1Generating patches with -p 2-------------------------- 3 4When "git-diff-index", "git-diff-tree", or "git-diff-files" are run 5with a '-p' option, "git diff" without the '--raw' option, or 6"git log" with the "-p" option, they 7do not produce the output described above; instead they produce a 8patch file. You can customize the creation of such patches via the 9GIT_EXTERNAL_DIFF and the GIT_DIFF_OPTS environment variables. 10 11What the -p option produces is slightly different from the traditional 12diff format: 13 141. It is preceded with a "git diff" header that looks like this: 15 16 diff --git a/file1 b/file2 17+ 18The `a/` and `b/` filenames are the same unless rename/copy is 19involved. Especially, even for a creation or a deletion, 20`/dev/null` is _not_ used in place of the `a/` or `b/` filenames. 21+ 22When rename/copy is involved, `file1` and `file2` show the 23name of the source file of the rename/copy and the name of 24the file that rename/copy produces, respectively. 25 262. It is followed by one or more extended header lines: 27 28 old mode <mode> 29 new mode <mode> 30 deleted file mode <mode> 31 new file mode <mode> 32 copy from <path> 33 copy to <path> 34 rename from <path> 35 rename to <path> 36 similarity index <number> 37 dissimilarity index <number> 38 index <hash>..<hash> <mode> 39+ 40File modes are printed as 6-digit octal numbers including the file type 41and file permission bits. 42+ 43Path names in extended headers do not include the `a/` and `b/` prefixes. 44+ 45The similarity index is the percentage of unchanged lines, and 46the dissimilarity index is the percentage of changed lines. It 47is a rounded down integer, followed by a percent sign. The 48similarity index value of 100% is thus reserved for two equal 49files, while 100% dissimilarity means that no line from the old 50file made it into the new one. 51+ 52The index line includes the SHA-1 checksum before and after the change. 53The <mode> is included if the file mode does not change; otherwise, 54separate lines indicate the old and the new mode. 55 563. TAB, LF, double quote and backslash characters in pathnames 57 are represented as `\t`, `\n`, `\"` and `\\`, respectively. 58 If there is need for such substitution then the whole 59 pathname is put in double quotes. 60 614. All the `file1` files in the output refer to files before the 62 commit, and all the `file2` files refer to files after the commit. 63 It is incorrect to apply each change to each file sequentially. For 64 example, this patch will swap a and b: 65 66 diff --git a/a b/b 67 rename from a 68 rename to b 69 diff --git a/b b/a 70 rename from b 71 rename to a 72 73 74combined diff format 75-------------------- 76 77"git-diff-tree", "git-diff-files" and "git-diff" can take '-c' or 78'--cc' option to produce 'combined diff'. For showing a merge commit 79with "git log -p", this is the default format; you can force showing 80full diff with the '-m' option. 81A 'combined diff' format looks like this: 82 83------------ 84diff --combined describe.c 85index fabadb8,cc95eb0..4866510 86--- a/describe.c 87+++ b/describe.c 88@@@ -98,20 -98,12 +98,20 @@@ 89 return (a_date > b_date) ? -1 : (a_date == b_date) ? 0 : 1; 90 } 91 92- static void describe(char *arg) 93 -static void describe(struct commit *cmit, int last_one) 94++static void describe(char *arg, int last_one) 95 { 96 + unsigned char sha1[20]; 97 + struct commit *cmit; 98 struct commit_list *list; 99 static int initialized = 0; 100 struct commit_name *n; 101 102 + if (get_sha1(arg, sha1) < 0) 103 + usage(describe_usage); 104 + cmit = lookup_commit_reference(sha1); 105 + if (!cmit) 106 + usage(describe_usage); 107 + 108 if (!initialized) { 109 initialized = 1; 110 for_each_ref(get_name); 111------------ 112 1131. It is preceded with a "git diff" header, that looks like 114 this (when '-c' option is used): 115 116 diff --combined file 117+ 118or like this (when '--cc' option is used): 119 120 diff --cc file 121 1222. It is followed by one or more extended header lines 123 (this example shows a merge with two parents): 124 125 index <hash>,<hash>..<hash> 126 mode <mode>,<mode>..<mode> 127 new file mode <mode> 128 deleted file mode <mode>,<mode> 129+ 130The `mode <mode>,<mode>..<mode>` line appears only if at least one of 131the <mode> is different from the rest. Extended headers with 132information about detected contents movement (renames and 133copying detection) are designed to work with diff of two 134<tree-ish> and are not used by combined diff format. 135 1363. It is followed by two-line from-file/to-file header 137 138 --- a/file 139 +++ b/file 140+ 141Similar to two-line header for traditional 'unified' diff 142format, `/dev/null` is used to signal created or deleted 143files. 144 1454. Chunk header format is modified to prevent people from 146 accidentally feeding it to `patch -p1`. Combined diff format 147 was created for review of merge commit changes, and was not 148 meant for apply. The change is similar to the change in the 149 extended 'index' header: 150 151 @@@ <from-file-range> <from-file-range> <to-file-range> @@@ 152+ 153There are (number of parents + 1) `@` characters in the chunk 154header for combined diff format. 155 156Unlike the traditional 'unified' diff format, which shows two 157files A and B with a single column that has `-` (minus -- 158appears in A but removed in B), `+` (plus -- missing in A but 159added to B), or `" "` (space -- unchanged) prefix, this format 160compares two or more files file1, file2,... with one file X, and 161shows how X differs from each of fileN. One column for each of 162fileN is prepended to the output line to note how X's line is 163different from it. 164 165A `-` character in the column N means that the line appears in 166fileN but it does not appear in the result. A `+` character 167in the column N means that the line appears in the result, 168and fileN does not have that line (in other words, the line was 169added, from the point of view of that parent). 170 171In the above example output, the function signature was changed 172from both files (hence two `-` removals from both file1 and 173file2, plus `++` to mean one line that was added does not appear 174in either file1 nor file2). Also eight other lines are the same 175from file1 but do not appear in file2 (hence prefixed with `{plus}`). 176 177When shown by `git diff-tree -c`, it compares the parents of a 178merge commit with the merge result (i.e. file1..fileN are the 179parents). When shown by `git diff-files -c`, it compares the 180two unresolved merge parents with the working tree file 181(i.e. file1 is stage 2 aka "our version", file2 is stage 3 aka 182"their version").