1 Git v2.44 Release Notes
2 =======================
4 Backward Compatibility Notes
6 * "git checkout -B <branch>" used to allow switching to a branch that
7 is in use on another worktree, but this was by mistake. The users
8 need to use "--ignore-other-worktrees" option.
11 UI, Workflows & Features
13 * "git add" and "git stash" learned to support the ":(attr:...)"
16 * "git rebase --autosquash" is now enabled for non-interactive rebase,
17 but it is still incompatible with the apply backend.
19 * Introduce "git replay", a tool meant on the server side without
20 working tree to recreate a history.
22 * "git merge-file" learned to take the "--diff-algorithm" option to
23 use algorithm different from the default "myers" diff.
25 * Command line completion (in contrib/) learned to complete path
26 arguments to the "add/set" subcommands of "git sparse-checkout"
29 * "git checkout -B <branch> [<start-point>]" allowed a branch that is
30 in use in another worktree to be updated and checked out, which
31 might be a bit unexpected. The rule has been tightened, which is a
32 breaking change. "--ignore-other-worktrees" option is required to
33 unbreak you, if you are used to the current behaviour that "-B"
36 * The builtin_objectmode attribute is populated for each path
37 without adding anything in .gitattributes files, which would be
38 useful in magic pathspec, e.g., ":(attr:builtin_objectmode=100755)"
39 to limit to executables.
41 * "git fetch" learned to pay attention to "fetch.all" configuration
42 variable, which pretends as if "--all" was passed from the command
43 line when no remote parameter was given.
45 * In addition to (rather cryptic) Security Identifiers, show username
46 and domain in the error message when we barf on mismatch between
47 the Git directory and the current user on Windows.
49 * The error message given when "git branch -d branch" fails due to
50 commits unique to the branch has been split into an error and a new
51 conditional advice message.
53 * When given an existing but unreadable file as a configuration file,
54 gitweb behaved as if the file did not exist at all, but now it
55 errors out. This is a change that may break backward compatibility.
57 * When $HOME/.gitconfig is missing but XDG config file is available, we
58 should write into the latter, not former. "git gc" and "git
59 maintenance" wrote into a wrong "global config" file, which have
62 * Define "special ref" as a very narrow set that consists of
63 FETCH_HEAD and MERGE_HEAD, and clarify everything else that used to
64 be classified as such are actually just pseudorefs.
66 * All conditional "advice" messages show how to turn them off, which
67 becomes repetitive. Setting advice.* configuration explicitly on
68 now omits the instruction part.
70 * The "disable repository discovery of a bare repository" check,
71 triggered by setting safe.bareRepository configuration variable to
72 'explicit', has been loosened to exclude the ".git/" directory inside
73 a non-bare repository from the check. So you can do "cd .git &&
74 git cmd" to run a Git command that works on a bare repository without
75 explicitly specifying $GIT_DIR now.
77 * The completion script (in contrib/) learned more options that can
78 be used with "git log".
80 * The labels on conflict markers for the common ancestor, our version,
81 and the other version are available to custom 3-way merge driver
82 via %S, %X, and %Y placeholders.
84 * The write codepath for the reftable data learned to honor
85 core.fsync configuration.
87 * The "--fsck-objects" option of "git index-pack" now can take the
88 optional parameter to tweak severity of different fsck errors.
90 * The wincred credential backend has been taught to support oauth
91 refresh token the same way as credential-cache and
92 credential-libsecret backends.
94 * Command line completion support (in contrib/) has been
95 updated for "git bisect".
97 * "git branch" and friends learned to use the formatted text as
98 sorting key, not the underlying timestamp value, when the --sort
99 option is used with author or committer timestamp with a format
100 specifier (e.g., "--sort=creatordate:format:%H:%M:%S").
102 * The command line completion script (in contrib/) learned to
103 complete configuration variable names better.
106 Performance, Internal Implementation, Development Support etc.
108 * Process to add some form of low-level unit tests has started.
110 * Add support for GitLab CI.
112 * "git for-each-ref --no-sort" still sorted the refs alphabetically
113 which paid non-trivial cost. It has been redefined to show output
114 in an unspecified order, to allow certain optimizations to take
117 * Simplify API implementation to delete references by eliminating
120 * Subject approxidate() and show_date() machinery to OSS-Fuzz.
122 * A new helper to let us pretend that we called lstat() when we know
123 our cache_entry is up-to-date via fsmonitor.
125 * The optimization based on fsmonitor in the "diff --cached"
126 codepath is resurrected with the "fake-lstat" introduced earlier.
128 * Test balloon to use C99 "bool" type from <stdbool.h> has been
131 * "git clone" has been prepared to allow cloning a repository with
132 non-default hash function into a repository that uses the reftable
135 * Streaming spans of packfile data used to be done only from a
136 single, primary, pack in a repository with multiple packfiles. It
137 has been extended to allow reuse from other packfiles, too.
139 * Comment updates to help developers not to attempt to modify
140 messages from plumbing commands that must stay constant.
142 It might make sense to reassess the plumbing needs every few years,
143 but that should be done as a separate effort.
145 * Move test-ctype helper to the unit-test framework.
147 * Instead of manually creating refs/ hierarchy on disk upon a
148 creation of a secondary worktree, which is only usable via the
149 files backend, use the refs API to populate it.
151 * CI for GitLab learned to drive macOS jobs.
153 * A few tests to "git commit -o <pathspec>" and "git commit -i
154 <pathspec>" has been added.
156 * Tests on ref API are moved around to prepare for reftable.
158 * The Makefile often had to say "-L$(path) -R$(path)" that repeats
159 the path to the same library directory for link time and runtime.
160 A Makefile template is used to reduce such repetition.
162 * The priority queue test has been migrated to the unit testing
165 * Setting `feature.experimental` opts the user into multi-pack reuse
168 * Squelch node.js 16 deprecation warnings from GitHub Actions CI
169 by updating actions/github-script and actions/checkout that use
172 * The mechanism to report the filename in the source code, used by
173 the unit-test machinery, assumed that the compiler expanded __FILE__
174 to the path to the source given to the $(CC), but some compilers
175 give full path, breaking the output. This has been corrected.
181 * The way CI testing used "prove" could lead to running the test
182 suite twice needlessly, which has been corrected.
184 * Update ref-related tests.
186 * "git format-patch --encode-email-headers" ignored the option when
187 preparing the cover letter, which has been corrected.
189 * Newer versions of Getopt::Long started giving warnings against our
190 (ab)use of it in "git send-email". Bump the minimum version
191 requirement for Perl to 5.8.1 (from September 2002) to allow
192 simplifying our implementation.
194 * Earlier we stopped relying on commit-graph that (still) records
195 information about commits that are lost from the object store,
196 which has negative performance implications. The default has been
197 flipped to disable this pessimization.
199 * Stale URLs have been updated to their current counterparts (or
200 archive.org) and HTTP links are replaced with working HTTPS links.
202 * trace2 streams used to record the URLs that potentially embed
203 authentication material, which has been corrected.
205 * The sample pre-commit hook that tries to catch introduction of new
206 paths that use potentially non-portable characters did not notice
207 an existing path getting renamed to such a problematic path, when
208 rename detection was enabled.
210 * The command line parser for the "log" family of commands was too
211 loose when parsing certain numbers, e.g., silently ignoring the
212 extra 'q' in "git log -n 1q" without complaining, which has been
215 * "git $cmd --end-of-options --rev -- --path" for some $cmd failed
216 to interpret "--rev" as a rev, and "--path" as a path. This was
217 fixed for many programs like "reset" and "checkout".
219 * "git bisect reset" has been taught to clean up state files and refs
220 even when BISECT_START file is gone.
222 * Some codepaths did not correctly parse configuration variables
223 specified with valueless "true", which has been corrected.
225 * Code clean-up for sanity checking of command line options for "git
228 * The code to parse the From e-mail header has been updated to avoid
231 * "git fetch --atomic" issued an unnecessary empty error message,
232 which has been corrected.
234 * Command line completion script (in contrib/) learned to work better
235 with the reftable backend.
237 * "git status" is taught to show both the branch being bisected and
238 being rebased when both are in effect at the same time.
240 * "git archive --list extra garbage" silently ignored excess command
241 line parameters, which has been corrected.
243 * "git sparse-checkout set" added default patterns even when the
244 patterns are being fed from the standard input, which has been
247 * "git sparse-checkout (add|set) --[no-]cone --end-of-options" did
248 not handle "--end-of-options" correctly after a recent update.
250 * Unlike other environment variables that took the usual
251 true/false/yes/no as well as 0/1, GIT_FLUSH only understood 0/1,
252 which has been corrected.
254 * Clearing in-core repository (happens during e.g., "git fetch
255 --recurse-submodules" with commit graph enabled) made in-core
256 commit object in an inconsistent state by discarding the necessary
257 data from commit-graph too early, which has been corrected.
259 * Update to a new feature recently added, "git show-ref --exists".
261 * oss-fuzz tests are built and run in CI.
262 (merge c4a9cf1df3 js/oss-fuzz-build-in-ci later to maint).
264 * Rename detection logic ignored the final line of a file if it is an
268 (merge 0188b2c8e0 pb/ci-github-skip-logs-for-broken-tests later to maint).
270 * "git diff --no-rename A B" did not disable rename detection but did
271 not trigger an error from the command line parser.
273 * "git archive --remote=<remote>" learned to talk over the smart
274 http (aka stateless) transport.
275 (merge 176cd68634 jx/remote-archive-over-smart-http later to maint).
277 * Fetching via protocol v0 over Smart HTTP transport sometimes failed
278 to correctly auto-follow tags.
279 (merge fba732c462 jk/fetch-auto-tag-following-fix later to maint).
281 * The documentation for the --exclude-per-directory option marked it
282 as deprecated, which confused readers into thinking there may be a
283 plan to remove it in the future, which was not our intention.
284 (merge 0009542cab jc/ls-files-doc-update later to maint).
286 * "git diff --no-index file1 file2" segfaulted while invoking the
287 external diff driver, which has been corrected.
289 * Rewrite //-comments to /* comments */ in files whose comments
290 prevalently use the latter.
292 * Cirrus CI jobs started breaking because we specified version of
293 FreeBSD that is no longer available, which has been corrected.
294 (merge 81fffb66d3 cb/use-freebsd-13-2-at-cirrus-ci later to maint).
296 * A caller called index_file_exists() that takes a string expressed
297 as <ptr, length> with a wrong length, which has been corrected.
298 (merge 156e28b36d jh/sparse-index-expand-to-path-fix later to maint).
300 * A failed "git tag -s" did not necessarily result in an error
301 depending on the crypto backend, which has been corrected.
303 * "git stash" sometimes was silent even when it failed due to
304 unwritable index file, which has been corrected.
306 * "git show-ref --verify" did not show things like "CHERRY_PICK_HEAD",
307 which has been corrected.
309 * Recent conversion to allow more than 0/1 in GIT_FLUSH broke the
310 mechanism by flipping what yes/no means by mistake, which has been
313 * The sequencer machinery does not use the ref API and instead
314 records names of certain objects it needs for its correct operation
315 in temporary files, which makes these objects susceptible to loss
316 by garbage collection. These temporary files have been added as
317 starting points for reachability analysis to fix this.
318 (merge bc7f5db896 pw/gc-during-rebase later to maint).
320 * "git cherry-pick" invoked during "git rebase -i" session lost
321 the authorship information, which has been corrected.
322 (merge e4301f73ff vn/rebase-with-cherry-pick-authorship later to maint).
324 * The code paths that call repo_read_object_file() have been
325 tightened to react to errors.
326 (merge 568459bf5e js/check-null-from-read-object-file later to maint).
328 * Other code cleanup, docfix, build fix, etc.
329 (merge 5aea3955bc rj/clarify-branch-doc-m later to maint).
330 (merge 9cce3be2df bk/bisect-doc-fix later to maint).
331 (merge 8430b438f6 vd/fsck-submodule-url-test later to maint).
332 (merge 3cb4384683 jc/t0091-with-unknown-git later to maint).
333 (merge 020456cb74 rs/receive-pack-remove-find-header later to maint).
334 (merge bc47139f4f la/trailer-cleanups later to maint).