1 Git v2.9.3 Release Notes
2 ========================
7 * A helper function that takes the contents of a commit object and
8 finds its subject line did not ignore leading blank lines, as is
9 commonly done by other codepaths. Make it ignore leading blank
12 * Git does not know what the contents in the index should be for a
13 path added with "git add -N" yet, so "git grep --cached" should not
14 show hits (or show lack of hits, with -L) in such a path, but that
15 logic does not apply to "git grep", i.e. searching in the working
16 tree files. But we did so by mistake, which has been corrected.
18 * "git rebase -i --autostash" did not restore the auto-stashed change
19 when the operation was aborted.
21 * "git commit --amend --allow-empty-message -S" for a commit without
22 any message body could have misidentified where the header of the
25 * More mark-up updates to typeset strings that are expected to
26 literally typed by the end user in fixed-width font.
28 * For a long time, we carried an in-code comment that said our
29 colored output would work only when we use fprintf/fputs on
30 Windows, which no longer is the case for the past few years.
32 * "gc.autoPackLimit" when set to 1 should not trigger a repacking
33 when there is only one pack, but the code counted poorly and did
36 * One part of "git am" had an oddball helper function that called
37 stuff from outside "his" as opposed to calling what we have "ours",
38 which was not gender-neutral and also inconsistent with the rest of
39 the system where outside stuff is usually called "theirs" in
42 * The test framework learned a new helper test_match_signal to
43 check an exit code from getting killed by an expected signal.
45 * "git blame -M" missed a single line that was moved within the file.
47 * Fix recently introduced codepaths that are involved in parallel
48 submodule operations, which gave up on reading too early, and
49 could have wasted CPU while attempting to write under a corner
52 * "git grep -i" has been taught to fold case in non-ascii locales
55 * A test that unconditionally used "mktemp" learned that the command
56 is not necessarily available everywhere.
58 * "git blame file" allowed the lineage of lines in the uncommitted,
59 unadded contents of "file" to be inspected, but it refused when
60 "file" did not appear in the current commit. When "file" was
61 created by renaming an existing file (but the change has not been
62 committed), this restriction was unnecessarily tight.
64 * "git add -N dir/file && git write-tree" produced an incorrect tree
65 when there are other paths in the same directory that sorts after
68 * "git fetch http://user:pass@host/repo..." scrubbed the userinfo
69 part, but "git push" didn't.
71 * An age old bug that caused "git diff --ignore-space-at-eol"
72 misbehave has been fixed.
74 * "git notes merge" had a code to see if a path exists (and fails if
75 it does) and then open the path for writing (when it doesn't).
76 Replace it with open with O_EXCL.
78 * "git pack-objects" and "git index-pack" mostly operate with off_t
79 when talking about the offset of objects in a packfile, but there
80 were a handful of places that used "unsigned long" to hold that
81 value, leading to an unintended truncation.
83 * Recent update to "git daemon" tries to enable the socket-level
84 KEEPALIVE, but when it is spawned via inetd, the standard input
85 file descriptor may not necessarily be connected to a socket.
86 Suppress an ENOTSOCK error from setsockopt().
88 * Recent FreeBSD stopped making perl available at /usr/bin/perl;
89 switch the default the built-in path to /usr/local/bin/perl on not
90 too ancient FreeBSD releases.
92 * "git status" learned to suggest "merge --abort" during a conflicted
93 merge, just like it already suggests "rebase --abort" during a
96 * The .c/.h sources are marked as such in our .gitattributes file so
97 that "git diff -W" and friends would work better.
99 * Existing autoconf generated test for the need to link with pthread
100 library did not check all the functions from pthread libraries;
101 recent FreeBSD has some functions in libc but not others, and we
102 mistakenly thought linking with libc is enough when it is not.
104 * Allow http daemon tests in Travis CI tests.
106 * Users of the parse_options_concat() API function need to allocate
107 extra slots in advance and fill them with OPT_END() when they want
108 to decide the set of supported options dynamically, which makes the
109 code error-prone and hard to read. This has been corrected by tweaking
110 the API to allocate and return a new copy of "struct option" array.
112 * The use of strbuf in "git rm" to build filename to remove was a bit
113 suboptimal, which has been fixed.
115 * "git commit --help" said "--no-verify" is only about skipping the
116 pre-commit hook, and failed to say that it also skipped the
119 * "git merge" in Git v2.9 was taught to forbid merging an unrelated
120 lines of history by default, but that is exactly the kind of thing
121 the "--rejoin" mode of "git subtree" (in contrib/) wants to do.
122 "git subtree" has been taught to use the "--allow-unrelated-histories"
123 option to override the default.
125 * The build procedure for "git persistent-https" helper (in contrib/)
126 has been updated so that it can be built with more recent versions
129 * There is an optimization used in "git diff $treeA $treeB" to borrow
130 an already checked-out copy in the working tree when it is known to
131 be the same as the blob being compared, expecting that open/mmap of
132 such a file is faster than reading it from the object store, which
133 involves inflating and applying delta. This however kicked in even
134 when the checked-out copy needs to go through the convert-to-git
135 conversion (including the clean filter), which defeats the whole
136 point of the optimization. The optimization has been disabled when
137 the conversion is necessary.
139 * "git -c grep.patternType=extended log --basic-regexp" misbehaved
140 because the internal API to access the grep machinery was not
143 * Windows port was failing some tests in t4130, due to the lack of
144 inum in the returned values by its lstat(2) emulation.
146 * The characters in the label shown for tags/refs for commits in
147 "gitweb" output are now properly escaped for proper HTML output.
149 * FreeBSD can lie when asked mtime of a directory, which made the
150 untracked cache code to fall back to a slow-path, which in turn
151 caused tests in t7063 to fail because it wanted to verify the
152 behaviour of the fast-path.
154 * Squelch compiler warnings for netmalloc (in compat/) library.
156 * The API documentation for hashmap was unclear if hashmap_entry
157 can be safely discarded without any other consideration. State
158 that it is safe to do so.
160 * Not-so-recent rewrite of "git am" that started making internal
161 calls into the commit machinery had an unintended regression, in
162 that no matter how many seconds it took to apply many patches, the
163 resulting committer timestamp for the resulting commits were all
166 * "git difftool <paths>..." started in a subdirectory failed to
167 interpret the paths relative to that directory, which has been
170 Also contains minor documentation updates and code clean-ups.