| blob | 14b6c70ed6a64cae8d75fd0de863f66597c67da8 |
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 * "branch --delete" has "branch -d" but "push --delete" does not.
21 * "git blame" learned to produce the progress eye-candy when it takes
22 too much time before emitting the first line of the result.
24 * "git grep" can now be configured (or told from the command line)
25 how many threads to use when searching in the working tree files.
27 * Some "git notes" operations, e.g. "git log --notes=<note>", should
28 be able to read notes from any tree-ish that is shaped like a notes
29 tree, but the notes infrastructure required that the argument must
30 be a ref under refs/notes/. Loosen it to require a valid ref only
31 when the operation would update the notes (in which case we must
32 have a place to store the updated notes tree, iow, a ref).
34 * "git grep" by default does not fall back to its "--no-index"
35 behaviour outside a directory under Git's control (otherwise the
36 user may by mistake end up running a huge recursive search); with a
37 new configuration (set in $HOME/.gitconfig--by definition this
38 cannot be set in the config file per project), this safety can be
39 disabled.
41 * "git pull --rebase" has been extended to allow invoking
42 "rebase -i".
44 * "git p4" learned to cope with the type of a file getting changed.
46 * "git format-patch" learned to notice format.outputDirectory
47 configuration variable. This allows "-o <dir>" option to be
48 omitted on the command line if you always use the same directory in
49 your workflow.
51 * "interpret-trailers" has been taught to optionally update a file in
52 place, instead of always writing the result to the standard output.
54 * Many commands that read files that are expected to contain text
55 that is generated (or can be edited) by the end user to control
56 their behaviour (e.g. "git grep -f <filename>") have been updated
57 to be more tolerant to lines that are terminated with CRLF (they
58 used to treat such a line to contain payload that ends with CR,
59 which is usually not what the users expect).
61 * "git notes merge" used to limit the source of the merged notes tree
62 to somewhere under refs/notes/ hierarchy, which was too limiting
63 when inventing a workflow to exchange notes with remote
64 repositories using remote-tracking notes trees (located in e.g.
65 refs/remote-notes/ or somesuch).
67 * "git ls-files" learned a new "--eol" option to help diagnose
68 end-of-line problems.
70 * "ls-remote" learned an option to show which branch the remote
71 repository advertises as its primary by pointing its HEAD at.
73 * New http.proxyAuthMethod configuration variable can be used to
74 specify what authentication method to use, as a way to work around
75 proxies that do not give error response expected by libcurl when
76 CURLAUTH_ANY is used. Also, the codepath for proxy authentication
77 has been taught to use credential API to store the authentication
78 material in user's keyrings.
80 * Update the untracked cache subsystem and change its primary UI from
81 "git update-index" to "git config".
83 * There were a few "now I am doing this thing" progress messages in
84 the TCP connection code that can be triggered by setting a verbose
85 option internally in the code, but "git fetch -v" and friends never
86 passed the verbose option down to that codepath.
88 * Clean/smudge filters defined in a configuration file of lower
89 precedence can now be overridden to be a pass-through no-op by
90 setting the variable to an empty string.
92 * A new "<branch>^{/!-<pattern>}" notation can be used to name a
93 commit that is reachable from <branch> that does not match the
94 given <pattern>.
96 * The "user.useConfigOnly" configuration variable can be used to
97 force the user to always set user.email & user.name configuration
98 variables, serving as a reminder for those who work on multiple
99 projects and do not want to put these in their $HOME/.gitconfig.
101 * "git fetch" and friends that make network connections can now be
102 told to only use ipv4 (or ipv6).
104 * Some authentication methods do not need username or password, but
105 libcurl needs some hint that it needs to perform authentication.
106 Supplying an empty username and password string is a valid way to
107 do so, but you can set the http.[<url>.]emptyAuth configuration
108 variable to achieve the same, if you find it cleaner.
110 * You can now set http.[<url>.]pinnedpubkey to specify the pinned
111 public key when building with recent enough versions of libcURL.
113 * The configuration system has been taught to phrase where it found a
114 bad configuration variable in a better way in its error messages.
115 "git config" learnt a new "--show-origin" option to indicate where
116 the values come from.
118 * The "credential-cache" daemon process used to run in whatever
119 directory it happened to start in, but this made umount(2)ing the
120 filesystem that houses the repository harder; now the process
121 chdir()s to the directory that house its own socket on startup.
123 * When "git submodule update" did not result in fetching the commit
124 object in the submodule that is referenced by the superproject, the
125 command learned to retry another fetch, specifically asking for
126 that commit that may not be connected to the refs it usually
127 fetches.
129 * "git merge-recursive" learned "--no-renames" option to disable its
130 rename detection logic.
132 * Across the transition at around Git version 2.0, the user used to
133 get a pretty loud warning when running "git push" without setting
134 push.default configuration variable. We no longer warn, given that
135 the transition is over long time ago.
137 * README has been renamed to README.md and its contents got tweaked
138 slightly to make it easier on the eyes.
141 Performance, Internal Implementation, Development Support etc.
143 * Add a framework to spawn a group of processes in parallel, and use
144 it to run "git fetch --recurse-submodules" in parallel.
146 * A slight update to the Makefile to mark "phoney" targets
147 as such correctly.
149 * In-core storage of the reverse index for .pack files (which lets
150 you go from a pack offset to an object name) has been streamlined.
152 * d95138e6 (setup: set env $GIT_WORK_TREE when work tree is set, like
153 $GIT_DIR, 2015-06-26) attempted to work around a glitch in alias
154 handling by overwriting GIT_WORK_TREE environment variable to
155 affect subprocesses when set_git_work_tree() gets called, which
156 resulted in a rather unpleasant regression to "clone" and "init".
157 Try to address the same issue by always restoring the environment
158 and respawning the real underlying command when handling alias.
160 * The low-level code that is used to create symbolic references has
161 been updated to share more code with the code that deals with
162 normal references.
164 * strbuf_getline() and friends have been redefined to make it easier
165 to identify which callsite of (new) strbuf_getline_lf() should
166 allow and silently ignore carriage-return at the end of the line to
167 help users on DOSsy systems.
169 * "git shortlog" used to accumulate various pieces of information
170 regardless of what was asked to be shown in the final output. It
171 has been optimized by noticing what need not to be collected
172 (e.g. there is no need to collect the log messages when showing
173 only the number of changes).
175 * "git checkout $branch" (and other operations that share the same
176 underlying machinery) has been optimized.
178 * Automated tests in Travis CI environment has been optimized by
179 persisting runtime statistics of previous "prove" run, executing
180 tests that take longer before other ones; this reduces the total
181 wallclock time.
183 * Test scripts have been updated to remove assumptions that are not
184 portable between Git for POSIX and Git for Windows, or to skip ones
185 with expectations that are not satisfiable on Git for Windows.
187 * Some calls to strcpy(3) triggers a false warning from static
188 analysers that are less intelligent than humans, and reducing the
189 number of these false hits helps us notice real issues. A few
190 calls to strcpy(3) in test-path-utils that are already safe has
191 been rewritten to avoid false wanings.
193 * Some calls to strcpy(3) triggers a false warning from static
194 analysers that are less intelligent than humans, and reducing the
195 number of these false hits helps us notice real issues. A few
196 calls to strcpy(3) in "git rerere" that are already safe has been
197 rewritten to avoid false wanings.
199 * The "name_path" API was an attempt to reduce the need to construct
200 the full path out of a series of path components while walking a
201 tree hierarchy, but over time made less efficient because the path
202 needs to be flattened, e.g. to be compared with another path that
203 is already flat. The API has been removed and its users have been
204 rewritten to simplify the overall code complexity.
206 * Help those who debug http(s) part of the system.
207 (merge 0054045 sp/remote-curl-ssl-strerror later to maint).
209 * The internal API to interact with "remote.*" configuration
210 variables has been streamlined.
212 * The ref-filter's format-parsing code has been refactored, in
213 preparation for "branch --format" and friends.
215 * Traditionally, the tests that try commands that work on the
216 contents in the working tree were named with "worktree" in their
217 filenames, but with the recent addition of "git worktree"
218 subcommand, whose tests are also named similarly, it has become
219 harder to tell them apart. The traditional tests have been renamed
220 to use "work-tree" instead in an attempt to differentiate them.
221 (merge 5549029 mg/work-tree-tests later to maint).
223 * Many codepaths forget to check return value from git_config_set();
224 the function is made to die() to make sure we do not proceed when
225 setting a configuration variable failed.
226 (merge 3d18064 ps/config-error later to maint).
228 * Handling of errors while writing into our internal asynchronous
229 process has been made more robust, which reduces flakiness in our
230 tests.
231 (merge 43f3afc jk/epipe-in-async later to maint).
233 * There is a new DEVELOPER knob that enables many compiler warning
234 options in the Makefile.
236 * The way the test scripts configure the Apache web server has been
237 updated to work also for Apache 2.4 running on RedHat derived
238 distros.
240 * Out of maintenance gcc on OSX 10.6 fails to compile the code in
241 'master'; work it around by using clang by default on the platform.
244 Also contains various documentation updates and code clean-ups.
247 Fixes since v2.7
248 ----------------
250 Unless otherwise noted, all the fixes since v2.7 in the maintenance
251 track are contained in this release (see the maintenance releases'
252 notes for details).
254 * An earlier change in 2.5.x-era broke users' hooks and aliases by
255 exporting GIT_WORK_TREE to point at the root of the working tree,
256 interfering when they tried to use a different working tree without
257 setting GIT_WORK_TREE environment themselves.
259 * The "exclude_list" structure has the usual "alloc, nr" pair of
260 fields to be used by ALLOC_GROW(), but clear_exclude_list() forgot
261 to reset 'alloc' to 0 when it cleared 'nr' to discard the managed
262 array.
264 * Paths that have been told the index about with "add -N" are not
265 quite yet in the index, but a few commands behaved as if they
266 already are in a harmful way.
268 * "git send-email" was confused by escaped quotes stored in the alias
269 files saved by "mutt", which has been corrected.
271 * A few unportable C construct have been spotted by clang compiler
272 and have been fixed.
274 * The documentation has been updated to hint the connection between
275 the '--signoff' option and DCO.
277 * "git reflog" incorrectly assumed that all objects that used to be
278 at the tip of a ref must be commits, which caused it to segfault.
280 * The ignore mechanism saw a few regressions around untracked file
281 listing and sparse checkout selection areas in 2.7.0; the change
282 that is responsible for the regression has been reverted.
284 * Another try to improve the ignore mechanism that lets you say "this
285 is excluded" and then later say "oh, no, this part (that is a
286 subset of the previous part) is not excluded". This has still a
287 known limitation, though.
289 * Some codepaths used fopen(3) when opening a fixed path in $GIT_DIR
290 (e.g. COMMIT_EDITMSG) that is meant to be left after the command is
291 done. This however did not work well if the repository is set to
292 be shared with core.sharedRepository and the umask of the previous
293 user is tighter. They have been made to work better by calling
294 unlink(2) and retrying after fopen(3) fails with EPERM.
296 * Asking gitweb for a nonexistent commit left a warning in the server
297 log.
299 Somebody may want to follow this up with an additional test, perhaps?
300 IIRC, we do test that no Perl warnings are given to the server log,
301 so this should have been caught if our test coverage were good.
303 * "git rebase", unlike all other callers of "gc --auto", did not
304 ignore the exit code from "gc --auto".
306 * Many codepaths that run "gc --auto" before exiting kept packfiles
307 mapped and left the file descriptors to them open, which was not
308 friendly to systems that cannot remove files that are open. They
309 now close the packs before doing so.
311 * A recent optimization to filter-branch in v2.7.0 introduced a
312 regression when --prune-empty filter is used, which has been
313 corrected.
315 * The description for SANITY prerequisite the test suite uses has
316 been clarified both in the comment and in the implementation.
318 * "git tag" started listing a tag "foo" as "tags/foo" when a branch
319 named "foo" exists in the same repository; remove this unnecessary
320 disambiguation, which is a regression introduced in v2.7.0.
322 * The way "git svn" uses auth parameter was broken by Subversion
323 1.9.0 and later.
325 * The "split" subcommand of "git subtree" (in contrib/) incorrectly
326 skipped merges when it shouldn't, which was corrected.
328 * A few options of "git diff" did not work well when the command was
329 run from a subdirectory.
331 * The command line completion learned a handful of additional options
332 and command specific syntax.
334 * dirname() emulation has been added, as Msys2 lacks it.
336 * The underlying machinery used by "ls-files -o" and other commands
337 have been taught not to create empty submodule ref cache for a
338 directory that is not a submodule. This removes a ton of wasted
339 CPU cycles.
341 * "git worktree" had a broken code that attempted to auto-fix
342 possible inconsistency that results from end-users moving a
343 worktree to different places without telling Git (the original
344 repository needs to maintain backpointers to its worktrees, but
345 "mv" run by end-users who are not familiar with that fact will
346 obviously not adjust them), which actually made things worse
347 when triggered.
349 * The low-level merge machinery has been taught to use CRLF line
350 termination when inserting conflict markers to merged contents that
351 are themselves CRLF line-terminated.
353 * "git push --force-with-lease" has been taught to report if the push
354 needed to force (or fast-forwarded).
356 * The emulated "yes" command used in our test scripts has been
357 tweaked not to spend too much time generating unnecessary output
358 that is not used, to help those who test on Windows where it would
359 not stop until it fills the pipe buffer due to lack of SIGPIPE.
361 * The documentation for "git clean" has been corrected; it mentioned
362 that .git/modules/* are removed by giving two "-f", which has never
363 been the case.
365 * The vimdiff backend for "git mergetool" has been tweaked to arrange
366 and number buffers in the order that would match the expectation of
367 majority of people who read left to right, then top down and assign
368 buffers 1 2 3 4 "mentally" to local base remote merge windows based
369 on that order.
371 * "git show 'HEAD:Foo[BAR]Baz'" did not interpret the argument as a
372 rev, i.e. the object named by the the pathname with wildcard
373 characters in a tree object.
374 (merge aac4fac nd/dwim-wildcards-as-pathspecs later to maint).
376 * "git rev-parse --git-common-dir" used in the worktree feature
377 misbehaved when run from a subdirectory.
378 (merge 17f1365 nd/git-common-dir-fix later to maint).
380 * "git worktree add -B <branchname>" did not work.
382 * The "v(iew)" subcommand of the interactive "git am -i" command was
383 broken in 2.6.0 timeframe when the command was rewritten in C.
384 (merge 708b8cc jc/am-i-v-fix later to maint).
386 * "git merge-tree" used to mishandle "both sides added" conflict with
387 its own "create a fake ancestor file that has the common parts of
388 what both sides have added and do a 3-way merge" logic; this has
389 been updated to use the usual "3-way merge with an empty blob as
390 the fake common ancestor file" approach used in the rest of the
391 system.
392 (merge 907681e jk/no-diff-emit-common later to maint).
394 * The memory ownership rule of fill_textconv() API, which was a bit
395 tricky, has been documented a bit better.
396 (merge a64e6a4 jk/more-comments-on-textconv later to maint).
398 * Update various codepaths to avoid manually-counted malloc().
399 (merge 08c95df jk/tighten-alloc later to maint).
401 * The documentation did not clearly state that the 'simple' mode is
402 now the default for "git push" when push.default configuration is
403 not set.
404 (merge f6b1fb3 mm/push-simple-doc later to maint).
406 * Recent versions of GNU grep are pickier when their input contains
407 arbitrary binary data, which some of our tests uses. Rewrite the
408 tests to sidestep the problem.
409 (merge 3b1442d jk/grep-binary-workaround-in-test later to maint).
411 * A helper function "git submodule" uses since v2.7.0 to list the
412 modules that match the pathspec argument given to its subcommands
413 (e.g. "submodule add <repo> <path>") has been fixed.
414 (merge 2b56bb7 sb/submodule-module-list-fix later to maint).
416 * "git config section.var value" to set a value in per-repository
417 configuration file failed when it was run outside any repository,
418 but didn't say the reason correctly.
419 (merge 638fa62 js/config-set-in-non-repository later to maint).
421 * The code to read the pack data using the offsets stored in the pack
422 idx file has been made more carefully check the validity of the
423 data in the idx.
424 (merge 7465feb jk/pack-idx-corruption-safety later to maint).
426 * Other minor clean-ups and documentation updates
427 (merge f459823 ak/extract-argv0-last-dir-sep later to maint).
428 (merge 63ca1c0 ak/git-strip-extension-from-dashed-command later to maint).
429 (merge 4867f11 ps/plug-xdl-merge-leak later to maint).
430 (merge 4938686 dt/initial-ref-xn-commit-doc later to maint).
431 (merge 9537f21 ma/update-hooks-sample-typofix later to maint).