summaryrefslogtreecommitdiff
path: root/Documentation/RelNotes/2.2.0.txt
blob: 5f0c421365d52134030d14580258059b27a536b9 (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
Git v2.2 Release Notes
======================

Updates since v2.1
------------------

Ports

 * Building on older MacOS X systems automatically sets
   the necessary NO_APPLE_COMMON_CRYPTO build-time option.


UI, Workflows & Features

 * "git archive" learned to filter what gets archived with pathspec.

 * "git config --edit --global" starts from a skeletal per-user
   configuration file contents, instead of a total blank, when the
   user does not already have any.  This immediately reduces the
   need for a later "Have you forgotten setting core.user?" and we
   can add more to the template as we gain more experience.

 * "git stash list -p" used to be almost always a no-op because each
   stash entry is represented as a merge commit.  It learned to show
   the difference between the base commit version and the working tree
   version, which is in line with what "git show" gives.

 * Sometimes users want to report a bug they experience on their
   repository, but they are not at liberty to share the contents of
   the repository.  "fast-export" was taught an "--anonymize" option
   to replace blob contents, names of people and paths and log
   messages with bland and simple strings to help them.

 * "log --date=iso" uses a slight variant of ISO 8601 format that is
   made more human readable.  A new "--date=iso-strict" option gives
   datetime output that is more strictly conformant.

 * A broken reimplementation of Git could write an invalid index that
   records both stage #0 and higher stage entries for the same path.
   We now notice and reject such an index, as there is no sensible
   fallback (we do not know if the broken tool wanted to resolve and
   forgot to remove higher stage entries, or if it wanted to unresolve
   and forgot to remove the stage#0 entry).

 * The "pre-receive" and "post-receive" hooks are no longer required
   to consume their input fully (not following this requirement used
   to result in intermittent errors in "git push").

 * The pretty-format specifier "%d", which expanded to " (tagname)"
   for a tagged commit, gained a cousin "%D" that just gives the
   "tagname" without frills.

 * "git push" learned "--signed" push, that allows a push (i.e.
   request to update the refs on the other side to point at a new
   history, together with the transmission of necessary objects) to be
   signed, so that it can be verified and audited, using the GPG
   signature of the person who pushed, that the tips of branches at a
   public repository really point the commits the pusher wanted to,
   without having to "trust" the server.

Performance, Internal Implementation, etc.

 * The API to manipulate the "refs" is currently undergoing a revamp
   to make it more transactional, with the eventual goal to allow
   all-or-none atomic updates and migrating the storage to something
   other than the traditional filesystem based one (e.g. databases).

 * We no longer attempt to keep track of individual dependencies to
   the header files in the build procedure, relying on automated
   dependency generation support from modern compilers.

 * In tests, we have been using NOT_{MINGW,CYGWIN} test prerequisites
   long before negated prerequisites e.g. !MINGW were invented.
   The former has been converted to the latter to avoid confusion.

 * Looking up remotes configuration in a repository with very many
   remotes defined has been optimized.

 * There are cases where you lock and open to write a file, close it
   to show the updated contents to external processes, and then have
   to update the file again while still holding the lock, but the
   lockfile API lacked support for such an access pattern.

 * The API to allocate the structure to keep track of commit
   decoration has been updated to make it less cumbersome to use.

 * An in-core caching layer to let us avoid reading the same
   configuration files number of times has been added.  A few commands
   have been converted to use this subsystem.

 * Various code paths have been cleaned up and simplified by using
   "strbuf", "starts_with()", and "skip_prefix()" APIs more.

 * A few codepaths that died when large blobs that would not fit in
   core are involved in their operation have been taught to punt
   instead, by e.g. marking too large a blob as not to be diffed.

 * A few more code paths in "commit" and "checkout" have been taught
   to repopulate the cache-tree in the index, to help speed up later
   "write-tree" (used in "commit") and "diff-index --cached" (used in
   "status").

 * A common programming mistake to assign the same short option name
   to two separate options is detected by parse_options() API to help
   developers.

 * The code path to write out the packed-refs file has been optimized,
   which especially matters in a repository with a large number of
   refs.

 * The check to see if a ref $F can be created by making sure no
   existing ref has $F/ as its prefix has been optimized, which
   especially matters in a repository with a large number of existing
   refs.

 * "git fsck" was taught to check contents of tag objects a bit more.

 * "git hash-object" was taught a "--literally" option to help
   debugging.

 * When running a required clean filter, we do not have to mmap the
   original before feeding the filter.  Instead, stream the file
   contents directly to the filter and process its output.

Also contains various documentation updates and code clean-ups.


Fixes since v2.1
----------------

Unless otherwise noted, all the fixes since v2.1 in the maintenance
track are contained in this release (see the maintenance releases'
notes for details).

 * "git log --pretty/format=" with an empty format string did not
   mean the more obvious "No output whatsoever" but "Use default
   format", which was counterintuitive.

 * Implementations of "tar" that do not understand an extended pax
   header would extract the contents of it in a regular file; make
   sure the permission bits of this file follows the same tar.umask
   configuration setting.

 * "git -c section.var command" and "git -c section.var= command"
   should pass the configuration differently (the former should be a
   boolean true, the latter should be an empty string).

 * Applying a patch not generated by Git in a subdirectory used to
   check the whitespace breakage using the attributes for incorrect
   paths. Also whitespace checks were performed even for paths
   excluded via "git apply --exclude=<path>" mechanism.

 * "git bundle create" with date-range specification were meant to
   exclude tags outside the range, but it didn't.

 * "git add x" where x that used to be a directory has become a
   symbolic link to a directory misbehaved.

 * The prompt script checked $GIT_DIR/ref/stash file to see if there
   is a stash, which was a no-no.

 * Pack-protocol documentation had a minor typo.

 * "git checkout -m" did not switch to another branch while carrying
   the local changes forward when a path was deleted from the index.

 * With sufficiently long refnames, "git fast-import" could have
   overflown an on-stack buffer.

 * After "pack-refs --prune" packed refs at the top-level, it failed
   to prune them.

 * Progress output from "git gc --auto" was visible in "git fetch -q".

 * We used to pass -1000 to poll(2), expecting it to also mean "no
   timeout", which should be spelled as -1.

 * "git rebase" documentation was unclear that it is required to
   specify on what <upstream> the rebase is to be done when telling it
   to first check out <branch>.
   (merge 95c6826 so/rebase-doc later to maint).

 * "git push" over HTTP transport had an artificial limit on number of
   refs that can be pushed imposed by the command line length.
   (merge 26be19b jk/send-pack-many-refspecs later to maint).

 * When receiving an invalid pack stream that records the same object
   twice, multiple threads got confused due to a race.
   (merge ab791dd jk/index-pack-threading-races later to maint).

 * An attempt to remove the entire tree in the "git fast-import" input
   stream caused it to misbehave.
   (merge 2668d69 mb/fast-import-delete-root later to maint).

 * Reachability check (used in "git prune" and friends) did not add a
   detached HEAD as a starting point to traverse objects still in use.
   (merge c40fdd0 mk/reachable-protect-detached-head later to maint).

 * "git config --add section.var val" used to lose existing
   section.var whose value was an empty string.
   (merge c1063be ta/config-add-to-empty-or-true-fix later to maint).

 * "git fsck" failed to report that it found corrupt objects via its
   exit status in some cases.
   (merge 30d1038 jk/fsck-exit-code-fix later to maint).

 * Use of "--verbose" option used to break "git branch --merged".
   (merge 12994dd jk/maint-branch-verbose-merged later to maint).

 * Some MUAs mangled a line in a message that begins with "From " to
   ">From " when writing to a mailbox file and feeding such an input
   to "git am" used to lose such a line.
   (merge 85de86a jk/mbox-from-line later to maint).

 * "rev-parse --verify --quiet $name" is meant to quietly exit with a
   non-zero status when $name is not a valid object name, but still
   gave error messages in some cases.