Documentation/diff-generate-patch.txt

   1 Generating patch text with -p
   2 -----------------------------
   3
   4 Running
   5 linkgit:git-diff[1],
   6 linkgit:git-log[1],
   7 linkgit:git-show[1],
   8 linkgit:git-diff-index[1],
   9 linkgit:git-diff-tree[1], or
  10 linkgit:git-diff-files[1]
  11 with the `-p` option produces patch text.
  12 You can customize the creation of patch text via the
  13 `GIT_EXTERNAL_DIFF` and the `GIT_DIFF_OPTS` environment variables
  14 (see linkgit:git[1]), and the `diff` attribute (see linkgit:gitattributes[5]).
  15
  16 What the -p option produces is slightly different from the traditional
  17 diff format:
  18
  19 1.   It is preceded with a "git diff" header that looks like this:
  20
  21        diff --git a/file1 b/file2
  22 +
  23 The `a/` and `b/` filenames are the same unless rename/copy is
  24 involved.  Especially, even for a creation or a deletion,
  25 `/dev/null` is _not_ used in place of the `a/` or `b/` filenames.
  26 +
  27 When rename/copy is involved, `file1` and `file2` show the
  28 name of the source file of the rename/copy and the name of
  29 the file that rename/copy produces, respectively.
  30
  31 2.   It is followed by one or more extended header lines:
  32
  33        old mode <mode>
  34        new mode <mode>
  35        deleted file mode <mode>
  36        new file mode <mode>
  37        copy from <path>
  38        copy to <path>
  39        rename from <path>
  40        rename to <path>
  41        similarity index <number>
  42        dissimilarity index <number>
  43        index <hash>..<hash> <mode>
  44 +
  45 File modes are printed as 6-digit octal numbers including the file type
  46 and file permission bits.
  47 +
  48 Path names in extended headers do not include the `a/` and `b/` prefixes.
  49 +
  50 The similarity index is the percentage of unchanged lines, and
  51 the dissimilarity index is the percentage of changed lines.  It
  52 is a rounded down integer, followed by a percent sign.  The
  53 similarity index value of 100% is thus reserved for two equal
  54 files, while 100% dissimilarity means that no line from the old
  55 file made it into the new one.
  56 +
  57 The index line includes the blob object names before and after the change.
  58 The <mode> is included if the file mode does not change; otherwise,
  59 separate lines indicate the old and the new mode.
  60
  61 3.  Pathnames with "unusual" characters are quoted as explained for
  62     the configuration variable `core.quotePath` (see
  63     linkgit:git-config[1]).
  64
  65 4.  All the `file1` files in the output refer to files before the
  66     commit, and all the `file2` files refer to files after the commit.
  67     It is incorrect to apply each change to each file sequentially.  For
  68     example, this patch will swap a and b:
  69
  70       diff --git a/a b/b
  71       rename from a
  72       rename to b
  73       diff --git a/b b/a
  74       rename from b
  75       rename to a
  76
  77 5.  Hunk headers mention the name of the function to which the hunk
  78     applies.  See "Defining a custom hunk-header" in
  79     linkgit:gitattributes[5] for details of how to tailor to this to
  80     specific languages.
  81
  82
  83 Combined diff format
  84 --------------------
  85
  86 Any diff-generating command can take the `-c` or `--cc` option to
  87 produce a 'combined diff' when showing a merge. This is the default
  88 format when showing merges with linkgit:git-diff[1] or
  89 linkgit:git-show[1]. Note also that you can give suitable
  90 `--diff-merges` option to any of these commands to force generation of
  91 diffs in specific format.
  92
  93 A "combined diff" format looks like this:
  94
  95 ------------
  96 diff --combined describe.c
  97 index fabadb8,cc95eb0..4866510
  98 --- a/describe.c
  99 +++ b/describe.c
 100 @@@ -98,20 -98,12 +98,20 @@@
 101         return (a_date > b_date) ? -1 : (a_date == b_date) ? 0 : 1;
 102   }
 103
 104 - static void describe(char *arg)
 105  -static void describe(struct commit *cmit, int last_one)
 106 ++static void describe(char *arg, int last_one)
 107   {
 108  +      unsigned char sha1[20];
 109  +      struct commit *cmit;
 110         struct commit_list *list;
 111         static int initialized = 0;
 112         struct commit_name *n;
 113
 114  +      if (get_sha1(arg, sha1) < 0)
 115  +              usage(describe_usage);
 116  +      cmit = lookup_commit_reference(sha1);
 117  +      if (!cmit)
 118  +              usage(describe_usage);
 119  +
 120         if (!initialized) {
 121                 initialized = 1;
 122                 for_each_ref(get_name);
 123 ------------
 124
 125 1.   It is preceded with a "git diff" header, that looks like
 126      this (when the `-c` option is used):
 127
 128        diff --combined file
 129 +
 130 or like this (when the `--cc` option is used):
 131
 132        diff --cc file
 133
 134 2.   It is followed by one or more extended header lines
 135      (this example shows a merge with two parents):
 136
 137        index <hash>,<hash>..<hash>
 138        mode <mode>,<mode>..<mode>
 139        new file mode <mode>
 140        deleted file mode <mode>,<mode>
 141 +
 142 The `mode <mode>,<mode>..<mode>` line appears only if at least one of
 143 the <mode> is different from the rest. Extended headers with
 144 information about detected contents movement (renames and
 145 copying detection) are designed to work with diff of two
 146 <tree-ish> and are not used by combined diff format.
 147
 148 3.   It is followed by two-line from-file/to-file header
 149
 150        --- a/file
 151        +++ b/file
 152 +
 153 Similar to two-line header for traditional 'unified' diff
 154 format, `/dev/null` is used to signal created or deleted
 155 files.
 156 +
 157 However, if the --combined-all-paths option is provided, instead of a
 158 two-line from-file/to-file you get a N+1 line from-file/to-file header,
 159 where N is the number of parents in the merge commit
 160
 161        --- a/file
 162        --- a/file
 163        --- a/file
 164        +++ b/file
 165 +
 166 This extended format can be useful if rename or copy detection is
 167 active, to allow you to see the original name of the file in different
 168 parents.
 169
 170 4.   Chunk header format is modified to prevent people from
 171      accidentally feeding it to `patch -p1`. Combined diff format
 172      was created for review of merge commit changes, and was not
 173      meant to be applied. The change is similar to the change in the
 174      extended 'index' header:
 175
 176        @@@ <from-file-range> <from-file-range> <to-file-range> @@@
 177 +
 178 There are (number of parents + 1) `@` characters in the chunk
 179 header for combined diff format.
 180
 181 Unlike the traditional 'unified' diff format, which shows two
 182 files A and B with a single column that has `-` (minus --
 183 appears in A but removed in B), `+` (plus -- missing in A but
 184 added to B), or `" "` (space -- unchanged) prefix, this format
 185 compares two or more files file1, file2,... with one file X, and
 186 shows how X differs from each of fileN.  One column for each of
 187 fileN is prepended to the output line to note how X's line is
 188 different from it.
 189
 190 A `-` character in the column N means that the line appears in
 191 fileN but it does not appear in the result.  A `+` character
 192 in the column N means that the line appears in the result,
 193 and fileN does not have that line (in other words, the line was
 194 added, from the point of view of that parent).
 195
 196 In the above example output, the function signature was changed
 197 from both files (hence two `-` removals from both file1 and
 198 file2, plus `++` to mean one line that was added does not appear
 199 in either file1 or file2).  Also eight other lines are the same
 200 from file1 but do not appear in file2 (hence prefixed with `+`).
 201
 202 When shown by `git diff-tree -c`, it compares the parents of a
 203 merge commit with the merge result (i.e. file1..fileN are the
 204 parents).  When shown by `git diff-files -c`, it compares the
 205 two unresolved merge parents with the working tree file
 206 (i.e. file1 is stage 2 aka "our version", file2 is stage 3 aka
 207 "their version").