Git 2.7-rc3
[git/git.git] / Documentation / RelNotes / 2.7.0.txt
1 Git 2.7 Release Notes
2 =====================
4 Updates since v2.6
5 ------------------
7 UI, Workflows & Features
9 * The appearance of "gitk", particularly on high DPI monitors, have
10 been improved. "gitk" also comes with an undated translation for
11 Swedish and Japanese.
13 * "git remote" learned "get-url" subcommand to show the URL for a
14 given remote name used for fetching and pushing.
16 * There was no way to defeat a configured rebase.autostash variable
17 from the command line, as "git rebase --no-autostash" was missing.
19 * "git log --date=local" used to only show the normal (default)
20 format in the local timezone. The command learned to take 'local'
21 as an instruction to use the local timezone with other formats,
23 * The refs used during a "git bisect" session is now per-worktree so
24 that independent bisect sessions can be done in different worktrees
25 created with "git worktree add".
27 * Users who are too busy to type three extra keystrokes to ask for
28 "git stash show -p" can now set stash.showPatch configuration
29 variable to true to always see the actual patch, not just the list
30 of paths affected with feel for the extent of damage via diffstat.
32 * "quiltimport" allows to specify the series file by honoring the
33 $QUILT_SERIES environment and also --series command line option.
35 * The use of 'good/bad' in "git bisect" made it confusing to use when
36 hunting for a state change that is not a regression (e.g. bugfix).
37 The command learned 'old/new' and then allows the end user to
38 say e.g. "bisect start --term-old=fast --term-new=slow" to find a
39 performance regression.
41 * "git interpret-trailers" can now run outside of a Git repository.
43 * "git p4" learned to reencode the pathname it uses to communicate
44 with the p4 depot with a new option.
46 * Give progress meter to "git filter-branch".
48 * Allow a later "!/abc/def" to override an earlier "/abc" that
49 appears in the same .gitignore file to make it easier to express
50 "everything in /abc directory is ignored, except for ...".
52 * Teach "git p4" to send large blobs outside the repository by
53 talking to Git LFS.
55 * Prepare for Git on-disk repository representation to undergo
56 backward incompatible changes by introducing a new repository
57 format version "1", with an extension mechanism.
59 * "git worktree" learned a "list" subcommand.
61 * "git clone --dissociate" learned that it can be used even when
62 "--reference" was not used at the same time.
64 * "git blame" learnt to take "--first-parent" and "--reverse" at the
65 same time when it makes sense.
67 * "git checkout" did not follow the usual "--[no-]progress"
68 convention and implemented only "--quiet" that is essentially
69 a superset of "--no-progress". Extend the command to support the
70 usual "--[no-]progress".
72 * The semantics of transfer.hideRefs configuration variable have been
73 extended to work better with the ref "namespace" feature that lets
74 you throw unrelated bunches of repositories in a single physical
75 repository and virtually serve them as separate ones.
77 * send-email config variables whose values are pathnames now go
78 through the ~username/ expansion.
80 * bash completion learnt to TAB-complete recipient addresses given
81 to send-email.
83 * The credential-cache daemon can be told to ignore SIGHUP to work
84 around issue when running Git from inside emacs.
86 * "git push" learned new configuration for doing "--recurse-submodules"
87 on each push.
89 * "format-patch" has learned a new option to zero-out the commit
90 object name on the mbox "From " line.
93 Performance, Internal Implementation, Development Support etc.
95 * The infrastructure to rewrite "git submodule" in C is being built
96 incrementally. Let's polish these early parts well enough and make
97 them graduate to 'next' and 'master', so that the more involved
98 follow-up can start cooking on a solid ground.
100 * Some features from "git tag -l" and "git branch -l" have been made
101 available to "git for-each-ref" so that eventually the unified
102 implementation can be shared across all three. The version merged
103 to the 'master' branch earlier had a performance regression in "tag
104 --contains", which has since been corrected.
106 * Because "test_when_finished" in our test framework queues the
107 clean-up tasks to be done in a shell variable, it should not be
108 used inside a subshell. Add a mechanism to allow 'bash' to catch
109 such uses, and fix the ones that were found.
111 * The debugging infrastructure for pkt-line based communication has
112 been improved to mark the side-band communication specifically.
114 * Update "git branch" that list existing branches, using the
115 ref-filter API that is shared with "git tag" and "git
116 for-each-ref".
118 * The test for various line-ending conversions has been enhanced.
120 * A few test scripts around "git p4" have been improved for
121 portability.
123 * Many allocations that is manually counted (correctly) that are
124 followed by strcpy/sprintf have been replaced with a less error
125 prone constructs such as xstrfmt.
127 * The internal stripspace() function has been moved to where it
128 logically belongs to, i.e. strbuf API, and the command line parser
129 of "git stripspace" has been updated to use the parse_options API.
131 * "git am" used to spawn "git mailinfo" via run_command() API once
132 per each patch, but learned to make a direct call to mailinfo()
133 instead.
135 * The implementation of "git mailinfo" was refactored so that a
136 mailinfo() function can be directly called from inside a process.
138 * With a "debug" helper, debugging of a single "git" invocation in
139 our test scripts has become a lot easier.
141 * The "configure" script did not test for -lpthread correctly, which
142 upset some linkers.
144 * Cross completed task off of subtree project's todo list.
146 * Test cleanups for the subtree project.
148 * Clean up style in an ancient test t9300.
150 * Work around some test flakiness with p4d.
152 * Fsck did not correctly detect a NUL-truncated header in a tag.
154 * Use a safer behavior when we hit errors verifying remote certificates.
156 * Speed up filter-branch for cases where we only care about rewriting
157 commits, not tree data.
159 * The parse-options API has been updated to make "-h" command line
160 option work more consistently in all commands.
162 * "git svn rebase/mkdirs" got optimized by keeping track of empty
163 directories better.
165 * Fix some racy client/server tests by treating SIGPIPE the same as a
166 normal non-zero exit.
168 * The necessary infrastructure to build topics using the free Travis
169 CI has been added. Developers forking from this topic (and enabling
170 Travis) can do their own builds, and we can turn on auto-builds for
171 git/git (including build-status for pull requests that people
172 open).
174 * The write(2) emulation for Windows learned to set errno to EPIPE
175 when necessary.
178 Also contains various documentation updates and code clean-ups.
181 Fixes since v2.6
182 ----------------
184 Unless otherwise noted, all the fixes since v2.6 in the maintenance
185 track are contained in this release (see the maintenance releases'
186 notes for details).
188 * Very small number of options take a parameter that is optional
189 (which is not a great UI element as they can only appear at the end
190 of the command line). Add notice to documentation of each and
191 every one of them.
193 * "git blame --first-parent v1.0..v2.0" was not rejected but did not
194 limit the blame to commits on the first parent chain.
196 * "git subtree" (in contrib/) now can take whitespaces in the
197 pathnames, not only in the in-tree pathname but the name of the
198 directory that the repository is in.
200 * The ssh transport, just like any other transport over the network,
201 did not clear GIT_* environment variables, but it is possible to
202 use SendEnv and AcceptEnv to leak them to the remote invocation of
203 Git, which is not a good idea at all. Explicitly clear them just
204 like we do for the local transport.
206 * Correct "git p4 --detect-labels" so that it does not fail to create
207 a tag that points at a commit that is also being imported.
209 * The Makefile always runs the library archiver with hardcoded "crs"
210 options, which was inconvenient for exotic platforms on which
211 people want to use programs with totally different set of command
212 line options.
214 * Customization to change the behaviour with "make -w" and "make -s"
215 in our Makefile was broken when they were used together.
217 * Allocation related functions and stdio are unsafe things to call
218 inside a signal handler, and indeed killing the pager can cause
219 glibc to deadlock waiting on allocation mutex as our signal handler
220 tries to free() some data structures in wait_for_pager(). Reduce
221 these unsafe calls.
223 * The way how --ref/--notes to specify the notes tree reference are
224 DWIMmed was not clearly documented.
226 * "git gc" used to barf when a symbolic ref has gone dangling
227 (e.g. the branch that used to be your upstream's default when you
228 cloned from it is now gone, and you did "fetch --prune").
230 * "git clone --dissociate" runs a big "git repack" process at the
231 end, and it helps to close file descriptors that are open on the
232 packs and their idx files before doing so on filesystems that
233 cannot remove a file that is still open.
235 * Description of the "log.follow" configuration variable in "git log"
236 documentation is now also copied to "git config" documentation.
238 * "git rebase -i" had a minor regression recently, which stopped
239 considering a line that begins with an indented '#' in its insn
240 sheet not a comment. Further, the code was still too picky on
241 Windows where CRLF left by the editor is turned into a trailing CR
242 on the line read via the "read" built-in command of bash. Both of
243 these issues are now fixed.
245 * After "git checkout --detach", "git status" reported a fairly
246 useless "HEAD detached at HEAD", instead of saying at which exact
247 commit.
249 * When "git send-email" wanted to talk over Net::SMTP::SSL,
250 Net::Cmd::datasend() did not like to be fed too many bytes at the
251 same time and failed to send messages. Send the payload one line
252 at a time to work around the problem.
254 * When "git am" was rewritten as a built-in, it stopped paying
255 attention to user.signingkey, which was fixed.
257 * It was not possible to use a repository-lookalike created by "git
258 worktree add" as a local source of "git clone".
260 * On a case insensitive filesystems, setting GIT_WORK_TREE variable
261 using a random cases that does not agree with what the filesystem
262 thinks confused Git that it wasn't inside the working tree.
264 * Performance-measurement tests did not work without an installed Git.
266 * A test script for the HTTP service had a timing dependent bug,
267 which was fixed.
269 * There were some classes of errors that "git fsck" diagnosed to its
270 standard error that did not cause it to exit with non-zero status.
272 * Work around "git p4" failing when the P4 depot records the contents
273 in UTF-16 without UTF-16 BOM.
275 * When "git gc --auto" is backgrounded, its diagnosis message is
276 lost. Save it to a file in $GIT_DIR and show it next time the "gc
277 --auto" is run.
279 * The submodule code has been taught to work better with separate
280 work trees created via "git worktree add".
282 * "git gc" is safe to run anytime only because it has the built-in
283 grace period to protect young objects. In order to run with no
284 grace period, the user must make sure that the repository is
285 quiescent.
287 * A recent "filter-branch --msg-filter" broke skipping of the commit
288 object header, which is fixed.
290 * The normalize_ceiling_entry() function does not muck with the end
291 of the path it accepts, and the real world callers do rely on that,
292 but a test insisted that the function drops a trailing slash.
294 * A test for interaction between untracked cache and sparse checkout
295 added in Git 2.5 days were flaky.
297 * A couple of commands still showed "[options]" in their usage string
298 to note where options should come on their command line, but we
299 spell that "[<options>]" in most places these days.
301 * The synopsis text and the usage string of subcommands that read
302 list of things from the standard input are often shown as if they
303 only take input from a file on a filesystem, which was misleading.
305 * "git am -3" had a small regression where it is aborted in its error
306 handling codepath when underlying merge-recursive failed in certain
307 ways, as it assumed that the internal call to merge-recursive will
308 never die, which is not the case (yet).
310 * The linkage order of libraries was wrong in places around libcurl.
312 * The name-hash subsystem that is used to cope with case insensitive
313 filesystems keeps track of directories and their on-filesystem
314 cases for all the paths in the index by holding a pointer to a
315 randomly chosen cache entry that is inside the directory (for its
316 ce->ce_name component). This pointer was not updated even when the
317 cache entry was removed from the index, leading to use after free.
318 This was fixed by recording the path for each directory instead of
319 borrowing cache entries and restructuring the API somewhat.
321 * "git merge-file" tried to signal how many conflicts it found, which
322 obviously would not work well when there are too many of them.
324 * The error message from "git blame --contents --reverse" incorrectly
325 talked about "--contents --children".
327 * "git imap-send" did not compile well with older version of cURL library.
329 * Merging a branch that removes a path and another that changes the
330 mode bits on the same path should have conflicted at the path, but
331 it didn't and silently favoured the removal.
333 * "git --literal-pathspecs add -u/-A" without any command line
334 argument misbehaved ever since Git 2.0.
336 * "git daemon" uses "run_command()" without "finish_command()", so it
337 needs to release resources itself, which it forgot to do.
339 * "git status --branch --short" accessed beyond the constant string
340 "HEAD", which has been corrected.
342 * We peek objects from submodule's object store by linking it to the
343 list of alternate object databases, but the code to do so forgot to
344 correctly initialize the list.
346 * The code to prepare the working tree side of temporary directory
347 for the "dir-diff" feature forgot that symbolic links need not be
348 copied (or symlinked) to the temporary area, as the code already
349 special cases and overwrites them. Besides, it was wrong to try
350 computing the object name of the target of symbolic link, which may
351 not even exist or may be a directory.
353 * A Range: request can be responded with a full response and when
354 asked properly libcurl knows how to strip the result down to the
355 requested range. However, we were hand-crafting a range request
356 and it did not kick in.
358 * Having a leftover .idx file without corresponding .pack file in
359 the repository hurts performance; "git gc" learned to prune them.
361 * Apple's common crypto implementation of SHA1_Update() does not take
362 more than 4GB at a time, and we now have a compile-time workaround
363 for it.
365 * Produce correct "dirty" marker for shell prompts, even when we
366 are on an orphan or an unborn branch.
368 * A build without NO_IPv6 used to use gethostbyname() when guessing
369 user's hostname, instead of getaddrinfo() that is used in other
370 codepaths in such a build.
372 * The exit code of git-fsck did not reflect some types of errors
373 found in packed objects, which has been corrected.
375 * The helper used to iterate over loose object directories to prune
376 stale objects did not closedir() immediately when it is done with a
377 directory--a callback such as the one used for "git prune" may want
378 to do rmdir(), but it would fail on open directory on platforms
379 such as WinXP.
381 * "git p4" used to import Perforce CLs that touch only paths outside
382 the client spec as empty commits. It has been corrected to ignore
383 them instead, with a new configuration git-p4.keepEmptyCommits as a
384 backward compatibility knob.
386 * The completion script (in contrib/) used to list "git column"
387 (which is not an end-user facing command) as one of the choices
388 (merge 160fcdb sg/completion-no-column later to maint).
390 * The error reporting from "git send-email", when SMTP TLS fails, has
391 been improved.
392 (merge 9d60524 jk/send-email-ssl-errors later to maint).
394 * When getpwuid() on the system returned NULL (e.g. the user is not
395 in the /etc/passwd file or other uid-to-name mappings), the
396 codepath to find who the user is to record it in the reflog barfed
397 and died. Loosen the check in this codepath, which already accepts
398 questionable ident string (e.g. host part of the e-mail address is
399 obviously bogus), and in general when we operate fmt_ident() function
400 in non-strict mode.
401 (merge 92bcbb9 jk/ident-loosen-getpwuid later to maint).
403 * "git symbolic-ref" forgot to report a failure with its exit status.
404 (merge f91b273 jk/symbolic-ref-maint later to maint).
406 * History traversal with "git log --source" that starts with an
407 annotated tag failed to report the tag as "source", due to an
408 old regression in the command line parser back in v2.2 days.
409 (merge 728350b jk/pending-keep-tag-name later to maint).
411 * "git p4" when interacting with multiple depots at the same time
412 used to incorrectly drop changes.
414 * Code clean-up, minor fixes etc.