a
248 file is "binary" and refuse to give line-oriented hits when we
249 expect it to, unless explicitly told with "-a" option. As our
250 scripted Porcelains use sane_grep wrapper for line-oriented data,
251 even when the line may contain non-ASCII payload we took from
252 end-user data, use "grep -a" to implement sane_grep wrapper when
253 using an implementation of "grep" that takes the "-a" option.
257 Also contains various documentation updates and code clean-ups.
260 Fixes since v2.7
261 ----------------
263 Unless otherwise noted, all the fixes since v2.7 in the maintenance
264 track are contained in this release (see the maintenance releases'
265 notes for details).
267 * An earlier change in 2.5.x-era broke users' hooks and aliases by
268 exporting GIT_WORK_TREE to point at the root of the working tree,
269 interfering when they tried to use a different working tree without
270 setting GIT_WORK_TREE environment themselves.
272 * The "exclude_list" structure has the usual "alloc, nr" pair of
273 fields to be used by ALLOC_GROW(), but clear_exclude_list() forgot
274 to reset 'alloc' to 0 when it cleared 'nr' to discard the managed
275 array.
277 * Paths that have been told the index about with "add -N" are not
278 quite yet in the index, but a few commands behaved as if they
279 already are in a harmful way.
281 * "git send-email" was confused by escaped quotes stored in the alias
282 files saved by "mutt", which has been corrected.
284 * A few unportable C construct have been spotted by clang compiler
285 and have been fixed.
287 * The documentation has been updated to hint the connection between
288 the '--signoff' option and DCO.
290 * "git reflog" incorrectly assumed that all objects that used to be
291 at the tip of a ref must be commits, which caused it to segfault.
293 * The ignore mechanism saw a few regressions around untracked file
294 listing and sparse checkout selection areas in 2.7.0; the change
295 that is responsible for the regression has been reverted.
297 * Some codepaths used fopen(3) when opening a fixed path in $GIT_DIR
298 (e.g. COMMIT_EDITMSG) that is meant to be left after the command is
299 done. This however did not work well if the repository is set to
300 be shared with core.sharedRepository and the umask of the previous
301 user is tighter. They have been made to work better by calling
302 unlink(2) and retrying after fopen(3) fails with EPERM.
304 * Asking gitweb for a nonexistent commit left a warning in the server
305 log.
307 Somebody may want to follow this up with an additional test, perhaps?
308 IIRC, we do test that no Perl warnings are given to the server log,
309 so this should have been caught if our test coverage were good.
311 * "git rebase", unlike all other callers of "gc --auto", did not
312 ignore the exit code from "gc --auto".
314 * Many codepaths that run "gc --auto" before exiting kept packfiles
315 mapped and left the file descriptors to them open, which was not
316 friendly to systems that cannot remove files that are open. They
317 now close the packs before doing so.
319 * A recent optimization to filter-branch in v2.7.0 introduced a
320 regression when --prune-empty filter is used, which has been
321 corrected.
323 * The description for SANITY prerequisite the test suite uses has
324 been clarified both in the comment and in the implementation.
326 * "git tag" started listing a tag "foo" as "tags/foo" when a branch
327 named "foo" exists in the same repository; remove this unnecessary
328 disambiguation, which is a regression introduced in v2.7.0.
330 * The way "git svn" uses auth parameter was broken by Subversion
331 1.9.0 and later.
333 * The "split" subcommand of "git subtree" (in contrib/) incorrectly
334 skipped merges when it shouldn't, which was corrected.
336 * A few options of "git diff" did not work well when the command was
337 run from a subdirectory.
339 * The command line completion learned a handful of additional options
340 and command specific syntax.
342 * dirname() emulation has been added, as Msys2 lacks it.
344 * The underlying machinery used by "ls-files -o" and other commands
345 has been taught not to create empty submodule ref cache for a
346 directory that is not a submodule. This removes a ton of wasted
347 CPU cycles.
349 * "git worktree" had a broken code that attempted to auto-fix
350 possible inconsistency that results from end-users moving a
351 worktree to different places without telling Git (the original
352 repository needs to maintain backpointers to its worktrees, but
353 "mv" run by end-users who are not familiar with that fact will
354 obviously not adjust them), which actually made things worse
355 when triggered.
357 * The low-level merge machinery has been taught to use CRLF line
358 termination when inserting conflict markers to merged contents that
359 are themselves CRLF line-terminated.
361 * "git push --force-with-lease" has been taught to report if the push
362 needed to force (or fast-forwarded).
364 * The emulated "yes" command used in our test scripts has been
365 tweaked not to spend too much time generating unnecessary output
366 that is not used, to help those who test on Windows where it would
367 not stop until it fills the pipe buffer due to lack of SIGPIPE.
369 * The documentation for "git clean" has been corrected; it mentioned
370 that .git/modules/* are removed by giving two "-f", which has never
371 been the case.
373 * The vimdiff backend for "git mergetool" has been tweaked to arrange
374 and number buffers in the order that would match the expectation of
375 majority of people who read left to right, then top down and assign
376 buffers 1 2 3 4 "mentally" to local base remote merge windows based
377 on that order.
379 * "git show 'HEAD:Foo[BAR]Baz'" did not interpret the argument as a
380 rev, i.e. the object named by the the pathname with wildcard
381 characters in a tree object.
382 (merge aac4fac nd/dwim-wildcards-as-pathspecs later to maint).
384 * "git rev-parse --git-common-dir" used in the worktree feature
385 misbehaved when run from a subdirectory.
386 (merge 17f1365 nd/git-common-dir-fix later to maint).
388 * "git worktree add -B <branchname>" did not work.
390 * The "v(iew)" subcommand of the interactive "git am -i" command was
391 broken in 2.6.0 timeframe when the command was rewritten in C.
392 (merge 708b8cc jc/am-i-v-fix later to maint).
394 * "git merge-tree" used to mishandle "both sides added" conflict with
395 its own "create a fake ancestor file that has the common parts of
396 what both sides have added and do a 3-way merge" logic; this has
397 been updated to use the usual "3-way merge with an empty blob as
398 the fake common ancestor file" approach used in the rest of the
399 system.
400 (merge 907681e jk/no-diff-emit-common later to maint).
402 * The memory ownership rule of fill_textconv() API, which was a bit
403 tricky, has been documented a bit better.
404 (merge a64e6a4 jk/more-comments-on-textconv later to maint).
406 * Update various codepaths to avoid manually-counted malloc().
407 (merge 08c95df jk/tighten-alloc later to maint).
409 * The documentation did not clearly state that the 'simple' mode is
410 now the default for "git push" when push.default configuration is
411 not set.
412 (merge f6b1fb3 mm/push-simple-doc later to maint).
414 * Recent versions of GNU grep are pickier when their input contains
415 arbitrary binary data, which some of our tests uses. Rewrite the
416 tests to sidestep the problem.
417 (merge 3b1442d jk/grep-binary-workaround-in-test later to maint).
419 * A helper function "git submodule" uses since v2.7.0 to list the
420 modules that match the pathspec argument given to its subcommands
421 (e.g. "submodule add <repo> <path>") has been fixed.
422 (merge 2b56bb7 sb/submodule-module-list-fix later to maint).
424 * "git config section.var value" to set a value in per-repository
425 configuration file failed when it was run outside any repository,
426 but didn't say the reason correctly.
427 (merge 638fa62 js/config-set-in-non-repository later to maint).
429 * The code to read the pack data using the offsets stored in the pack
430 idx file has been made more carefully check the validity of the
431 data in the idx.
432 (merge 7465feb jk/pack-idx-corruption-safety later to maint).
434 * Other minor clean-ups and documentation updates
435 (merge f459823 ak/extract-argv0-last-dir-sep later to maint).
436 (merge 63ca1c0 ak/git-strip-extension-from-dashed-command later to maint).
437 (merge 4867f11 ps/plug-xdl-merge-leak later to maint).
438 (merge 4938686 dt/initial-ref-xn-commit-doc later to maint).
439 (merge 9537f21 ma/update-hooks-sample-typofix later to maint).