xref: /freebsd/contrib/diff/doc/diff.7 (revision f39bffc62c1395bde25d152c7f68fdf7cbaab414)
1.Dd 2015-03-02
2.Dt DIFF 7
3.Os
4.Sh NAME
5.Nm diff
6.Nd Comparing and Merging Files
7.Sh  Comparing and Merging Files
8.Sh  Overview
9Computer users often find occasion to ask how two files differ. Perhaps one
10file is a newer version of the other file. Or maybe the two files started
11out as identical copies but were changed by different people.
12.Pp
13You can use the
14.Xr diff
15command to show differences between two files, or each corresponding file
16in two directories.
17.Xr diff
18outputs differences between files line by line in any of several formats,
19selectable by command line options. This set of differences is often called
20a
21.Em diff
22or
23.Em patch .
24For files that are identical,
25.Xr diff
26normally produces no output; for binary (non-text) files,
27.Xr diff
28normally reports only that they are different.
29.Pp
30You can use the
31.Xr cmp
32command to show the byte and line numbers where two files differ.
33.Xr cmp
34can also show all the bytes that differ between the two files, side by side.
35A way to compare two files character by character is the Emacs command
36.Li M-x compare-windows .
37See Section.Dq Other Window ,
38for more information on that command.
39.Pp
40You can use the
41.Xr diff3
42command to show differences among three files. When two people have made independent
43changes to a common original,
44.Xr diff3
45can report the differences between the original and the two changed versions,
46and can produce a merged file that contains both persons' changes together
47with warnings about conflicts.
48.Pp
49You can use the
50.Xr sdiff
51command to merge two files interactively.
52.Pp
53You can use the set of differences produced by
54.Xr diff
55to distribute updates to text files (such as program source code) to other
56people. This method is especially useful when the differences are small compared
57to the complete files. Given
58.Xr diff
59output, you can use the
60.Xr patch
61program to update, or
62.Em patch ,
63a copy of the file. If you think of
64.Xr diff
65as subtracting one file from another to produce their difference, you can
66think of
67.Xr patch
68as adding the difference to one file to reproduce the other.
69.Pp
70This manual first concentrates on making diffs, and later shows how to use
71diffs to update files.
72.Pp
73GNU
74.Xr diff
75was written by Paul Eggert, Mike Haertel, David Hayes, Richard Stallman, and
76Len Tower. Wayne Davison designed and implemented the unified output format.
77The basic algorithm is described by Eugene W. Myers in \(lqAn O(ND) Difference
78Algorithm and its Variations\(rq,
79.Em Algorithmica
80Vol. 1 No. 2, 1986, pp. 251--266; and in \(lqA File Comparison Program\(rq, Webb Miller
81and Eugene W. Myers,
82.Em Software---Practice and Experience
83Vol. 15 No. 11, 1985, pp. 1025--1040. The algorithm was independently discovered
84as described by E. Ukkonen in \(lqAlgorithms for Approximate String Matching\(rq,
85.Em Information and Control
86Vol. 64, 1985, pp. 100--118. Unless the
87.Op --minimal
88option is used,
89.Xr diff
90uses a heuristic by Paul Eggert that limits the cost to O(N^1.5 log N) at
91the price of producing suboptimal output for large inputs with many differences.
92Related algorithms are surveyed by Alfred V. Aho in section 6.3 of \(lqAlgorithms
93for Finding Patterns in Strings\(rq,
94.Em Handbook of Theoretical Computer Science
95(Jan Van Leeuwen, ed.), Vol. A,
96.Em Algorithms and Complexity ,
97Elsevier/MIT Press, 1990, pp. 255--300.
98.Pp
99GNU
100.Xr diff3
101was written by Randy Smith. GNU
102.Xr sdiff
103was written by Thomas Lord. GNU
104.Xr cmp
105was written by Torbj\(:orn Granlund and David MacKenzie.
106.Pp
107GNU
108.Xr patch
109was written mainly by Larry Wall and Paul Eggert; several GNU enhancements
110were contributed by Wayne Davison and David MacKenzie. Parts of this manual
111are adapted from a manual page written by Larry Wall, with his permission.
112.Pp
113.Sh  What Comparison Means
114There are several ways to think about the differences between two files. One
115way to think of the differences is as a series of lines that were deleted
116from, inserted in, or changed in one file to produce the other file.
117.Xr diff
118compares two files line by line, finds groups of lines that differ, and reports
119each group of differing lines. It can report the differing lines in several
120formats, which have different purposes.
121.Pp
122GNU
123.Xr diff
124can show whether files are different without detailing the differences. It
125also provides ways to suppress certain kinds of differences that are not important
126to you. Most commonly, such differences are changes in the amount of white
127space between words or lines.
128.Xr diff
129also provides ways to suppress differences in alphabetic case or in lines
130that match a regular expression that you provide. These options can accumulate;
131for example, you can ignore changes in both white space and alphabetic case.
132.Pp
133Another way to think of the differences between two files is as a sequence
134of pairs of bytes that can be either identical or different.
135.Xr cmp
136reports the differences between two files byte by byte, instead of line by
137line. As a result, it is often more useful than
138.Xr diff
139for comparing binary files. For text files,
140.Xr cmp
141is useful mainly when you want to know only whether two files are identical,
142or whether one file is a prefix of the other.
143.Pp
144To illustrate the effect that considering changes byte by byte can have compared
145with considering them line by line, think of what happens if a single newline
146character is added to the beginning of a file. If that file is then compared
147with an otherwise identical file that lacks the newline at the beginning,
148.Xr diff
149will report that a blank line has been added to the file, while
150.Xr cmp
151will report that almost every byte of the two files differs.
152.Pp
153.Xr diff3
154normally compares three input files line by line, finds groups of lines that
155differ, and reports each group of differing lines. Its output is designed
156to make it easy to inspect two different sets of changes to the same file.
157.Pp
158.Ss  Hunks
159When comparing two files,
160.Xr diff
161finds sequences of lines common to both files, interspersed with groups of
162differing lines called
163.Em hunks .
164Comparing two identical files yields one sequence of common lines and no hunks,
165because no lines differ. Comparing two entirely different files yields no
166common lines and one large hunk that contains all lines of both files. In
167general, there are many ways to match up lines between two given files.
168.Xr diff
169tries to minimize the total hunk size by finding large sequences of common
170lines interspersed with small hunks of differing lines.
171.Pp
172For example, suppose the file
173.Pa F
174contains the three lines
175.Li a ,
176.Li b ,
177.Li c ,
178and the file
179.Pa G
180contains the same three lines in reverse order
181.Li c ,
182.Li b ,
183.Li a .
184If
185.Xr diff
186finds the line
187.Li c
188as common, then the command
189.Li diff F G
190produces this output:
191.Pp
192.Bd -literal -offset indent
1931,2d0
194< a
195< b
1963a2,3
197> b
198> a
199.Ed
200.Pp
201But if
202.Xr diff
203notices the common line
204.Li b
205instead, it produces this output:
206.Pp
207.Bd -literal -offset indent
2081c1
209< a
210---
211> c
2123c3
213< c
214---
215> a
216.Ed
217.Pp
218It is also possible to find
219.Li a
220as the common line.
221.Xr diff
222does not always find an optimal matching between the files; it takes shortcuts
223to run faster. But its output is usually close to the shortest possible. You
224can adjust this tradeoff with the
225.Op -d
226or
227.Op --minimal
228option (see Section
229.Dq diff Performance ) .
230.Pp
231.Ss  Suppressing Differences in Blank and Tab Spacing
232The
233.Op -E
234or
235.Op --ignore-tab-expansion
236option ignores the distinction between tabs and spaces on input. A tab is
237considered to be equivalent to the number of spaces to the next tab stop (see Section
238.Dq Tabs ) .
239.Pp
240The
241.Op -b
242or
243.Op --ignore-space-change
244option is stronger. It ignores white space at line end, and considers all
245other sequences of one or more white space characters within a line to be
246equivalent. With this option,
247.Xr diff
248considers the following two lines to be equivalent, where
249.Li $
250denotes the line end:
251.Pp
252.Bd -literal -offset indent
253Here lyeth  muche rychnesse  in lytell space.   -- John Heywood$
254Here lyeth muche rychnesse in lytell space. -- John Heywood   $
255.Ed
256.Pp
257The
258.Op -w
259or
260.Op --ignore-all-space
261option is stronger still. It ignores differences even if one line has white
262space where the other line has none.
263.Em White space
264characters include tab, newline, vertical tab, form feed, carriage return,
265and space; some locales may define additional characters to be white space.
266With this option,
267.Xr diff
268considers the following two lines to be equivalent, where
269.Li $
270denotes the line end and
271.Li ^M
272denotes a carriage return:
273.Pp
274.Bd -literal -offset indent
275Here lyeth  muche  rychnesse in lytell space.--  John Heywood$
276  He relyeth much erychnes  seinly tells pace.  --John Heywood   ^M$
277.Ed
278.Pp
279.Ss  Suppressing Differences Whose Lines Are All Blank
280The
281.Op -B
282or
283.Op --ignore-blank-lines
284option ignores changes that consist entirely of blank lines. With this option,
285for example, a file containing
286.Bd -literal -offset indent
2871.  A point is that which has no part.
288
2892.  A line is breadthless length.
290-- Euclid, The Elements, I
291.Ed
292is considered identical to a file containing
293.Bd -literal -offset indent
2941.  A point is that which has no part.
2952.  A line is breadthless length.
296
297
298-- Euclid, The Elements, I
299.Ed
300.Pp
301Normally this option affects only lines that are completely empty, but if
302you also specify the
303.Op -b
304or
305.Op --ignore-space-change
306option, or the
307.Op -w
308or
309.Op --ignore-all-space
310option, lines are also affected if they look empty but contain white space.
311In other words,
312.Op -B
313is equivalent to
314.Li -I '^$'
315by default, but it is equivalent to
316.Op -I '^[[:space:]]*$'
317if
318.Op -b
319or
320.Op -w
321is also specified.
322.Pp
323.Ss  Suppressing Differences Whose Lines All Match a Regular Expression
324To ignore insertions and deletions of lines that match a
325.Xr grep
326-style regular expression, use the
327.Op -I Va regexp
328or
329.Op --ignore-matching-lines= Va regexp
330option. You should escape regular expressions that contain shell metacharacters
331to prevent the shell from expanding them. For example,
332.Li diff -I '^[[:digit:]]'
333ignores all changes to lines beginning with a digit.
334.Pp
335However,
336.Op -I
337only ignores the insertion or deletion of lines that contain the regular expression
338if every changed line in the hunk---every insertion and every deletion---matches
339the regular expression. In other words, for each nonignorable change,
340.Xr diff
341prints the complete set of changes in its vicinity, including the ignorable
342ones.
343.Pp
344You can specify more than one regular expression for lines to ignore by using
345more than one
346.Op -I
347option.
348.Xr diff
349tries to match each line against each regular expression.
350.Pp
351.Ss  Suppressing Case Differences
352GNU
353.Xr diff
354can treat lower case letters as equivalent to their upper case counterparts,
355so that, for example, it considers
356.Li Funky Stuff ,
357.Li funky STUFF ,
358and
359.Li fUNKy stuFf
360to all be the same. To request this, use the
361.Op -i
362or
363.Op --ignore-case
364option.
365.Pp
366.Ss  Summarizing Which Files Differ
367When you only want to find out whether files are different, and you don't
368care what the differences are, you can use the summary output format. In this
369format, instead of showing the differences between the files,
370.Xr diff
371simply reports whether files differ. The
372.Op -q
373or
374.Op --brief
375option selects this output format.
376.Pp
377This format is especially useful when comparing the contents of two directories.
378It is also much faster than doing the normal line by line comparisons, because
379.Xr diff
380can stop analyzing the files as soon as it knows that there are any differences.
381.Pp
382You can also get a brief indication of whether two files differ by using
383.Xr cmp .
384For files that are identical,
385.Xr cmp
386produces no output. When the files differ, by default,
387.Xr cmp
388outputs the byte and line number where the first difference occurs, or reports
389that one file is a prefix of the other. You can use the
390.Op -s ,
391.Op --quiet ,
392or
393.Op --silent
394option to suppress that information, so that
395.Xr cmp
396produces no output and reports whether the files differ using only its exit
397status (see Section
398.Dq Invoking cmp ) .
399.Pp
400Unlike
401.Xr diff ,
402.Xr cmp
403cannot compare directories; it can only compare two files.
404.Pp
405.Ss  Binary Files and Forcing Text Comparisons
406If
407.Xr diff
408thinks that either of the two files it is comparing is binary (a non-text
409file), it normally treats that pair of files much as if the summary output
410format had been selected (see Section
411.Dq Brief ) ,
412and reports only that the binary files are different. This is because line
413by line comparisons are usually not meaningful for binary files.
414.Pp
415.Xr diff
416determines whether a file is text or binary by checking the first few bytes
417in the file; the exact number of bytes is system dependent, but it is typically
418several thousand. If every byte in that part of the file is non-null,
419.Xr diff
420considers the file to be text; otherwise it considers the file to be binary.
421.Pp
422Sometimes you might want to force
423.Xr diff
424to consider files to be text. For example, you might be comparing text files
425that contain null characters;
426.Xr diff
427would erroneously decide that those are non-text files. Or you might be comparing
428documents that are in a format used by a word processing system that uses
429null characters to indicate special formatting. You can force
430.Xr diff
431to consider all files to be text files, and compare them line by line, by
432using the
433.Op -a
434or
435.Op --text
436option. If the files you compare using this option do not in fact contain
437text, they will probably contain few newline characters, and the
438.Xr diff
439output will consist of hunks showing differences between long lines of whatever
440characters the files contain.
441.Pp
442You can also force
443.Xr diff
444to report only whether files differ (but not how). Use the
445.Op -q
446or
447.Op --brief
448option for this.
449.Pp
450Normally, differing binary files count as trouble because the resulting
451.Xr diff
452output does not capture all the differences. This trouble causes
453.Xr diff
454to exit with status 2. However, this trouble cannot occur with the
455.Op -a
456or
457.Op --text
458option, or with the
459.Op -q
460or
461.Op --brief
462option, as these options both cause
463.Xr diff
464to generate a form of output that represents differences as requested.
465.Pp
466In operating systems that distinguish between text and binary files,
467.Xr diff
468normally reads and writes all data as text. Use the
469.Op --binary
470option to force
471.Xr diff
472to read and write binary data instead. This option has no effect on a POSIX-compliant
473system like GNU or traditional Unix. However, many personal computer operating
474systems represent the end of a line with a carriage return followed by a newline.
475On such systems,
476.Xr diff
477normally ignores these carriage returns on input and generates them at the
478end of each output line, but with the
479.Op --binary
480option
481.Xr diff
482treats each carriage return as just another input character, and does not
483generate a carriage return at the end of each output line. This can be useful
484when dealing with non-text files that are meant to be interchanged with POSIX-compliant
485systems.
486.Pp
487The
488.Op --strip-trailing-cr
489causes
490.Xr diff
491to treat input lines that end in carriage return followed by newline as if
492they end in plain newline. This can be useful when comparing text that is
493imperfectly imported from many personal computer operating systems. This option
494affects how lines are read, which in turn affects how they are compared and
495output.
496.Pp
497If you want to compare two files byte by byte, you can use the
498.Xr cmp
499program with the
500.Op -l
501or
502.Op --verbose
503option to show the values of each differing byte in the two files. With GNU
504.Xr cmp ,
505you can also use the
506.Op -b
507or
508.Op --print-bytes
509option to show the ASCII representation of those bytes.See Section
510.Dq Invoking cmp ,
511for more information.
512.Pp
513If
514.Xr diff3
515thinks that any of the files it is comparing is binary (a non-text file),
516it normally reports an error, because such comparisons are usually not useful.
517.Xr diff3
518uses the same test as
519.Xr diff
520to decide whether a file is binary. As with
521.Xr diff ,
522if the input files contain a few non-text bytes but otherwise are like text
523files, you can force
524.Xr diff3
525to consider all files to be text files and compare them line by line by using
526the
527.Op -a
528or
529.Op --text
530option.
531.Pp
532.Sh  Xr diff Output Formats
533.Xr diff
534has several mutually exclusive options for output format. The following sections
535describe each format, illustrating how
536.Xr diff
537reports the differences between two sample input files.
538.Pp
539.Ss  Two Sample Input Files
540Here are two sample files that we will use in numerous examples to illustrate
541the output of
542.Xr diff
543and how various options can change it.
544.Pp
545This is the file
546.Pa lao :
547.Pp
548.Bd -literal -offset indent
549The Way that can be told of is not the eternal Way;
550The name that can be named is not the eternal name.
551The Nameless is the origin of Heaven and Earth;
552The Named is the mother of all things.
553Therefore let there always be non-being,
554  so we may see their subtlety,
555And let there always be being,
556  so we may see their outcome.
557The two are the same,
558But after they are produced,
559  they have different names.
560.Ed
561.Pp
562This is the file
563.Pa tzu :
564.Pp
565.Bd -literal -offset indent
566The Nameless is the origin of Heaven and Earth;
567The named is the mother of all things.
568
569Therefore let there always be non-being,
570  so we may see their subtlety,
571And let there always be being,
572  so we may see their outcome.
573The two are the same,
574But after they are produced,
575  they have different names.
576They both may be called deep and profound.
577Deeper and more profound,
578The door of all subtleties!
579.Ed
580.Pp
581In this example, the first hunk contains just the first two lines of
582.Pa lao ,
583the second hunk contains the fourth line of
584.Pa lao
585opposing the second and third lines of
586.Pa tzu ,
587and the last hunk contains just the last three lines of
588.Pa tzu .
589.Pp
590.Ss  Showing Differences in Their Context
591Usually, when you are looking at the differences between files, you will also
592want to see the parts of the files near the lines that differ, to help you
593understand exactly what has changed. These nearby parts of the files are called
594the
595.Em context .
596.Pp
597GNU
598.Xr diff
599provides two output formats that show context around the differing lines:
600.Em context format
601and
602.Em unified format .
603It can optionally show in which function or section of the file the differing
604lines are found.
605.Pp
606If you are distributing new versions of files to other people in the form
607of
608.Xr diff
609output, you should use one of the output formats that show context so that
610they can apply the diffs even if they have made small changes of their own
611to the files.
612.Xr patch
613can apply the diffs in this case by searching in the files for the lines of
614context around the differing lines; if those lines are actually a few lines
615away from where the diff says they are,
616.Xr patch
617can adjust the line numbers accordingly and still apply the diff correctly.See Section
618.Dq Imperfect ,
619for more information on using
620.Xr patch
621to apply imperfect diffs.
622.Pp
623.Em  Context Format
624.Pp
625The context output format shows several lines of context around the lines
626that differ. It is the standard format for distributing updates to source
627code.
628.Pp
629To select this output format, use the
630.Op -C Va lines ,
631.Op --context[= Va lines] ,
632or
633.Op -c
634option. The argument
635.Va lines
636that some of these options take is the number of lines of context to show.
637If you do not specify
638.Va lines ,
639it defaults to three. For proper operation,
640.Xr patch
641typically needs at least two lines of context.
642.Pp
643.No  An Example of Context Format
644.Pp
645Here is the output of
646.Li diff -c lao tzu
647(see Section
648.Dq Sample diff Input ,
649for the complete contents of the two files). Notice that up to three lines
650that are not different are shown around each line that is different; they
651are the context lines. Also notice that the first two hunks have run together,
652because their contents overlap.
653.Pp
654.Bd -literal -offset indent
655*** lao	2002-02-21 23:30:39.942229878 -0800
656--- tzu	2002-02-21 23:30:50.442260588 -0800
657***************
658*** 1,7 ****
659- The Way that can be told of is not the eternal Way;
660- The name that can be named is not the eternal name.
661  The Nameless is the origin of Heaven and Earth;
662! The Named is the mother of all things.
663  Therefore let there always be non-being,
664    so we may see their subtlety,
665  And let there always be being,
666--- 1,6 ----
667  The Nameless is the origin of Heaven and Earth;
668! The named is the mother of all things.
669!
670  Therefore let there always be non-being,
671    so we may see their subtlety,
672  And let there always be being,
673***************
674*** 9,11 ****
675--- 8,13 ----
676  The two are the same,
677  But after they are produced,
678    they have different names.
679+ They both may be called deep and profound.
680+ Deeper and more profound,
681+ The door of all subtleties!
682.Ed
683.Pp
684.No  An Example of Context Format with Less Context
685.Pp
686Here is the output of
687.Li diff -C 1 lao tzu
688(see Section
689.Dq Sample diff Input ,
690for the complete contents of the two files). Notice that at most one context
691line is reported here.
692.Pp
693.Bd -literal -offset indent
694*** lao	2002-02-21 23:30:39.942229878 -0800
695--- tzu	2002-02-21 23:30:50.442260588 -0800
696***************
697*** 1,5 ****
698- The Way that can be told of is not the eternal Way;
699- The name that can be named is not the eternal name.
700  The Nameless is the origin of Heaven and Earth;
701! The Named is the mother of all things.
702  Therefore let there always be non-being,
703--- 1,4 ----
704  The Nameless is the origin of Heaven and Earth;
705! The named is the mother of all things.
706!
707  Therefore let there always be non-being,
708***************
709*** 11 ****
710--- 10,13 ----
711    they have different names.
712+ They both may be called deep and profound.
713+ Deeper and more profound,
714+ The door of all subtleties!
715.Ed
716.Pp
717.No  Detailed Description of Context Format
718.Pp
719The context output format starts with a two-line header, which looks like
720this:
721.Pp
722.Bd -literal -offset indent
723*** from-file from-file-modification-time
724--- to-file to-file-modification time
725.Ed
726.Pp
727The time stamp normally looks like
728.Li 2002-02-21 23:30:39.942229878 -0800
729to indicate the date, time with fractional seconds, and time zone in
730.Lk ftp://ftp.isi.edu/in-notes/rfc2822.txt .
731(The fractional seconds are omitted on hosts that do not support fractional
732time stamps.) However, a traditional time stamp like
733.Li Thu Feb 21 23:30:39 2002
734is used if the
735.Ev LC_TIME
736locale category is either
737.Li C
738or
739.Li POSIX .
740.Pp
741You can change the header's content with the
742.Op --label= Va label
743option; see Alternate Names.
744.Pp
745Next come one or more hunks of differences; each hunk shows one area where
746the files differ. Context format hunks look like this:
747.Pp
748.Bd -literal -offset indent
749***************
750*** from-file-line-numbers ****
751  from-file-line
752  from-file-line...
753--- to-file-line-numbers ----
754  to-file-line
755  to-file-line...
756.Ed
757.Pp
758If a hunk contains two or more lines, its line numbers look like
759.Li  Va start, Va end .
760Otherwise only its end line number appears. An empty hunk is considered to
761end at the line that precedes the hunk.
762.Pp
763The lines of context around the lines that differ start with two space characters.
764The lines that differ between the two files start with one of the following
765indicator characters, followed by a space character:
766.Pp
767.Bl -tag -width Ds
768.It  !
769A line that is part of a group of one or more lines that changed between the
770two files. There is a corresponding group of lines marked with
771.Li !
772in the part of this hunk for the other file.
773.Pp
774.It  +
775An \(lqinserted\(rq line in the second file that corresponds to nothing in the first
776file.
777.Pp
778.It  -
779A \(lqdeleted\(rq line in the first file that corresponds to nothing in the second
780file.
781.El
782.Pp
783If all of the changes in a hunk are insertions, the lines of
784.Va from-file
785are omitted. If all of the changes are deletions, the lines of
786.Va to-file
787are omitted.
788.Pp
789.Em  Unified Format
790.Pp
791The unified output format is a variation on the context format that is more
792compact because it omits redundant context lines. To select this output format,
793use the
794.Op -U Va lines ,
795.Op --unified[= Va lines] ,
796or
797.Op -u
798option. The argument
799.Va lines
800is the number of lines of context to show. When it is not given, it defaults
801to three.
802.Pp
803At present, only GNU
804.Xr diff
805can produce this format and only GNU
806.Xr patch
807can automatically apply diffs in this format. For proper operation,
808.Xr patch
809typically needs at least three lines of context.
810.Pp
811.No  An Example of Unified Format
812.Pp
813Here is the output of the command
814.Li diff -u lao tzu
815(see Section
816.Dq Sample diff Input ,
817for the complete contents of the two files):
818.Pp
819.Bd -literal -offset indent
820--- lao	2002-02-21 23:30:39.942229878 -0800
821+++ tzu	2002-02-21 23:30:50.442260588 -0800
822@@ -1,7 +1,6 @@
823-The Way that can be told of is not the eternal Way;
824-The name that can be named is not the eternal name.
825 The Nameless is the origin of Heaven and Earth;
826-The Named is the mother of all things.
827+The named is the mother of all things.
828+
829 Therefore let there always be non-being,
830   so we may see their subtlety,
831 And let there always be being,
832@@ -9,3 +8,6 @@
833 The two are the same,
834 But after they are produced,
835   they have different names.
836+They both may be called deep and profound.
837+Deeper and more profound,
838+The door of all subtleties!
839.Ed
840.Pp
841.No  Detailed Description of Unified Format
842.Pp
843The unified output format starts with a two-line header, which looks like
844this:
845.Pp
846.Bd -literal -offset indent
847--- from-file from-file-modification-time
848+++ to-file to-file-modification-time
849.Ed
850.Pp
851The time stamp looks like
852.Li 2002-02-21 23:30:39.942229878 -0800
853to indicate the date, time with fractional seconds, and time zone. The fractional
854seconds are omitted on hosts that do not support fractional time stamps.
855.Pp
856You can change the header's content with the
857.Op --label= Va label
858option; seeSee Section
859.Dq Alternate Names .
860.Pp
861Next come one or more hunks of differences; each hunk shows one area where
862the files differ. Unified format hunks look like this:
863.Pp
864.Bd -literal -offset indent
865@@ from-file-line-numbers to-file-line-numbers @@
866 line-from-either-file
867 line-from-either-file...
868.Ed
869.Pp
870If a hunk contains just one line, only its start line number appears. Otherwise
871its line numbers look like
872.Li  Va start, Va count .
873An empty hunk is considered to start at the line that follows the hunk.
874.Pp
875If a hunk and its context contain two or more lines, its line numbers look
876like
877.Li  Va start, Va count .
878Otherwise only its end line number appears. An empty hunk is considered to
879end at the line that precedes the hunk.
880.Pp
881The lines common to both files begin with a space character. The lines that
882actually differ between the two files have one of the following indicator
883characters in the left print column:
884.Pp
885.Bl -tag -width Ds
886.It  +
887A line was added here to the first file.
888.Pp
889.It  -
890A line was removed here from the first file.
891.El
892.Pp
893.Em  Showing Which Sections Differences Are in
894.Pp
895Sometimes you might want to know which part of the files each change falls
896in. If the files are source code, this could mean which function was changed.
897If the files are documents, it could mean which chapter or appendix was changed.
898GNU
899.Xr diff
900can show this by displaying the nearest section heading line that precedes
901the differing lines. Which lines are \(lqsection headings\(rq is determined by a regular
902expression.
903.Pp
904.No  Showing Lines That Match Regular Expressions
905.Pp
906To show in which sections differences occur for files that are not source
907code for C or similar languages, use the
908.Op -F Va regexp
909or
910.Op --show-function-line= Va regexp
911option.
912.Xr diff
913considers lines that match the
914.Xr grep
915-style regular expression
916.Va regexp
917to be the beginning of a section of the file. Here are suggested regular expressions
918for some common languages:
919.Pp
920.Bl -tag -width Ds
921.It  ^[[:alpha:]$_]
922C, C++, Prolog
923.It  ^(
924Lisp
925.It  ^@node
926Texinfo
927.El
928.Pp
929This option does not automatically select an output format; in order to use
930it, you must select the context format (see Section
931.Dq Context Format )
932or unified format (see Section
933.Dq Unified Format ) .
934In other output formats it has no effect.
935.Pp
936The
937.Op -F
938or
939.Op --show-function-line
940option finds the nearest unchanged line that precedes each hunk of differences
941and matches the given regular expression. Then it adds that line to the end
942of the line of asterisks in the context format, or to the
943.Li @@
944line in unified format. If no matching line exists, this option leaves the
945output for that hunk unchanged. If that line is more than 40 characters long,
946it outputs only the first 40 characters. You can specify more than one regular
947expression for such lines;
948.Xr diff
949tries to match each line against each regular expression, starting with the
950last one given. This means that you can use
951.Op -p
952and
953.Op -F
954together, if you wish.
955.Pp
956.No  Showing C Function Headings
957.Pp
958To show in which functions differences occur for C and similar languages,
959you can use the
960.Op -p
961or
962.Op --show-c-function
963option. This option automatically defaults to the context output format (see Section
964.Dq Context Format ) ,
965with the default number of lines of context. You can override that number
966with
967.Op -C Va lines
968elsewhere in the command line. You can override both the format and the number
969with
970.Op -U Va lines
971elsewhere in the command line.
972.Pp
973The
974.Op -p
975or
976.Op --show-c-function
977option is equivalent to
978.Op -F '^[[:alpha:]$_]'
979if the unified format is specified, otherwise
980.Op -c -F '^[[:alpha:]$_]'
981(see Section
982.Dq Specified Headings ) .
983GNU
984.Xr diff
985provides this option for the sake of convenience.
986.Pp
987.Em  Showing Alternate File Names
988.Pp
989If you are comparing two files that have meaningless or uninformative names,
990you might want
991.Xr diff
992to show alternate names in the header of the context and unified output formats.
993To do this, use the
994.Op --label= Va label
995option. The first time you give this option, its argument replaces the name
996and date of the first file in the header; the second time, its argument replaces
997the name and date of the second file. If you give this option more than twice,
998.Xr diff
999reports an error. The
1000.Op --label
1001option does not affect the file names in the
1002.Xr pr
1003header when the
1004.Op -l
1005or
1006.Op --paginate
1007option is used (see Section
1008.Dq Pagination ) .
1009.Pp
1010Here are the first two lines of the output from
1011.Li diff -C 2 --label=original --label=modified lao tzu :
1012.Pp
1013.Bd -literal -offset indent
1014*** original
1015--- modified
1016.Ed
1017.Pp
1018.Ss  Showing Differences Side by Side
1019.Xr diff
1020can produce a side by side difference listing of two files. The files are
1021listed in two columns with a gutter between them. The gutter contains one
1022of the following markers:
1023.Pp
1024.Bl -tag -width Ds
1025.It  white space
1026The corresponding lines are in common. That is, either the lines are identical,
1027or the difference is ignored because of one of the
1028.Op --ignore
1029options (see Section
1030.Dq White Space ) .
1031.Pp
1032.It  Li |
1033The corresponding lines differ, and they are either both complete or both
1034incomplete.
1035.Pp
1036.It  Li <
1037The files differ and only the first file contains the line.
1038.Pp
1039.It  Li >
1040The files differ and only the second file contains the line.
1041.Pp
1042.It  Li (
1043Only the first file contains the line, but the difference is ignored.
1044.Pp
1045.It  Li )
1046Only the second file contains the line, but the difference is ignored.
1047.Pp
1048.It  Li \e
1049The corresponding lines differ, and only the first line is incomplete.
1050.Pp
1051.It  Li /
1052The corresponding lines differ, and only the second line is incomplete.
1053.El
1054.Pp
1055Normally, an output line is incomplete if and only if the lines that it contains
1056are incomplete;See Section
1057.Dq Incomplete Lines .
1058However, when an output line represents two differing lines, one might be
1059incomplete while the other is not. In this case, the output line is complete,
1060but its the gutter is marked
1061.Li \e
1062if the first line is incomplete,
1063.Li /
1064if the second line is.
1065.Pp
1066Side by side format is sometimes easiest to read, but it has limitations.
1067It generates much wider output than usual, and truncates lines that are too
1068long to fit. Also, it relies on lining up output more heavily than usual,
1069so its output looks particularly bad if you use varying width fonts, nonstandard
1070tab stops, or nonprinting characters.
1071.Pp
1072You can use the
1073.Xr sdiff
1074command to interactively merge side by side differences.See Section
1075.Dq Interactive Merging ,
1076for more information on merging files.
1077.Pp
1078.Em  Controlling Side by Side Format
1079.Pp
1080The
1081.Op -y
1082or
1083.Op --side-by-side
1084option selects side by side format. Because side by side output lines contain
1085two input lines, the output is wider than usual: normally 130 print columns,
1086which can fit onto a traditional printer line. You can set the width of the
1087output with the
1088.Op -W Va columns
1089or
1090.Op --width= Va columns
1091option. The output is split into two halves of equal width, separated by a
1092small gutter to mark differences; the right half is aligned to a tab stop
1093so that tabs line up. Input lines that are too long to fit in half of an output
1094line are truncated for output.
1095.Pp
1096The
1097.Op --left-column
1098option prints only the left column of two common lines. The
1099.Op --suppress-common-lines
1100option suppresses common lines entirely.
1101.Pp
1102.Em  An Example of Side by Side Format
1103.Pp
1104Here is the output of the command
1105.Li diff -y -W 72 lao tzu
1106(see Section
1107.Dq Sample diff Input ,
1108for the complete contents of the two files).
1109.Pp
1110.Bd -literal -offset indent
1111The Way that can be told of is n   <
1112The name that can be named is no   <
1113The Nameless is the origin of He        The Nameless is the origin of He
1114The Named is the mother of all t   |    The named is the mother of all t
1115                                   >
1116Therefore let there always be no        Therefore let there always be no
1117  so we may see their subtlety,           so we may see their subtlety,
1118And let there always be being,          And let there always be being,
1119  so we may see their outcome.            so we may see their outcome.
1120The two are the same,                   The two are the same,
1121But after they are produced,            But after they are produced,
1122  they have different names.              they have different names.
1123                                   >    They both may be called deep and
1124                                   >    Deeper and more profound,
1125                                   >    The door of all subtleties!
1126.Ed
1127.Pp
1128.Ss  Showing Differences Without Context
1129The \(lqnormal\(rq
1130.Xr diff
1131output format shows each hunk of differences without any surrounding context.
1132Sometimes such output is the clearest way to see how lines have changed, without
1133the clutter of nearby unchanged lines (although you can get similar results
1134with the context or unified formats by using 0 lines of context). However,
1135this format is no longer widely used for sending out patches; for that purpose,
1136the context format (see Section
1137.Dq Context Format )
1138and the unified format (see Section
1139.Dq Unified Format )
1140are superior. Normal format is the default for compatibility with older versions
1141of
1142.Xr diff
1143and the POSIX standard. Use the
1144.Op --normal
1145option to select this output format explicitly.
1146.Pp
1147.Em  An Example of Normal Format
1148.Pp
1149Here is the output of the command
1150.Li diff lao tzu
1151(see Section
1152.Dq Sample diff Input ,
1153for the complete contents of the two files). Notice that it shows only the
1154lines that are different between the two files.
1155.Pp
1156.Bd -literal -offset indent
11571,2d0
1158< The Way that can be told of is not the eternal Way;
1159< The name that can be named is not the eternal name.
11604c2,3
1161< The Named is the mother of all things.
1162---
1163> The named is the mother of all things.
1164>
116511a11,13
1166> They both may be called deep and profound.
1167> Deeper and more profound,
1168> The door of all subtleties!
1169.Ed
1170.Pp
1171.Em  Detailed Description of Normal Format
1172.Pp
1173The normal output format consists of one or more hunks of differences; each
1174hunk shows one area where the files differ. Normal format hunks look like
1175this:
1176.Pp
1177.Bd -literal -offset indent
1178change-command
1179< from-file-line
1180< from-file-line...
1181---
1182> to-file-line
1183> to-file-line...
1184.Ed
1185.Pp
1186There are three types of change commands. Each consists of a line number or
1187comma-separated range of lines in the first file, a single character indicating
1188the kind of change to make, and a line number or comma-separated range of
1189lines in the second file. All line numbers are the original line numbers in
1190each file. The types of change commands are:
1191.Pp
1192.Bl -tag -width Ds
1193.It  Va la Va r
1194Add the lines in range
1195.Va r
1196of the second file after line
1197.Va l
1198of the first file. For example,
1199.Li 8a12,15
1200means append lines 12--15 of file 2 after line 8 of file 1; or, if changing
1201file 2 into file 1, delete lines 12--15 of file 2.
1202.Pp
1203.It  Va fc Va t
1204Replace the lines in range
1205.Va f
1206of the first file with lines in range
1207.Va t
1208of the second file. This is like a combined add and delete, but more compact.
1209For example,
1210.Li 5,7c8,10
1211means change lines 5--7 of file 1 to read as lines 8--10 of file 2; or, if
1212changing file 2 into file 1, change lines 8--10 of file 2 to read as lines
12135--7 of file 1.
1214.Pp
1215.It  Va rd Va l
1216Delete the lines in range
1217.Va r
1218from the first file; line
1219.Va l
1220is where they would have appeared in the second file had they not been deleted.
1221For example,
1222.Li 5,7d3
1223means delete lines 5--7 of file 1; or, if changing file 2 into file 1, append
1224lines 5--7 of file 1 after line 3 of file 2.
1225.El
1226.Pp
1227.Ss  Making Edit Scripts
1228Several output modes produce command scripts for editing
1229.Va from-file
1230to produce
1231.Va to-file .
1232.Pp
1233.Em  Xr ed Scripts
1234.Pp
1235.Xr diff
1236can produce commands that direct the
1237.Xr ed
1238text editor to change the first file into the second file. Long ago, this
1239was the only output mode that was suitable for editing one file into another
1240automatically; today, with
1241.Xr patch ,
1242it is almost obsolete. Use the
1243.Op -e
1244or
1245.Op --ed
1246option to select this output format.
1247.Pp
1248Like the normal format (see Section
1249.Dq Normal ) ,
1250this output format does not show any context; unlike the normal format, it
1251does not include the information necessary to apply the diff in reverse (to
1252produce the first file if all you have is the second file and the diff).
1253.Pp
1254If the file
1255.Pa d
1256contains the output of
1257.Li diff -e old new ,
1258then the command
1259.Li (cat d && echo w) | ed - old
1260edits
1261.Pa old
1262to make it a copy of
1263.Pa new .
1264More generally, if
1265.Pa d1 ,
1266.Pa d2 ,
1267\&...,
1268.Pa dN
1269contain the outputs of
1270.Li diff -e old new1 ,
1271.Li diff -e new1 new2 ,
1272\&...,
1273.Li diff -e newN-1 newN ,
1274respectively, then the command
1275.Li (cat d1 d2 ... dN && echo w) | ed - old
1276edits
1277.Pa old
1278to make it a copy of
1279.Pa newN .
1280.Pp
1281.No  Example Xr ed Script
1282.Pp
1283Here is the output of
1284.Li diff -e lao tzu
1285(see Section
1286.Dq Sample diff Input ,
1287for the complete contents of the two files):
1288.Pp
1289.Bd -literal -offset indent
129011a
1291They both may be called deep and profound.
1292Deeper and more profound,
1293The door of all subtleties!
1294\&.
12954c
1296The named is the mother of all things.
1297
1298\&.
12991,2d
1300.Ed
1301.Pp
1302.No  Detailed Description of Xr ed Format
1303.Pp
1304The
1305.Xr ed
1306output format consists of one or more hunks of differences. The changes closest
1307to the ends of the files come first so that commands that change the number
1308of lines do not affect how
1309.Xr ed
1310interprets line numbers in succeeding commands.
1311.Xr ed
1312format hunks look like this:
1313.Pp
1314.Bd -literal -offset indent
1315change-command
1316to-file-line
1317to-file-line...
1318\&.
1319.Ed
1320.Pp
1321Because
1322.Xr ed
1323uses a single period on a line to indicate the end of input, GNU
1324.Xr diff
1325protects lines of changes that contain a single period on a line by writing
1326two periods instead, then writing a subsequent
1327.Xr ed
1328command to change the two periods into one. The
1329.Xr ed
1330format cannot represent an incomplete line, so if the second file ends in
1331a changed incomplete line,
1332.Xr diff
1333reports an error and then pretends that a newline was appended.
1334.Pp
1335There are three types of change commands. Each consists of a line number or
1336comma-separated range of lines in the first file and a single character indicating
1337the kind of change to make. All line numbers are the original line numbers
1338in the file. The types of change commands are:
1339.Pp
1340.Bl -tag -width Ds
1341.It  Va la
1342Add text from the second file after line
1343.Va l
1344in the first file. For example,
1345.Li 8a
1346means to add the following lines after line 8 of file 1.
1347.Pp
1348.It  Va rc
1349Replace the lines in range
1350.Va r
1351in the first file with the following lines. Like a combined add and delete,
1352but more compact. For example,
1353.Li 5,7c
1354means change lines 5--7 of file 1 to read as the text file 2.
1355.Pp
1356.It  Va rd
1357Delete the lines in range
1358.Va r
1359from the first file. For example,
1360.Li 5,7d
1361means delete lines 5--7 of file 1.
1362.El
1363.Pp
1364.Em  Forward Xr ed Scripts
1365.Pp
1366.Xr diff
1367can produce output that is like an
1368.Xr ed
1369script, but with hunks in forward (front to back) order. The format of the
1370commands is also changed slightly: command characters precede the lines they
1371modify, spaces separate line numbers in ranges, and no attempt is made to
1372disambiguate hunk lines consisting of a single period. Like
1373.Xr ed
1374format, forward
1375.Xr ed
1376format cannot represent incomplete lines.
1377.Pp
1378Forward
1379.Xr ed
1380format is not very useful, because neither
1381.Xr ed
1382nor
1383.Xr patch
1384can apply diffs in this format. It exists mainly for compatibility with older
1385versions of
1386.Xr diff .
1387Use the
1388.Op -f
1389or
1390.Op --forward-ed
1391option to select it.
1392.Pp
1393.Em  RCS Scripts
1394.Pp
1395The RCS output format is designed specifically for use by the Revision Control
1396System, which is a set of free programs used for organizing different versions
1397and systems of files. Use the
1398.Op -n
1399or
1400.Op --rcs
1401option to select this output format. It is like the forward
1402.Xr ed
1403format (see Section
1404.Dq Forward ed ) ,
1405but it can represent arbitrary changes to the contents of a file because it
1406avoids the forward
1407.Xr ed
1408format's problems with lines consisting of a single period and with incomplete
1409lines. Instead of ending text sections with a line consisting of a single
1410period, each command specifies the number of lines it affects; a combination
1411of the
1412.Li a
1413and
1414.Li d
1415commands are used instead of
1416.Li c .
1417Also, if the second file ends in a changed incomplete line, then the output
1418also ends in an incomplete line.
1419.Pp
1420Here is the output of
1421.Li diff -n lao tzu
1422(see Section
1423.Dq Sample diff Input ,
1424for the complete contents of the two files):
1425.Pp
1426.Bd -literal -offset indent
1427d1 2
1428d4 1
1429a4 2
1430The named is the mother of all things.
1431
1432a11 3
1433They both may be called deep and profound.
1434Deeper and more profound,
1435The door of all subtleties!
1436.Ed
1437.Pp
1438.Ss  Merging Files with If-then-else
1439You can use
1440.Xr diff
1441to merge two files of C source code. The output of
1442.Xr diff
1443in this format contains all the lines of both files. Lines common to both
1444files are output just once; the differing parts are separated by the C preprocessor
1445directives
1446.Li #ifdef Va name
1447or
1448.Li #ifndef Va name ,
1449.Li #else ,
1450and
1451.Li #endif .
1452When compiling the output, you select which version to use by either defining
1453or leaving undefined the macro
1454.Va name .
1455.Pp
1456To merge two files, use
1457.Xr diff
1458with the
1459.Op -D Va name
1460or
1461.Op --ifdef= Va name
1462option. The argument
1463.Va name
1464is the C preprocessor identifier to use in the
1465.Li #ifdef
1466and
1467.Li #ifndef
1468directives.
1469.Pp
1470For example, if you change an instance of
1471.Li wait (&s)
1472to
1473.Li waitpid (-1, &s, 0)
1474and then merge the old and new files with the
1475.Op --ifdef=HAVE_WAITPID
1476option, then the affected part of your code might look like this:
1477.Pp
1478.Bd -literal -offset indent
1479    do {
1480#ifndef HAVE_WAITPID
1481        if ((w = wait (&s)) < 0  &&  errno != EINTR)
1482#else /* HAVE_WAITPID */
1483        if ((w = waitpid (-1, &s, 0)) < 0  &&  errno != EINTR)
1484#endif /* HAVE_WAITPID */
1485            return w;
1486    } while (w != child);
1487.Ed
1488.Pp
1489You can specify formats for languages other than C by using line group formats
1490and line formats, as described in the next sections.
1491.Pp
1492.Em  Line Group Formats
1493.Pp
1494Line group formats let you specify formats suitable for many applications
1495that allow if-then-else input, including programming languages and text formatting
1496languages. A line group format specifies the output format for a contiguous
1497group of similar lines.
1498.Pp
1499For example, the following command compares the TeX files
1500.Pa old
1501and
1502.Pa new ,
1503and outputs a merged file in which old regions are surrounded by
1504.Li \ebegin{em}
1505-
1506.Li \eend{em}
1507lines, and new regions are surrounded by
1508.Li \ebegin{bf}
1509-
1510.Li \eend{bf}
1511lines.
1512.Pp
1513.Bd -literal -offset indent
1514diff \e
1515   --old-group-format='\ebegin{em}
1516%<\eend{em}
1517\&' \e
1518   --new-group-format='\ebegin{bf}
1519%>\eend{bf}
1520\&' \e
1521   old new
1522.Ed
1523.Pp
1524The following command is equivalent to the above example, but it is a little
1525more verbose, because it spells out the default line group formats.
1526.Pp
1527.Bd -literal -offset indent
1528diff \e
1529   --old-group-format='\ebegin{em}
1530%<\eend{em}
1531\&' \e
1532   --new-group-format='\ebegin{bf}
1533%>\eend{bf}
1534\&' \e
1535   --unchanged-group-format='%=' \e
1536   --changed-group-format='\ebegin{em}
1537%<\eend{em}
1538\ebegin{bf}
1539%>\eend{bf}
1540\&' \e
1541   old new
1542.Ed
1543.Pp
1544Here is a more advanced example, which outputs a diff listing with headers
1545containing line numbers in a \(lqplain English\(rq style.
1546.Pp
1547.Bd -literal -offset indent
1548diff \e
1549   --unchanged-group-format=\(rq \e
1550   --old-group-format='-------- %dn line%(n=1?:s) deleted at %df:
1551%<' \e
1552   --new-group-format='-------- %dN line%(N=1?:s) added after %de:
1553%>' \e
1554   --changed-group-format='-------- %dn line%(n=1?:s) changed at %df:
1555%<-------- to:
1556%>' \e
1557   old new
1558.Ed
1559.Pp
1560To specify a line group format, use
1561.Xr diff
1562with one of the options listed below. You can specify up to four line group
1563formats, one for each kind of line group. You should quote
1564.Va format ,
1565because it typically contains shell metacharacters.
1566.Pp
1567.Bl -tag -width Ds
1568.It  --old-group-format= Va format
1569These line groups are hunks containing only lines from the first file. The
1570default old group format is the same as the changed group format if it is
1571specified; otherwise it is a format that outputs the line group as-is.
1572.Pp
1573.It  --new-group-format= Va format
1574These line groups are hunks containing only lines from the second file. The
1575default new group format is same as the changed group format if it is specified;
1576otherwise it is a format that outputs the line group as-is.
1577.Pp
1578.It  --changed-group-format= Va format
1579These line groups are hunks containing lines from both files. The default
1580changed group format is the concatenation of the old and new group formats.
1581.Pp
1582.It  --unchanged-group-format= Va format
1583These line groups contain lines common to both files. The default unchanged
1584group format is a format that outputs the line group as-is.
1585.El
1586.Pp
1587In a line group format, ordinary characters represent themselves; conversion
1588specifications start with
1589.Li %
1590and have one of the following forms.
1591.Pp
1592.Bl -tag -width Ds
1593.It  %<
1594stands for the lines from the first file, including the trailing newline.
1595Each line is formatted according to the old line format (see Section
1596.Dq Line Formats ) .
1597.Pp
1598.It  %>
1599stands for the lines from the second file, including the trailing newline.
1600Each line is formatted according to the new line format.
1601.Pp
1602.It  %=
1603stands for the lines common to both files, including the trailing newline.
1604Each line is formatted according to the unchanged line format.
1605.Pp
1606.It  %%
1607stands for
1608.Li % .
1609.Pp
1610.It  %c' Va C'
1611where
1612.Va C
1613is a single character, stands for
1614.Va C .
1615.Va C
1616may not be a backslash or an apostrophe. For example,
1617.Li %c':'
1618stands for a colon, even inside the then-part of an if-then-else format, which
1619a colon would normally terminate.
1620.Pp
1621.It  %c'\e Va O'
1622where
1623.Va O
1624is a string of 1, 2, or 3 octal digits, stands for the character with octal
1625code
1626.Va O .
1627For example,
1628.Li %c'\e0'
1629stands for a null character.
1630.Pp
1631.It  Va F Va n
1632where
1633.Va F
1634is a
1635.Li printf
1636conversion specification and
1637.Va n
1638is one of the following letters, stands for
1639.Va n
1640\&'s value formatted with
1641.Va F .
1642.Pp
1643.Bl -tag -width Ds
1644.It  e
1645The line number of the line just before the group in the old file.
1646.Pp
1647.It  f
1648The line number of the first line in the group in the old file; equals
1649.Va e
1650+ 1.
1651.Pp
1652.It  l
1653The line number of the last line in the group in the old file.
1654.Pp
1655.It  m
1656The line number of the line just after the group in the old file; equals
1657.Va l
1658+ 1.
1659.Pp
1660.It  n
1661The number of lines in the group in the old file; equals
1662.Va l
1663-
1664.Va f
1665+ 1.
1666.Pp
1667.It  E, F, L, M, N
1668Likewise, for lines in the new file.
1669.Pp
1670.El
1671The
1672.Li printf
1673conversion specification can be
1674.Li %d ,
1675.Li %o ,
1676.Li %x ,
1677or
1678.Li %X ,
1679specifying decimal, octal, lower case hexadecimal, or upper case hexadecimal
1680output respectively. After the
1681.Li %
1682the following options can appear in sequence: a series of zero or more flags;
1683an integer specifying the minimum field width; and a period followed by an
1684optional integer specifying the minimum number of digits. The flags are
1685.Li -
1686for left-justification,
1687.Li '
1688for separating the digit into groups as specified by the
1689.Ev LC_NUMERIC
1690locale category, and
1691.Li 0
1692for padding with zeros instead of spaces. For example,
1693.Li %5dN
1694prints the number of new lines in the group in a field of width 5 characters,
1695using the
1696.Li printf
1697format
1698.Li "%5d" .
1699.Pp
1700.It  ( Va A= Va B? Va T: Va E)
1701If
1702.Va A
1703equals
1704.Va B
1705then
1706.Va T
1707else
1708.Va E .
1709.Va A
1710and
1711.Va B
1712are each either a decimal constant or a single letter interpreted as above.
1713This format spec is equivalent to
1714.Va T
1715if
1716.Va A
1717\&'s value equals
1718.Va B
1719\&'s; otherwise it is equivalent to
1720.Va E .
1721.Pp
1722For example,
1723.Li %(N=0?no:%dN) line%(N=1?:s)
1724is equivalent to
1725.Li no lines
1726if
1727.Va N
1728(the number of lines in the group in the new file) is 0, to
1729.Li 1 line
1730if
1731.Va N
1732is 1, and to
1733.Li %dN lines
1734otherwise.
1735.El
1736.Pp
1737.Em  Line Formats
1738.Pp
1739Line formats control how each line taken from an input file is output as part
1740of a line group in if-then-else format.
1741.Pp
1742For example, the following command outputs text with a one-character change
1743indicator to the left of the text. The first character of output is
1744.Li -
1745for deleted lines,
1746.Li |
1747for added lines, and a space for unchanged lines. The formats contain newline
1748characters where newlines are desired on output.
1749.Pp
1750.Bd -literal -offset indent
1751diff \e
1752   --old-line-format='-%l
1753\&' \e
1754   --new-line-format='|%l
1755\&' \e
1756   --unchanged-line-format=' %l
1757\&' \e
1758   old new
1759.Ed
1760.Pp
1761To specify a line format, use one of the following options. You should quote
1762.Va format ,
1763since it often contains shell metacharacters.
1764.Pp
1765.Bl -tag -width Ds
1766.It  --old-line-format= Va format
1767formats lines just from the first file.
1768.Pp
1769.It  --new-line-format= Va format
1770formats lines just from the second file.
1771.Pp
1772.It  --unchanged-line-format= Va format
1773formats lines common to both files.
1774.Pp
1775.It  --line-format= Va format
1776formats all lines; in effect, it sets all three above options simultaneously.
1777.El
1778.Pp
1779In a line format, ordinary characters represent themselves; conversion specifications
1780start with
1781.Li %
1782and have one of the following forms.
1783.Pp
1784.Bl -tag -width Ds
1785.It  %l
1786stands for the contents of the line, not counting its trailing newline (if
1787any). This format ignores whether the line is incomplete;See Section
1788.Dq Incomplete Lines .
1789.Pp
1790.It  %L
1791stands for the contents of the line, including its trailing newline (if any).
1792If a line is incomplete, this format preserves its incompleteness.
1793.Pp
1794.It  %%
1795stands for
1796.Li % .
1797.Pp
1798.It  %c' Va C'
1799where
1800.Va C
1801is a single character, stands for
1802.Va C .
1803.Va C
1804may not be a backslash or an apostrophe. For example,
1805.Li %c':'
1806stands for a colon.
1807.Pp
1808.It  %c'\e Va O'
1809where
1810.Va O
1811is a string of 1, 2, or 3 octal digits, stands for the character with octal
1812code
1813.Va O .
1814For example,
1815.Li %c'\e0'
1816stands for a null character.
1817.Pp
1818.It  Va Fn
1819where
1820.Va F
1821is a
1822.Li printf
1823conversion specification, stands for the line number formatted with
1824.Va F .
1825For example,
1826.Li %.5dn
1827prints the line number using the
1828.Li printf
1829format
1830.Li "%.5d" .
1831See Section.Dq Line Group Formats ,
1832for more about printf conversion specifications.
1833.Pp
1834.El
1835The default line format is
1836.Li %l
1837followed by a newline character.
1838.Pp
1839If the input contains tab characters and it is important that they line up
1840on output, you should ensure that
1841.Li %l
1842or
1843.Li %L
1844in a line format is just after a tab stop (e.g. by preceding
1845.Li %l
1846or
1847.Li %L
1848with a tab character), or you should use the
1849.Op -t
1850or
1851.Op --expand-tabs
1852option.
1853.Pp
1854Taken together, the line and line group formats let you specify many different
1855formats. For example, the following command uses a format similar to normal
1856.Xr diff
1857format. You can tailor this command to get fine control over
1858.Xr diff
1859output.
1860.Pp
1861.Bd -literal -offset indent
1862diff \e
1863   --old-line-format='< %l
1864\&' \e
1865   --new-line-format='> %l
1866\&' \e
1867   --old-group-format='%df%(f=l?:,%dl)d%dE
1868%<' \e
1869   --new-group-format='%dea%dF%(F=L?:,%dL)
1870%>' \e
1871   --changed-group-format='%df%(f=l?:,%dl)c%dF%(F=L?:,%dL)
1872%<---
1873%>' \e
1874   --unchanged-group-format=\(rq \e
1875   old new
1876.Ed
1877.Pp
1878.Em  An Example of If-then-else Format
1879.Pp
1880Here is the output of
1881.Li diff -DTWO lao tzu
1882(see Section
1883.Dq Sample diff Input ,
1884for the complete contents of the two files):
1885.Pp
1886.Bd -literal -offset indent
1887#ifndef TWO
1888The Way that can be told of is not the eternal Way;
1889The name that can be named is not the eternal name.
1890#endif /* ! TWO */
1891The Nameless is the origin of Heaven and Earth;
1892#ifndef TWO
1893The Named is the mother of all things.
1894#else /* TWO */
1895The named is the mother of all things.
1896
1897#endif /* TWO */
1898Therefore let there always be non-being,
1899  so we may see their subtlety,
1900And let there always be being,
1901  so we may see their outcome.
1902The two are the same,
1903But after they are produced,
1904  they have different names.
1905#ifdef TWO
1906They both may be called deep and profound.
1907Deeper and more profound,
1908The door of all subtleties!
1909#endif /* TWO */
1910.Ed
1911.Pp
1912.Em  Detailed Description of If-then-else Format
1913.Pp
1914For lines common to both files,
1915.Xr diff
1916uses the unchanged line group format. For each hunk of differences in the
1917merged output format, if the hunk contains only lines from the first file,
1918.Xr diff
1919uses the old line group format; if the hunk contains only lines from the second
1920file,
1921.Xr diff
1922uses the new group format; otherwise,
1923.Xr diff
1924uses the changed group format.
1925.Pp
1926The old, new, and unchanged line formats specify the output format of lines
1927from the first file, lines from the second file, and lines common to both
1928files, respectively.
1929.Pp
1930The option
1931.Op --ifdef= Va name
1932is equivalent to the following sequence of options using shell syntax:
1933.Pp
1934.Bd -literal -offset indent
1935--old-group-format='#ifndef name
1936%<#endif /* ! name */
1937\&' \e
1938--new-group-format='#ifdef name
1939%>#endif /* name */
1940\&' \e
1941--unchanged-group-format='%=' \e
1942--changed-group-format='#ifndef name
1943%<#else /* name */
1944%>#endif /* name */
1945\&'
1946.Ed
1947.Pp
1948You should carefully check the
1949.Xr diff
1950output for proper nesting. For example, when using the
1951.Op -D Va name
1952or
1953.Op --ifdef= Va name
1954option, you should check that if the differing lines contain any of the C
1955preprocessor directives
1956.Li #ifdef ,
1957.Li #ifndef ,
1958.Li #else ,
1959.Li #elif ,
1960or
1961.Li #endif ,
1962they are nested properly and match. If they don't, you must make corrections
1963manually. It is a good idea to carefully check the resulting code anyway to
1964make sure that it really does what you want it to; depending on how the input
1965files were produced, the output might contain duplicate or otherwise incorrect
1966code.
1967.Pp
1968The
1969.Xr patch
1970.Op -D Va name
1971option behaves like the
1972.Xr diff
1973.Op -D Va name
1974option, except it operates on a file and a diff to produce a merged file;See Section
1975.Dq patch Options .
1976.Pp
1977.Sh  Incomplete Lines
1978When an input file ends in a non-newline character, its last line is called
1979an
1980.Em incomplete line
1981because its last character is not a newline. All other lines are called
1982.Em full lines
1983and end in a newline character. Incomplete lines do not match full lines unless
1984differences in white space are ignored (see Section
1985.Dq White Space ) .
1986.Pp
1987An incomplete line is normally distinguished on output from a full line by
1988a following line that starts with
1989.Li \e .
1990However, the RCS format (see Section
1991.Dq RCS )
1992outputs the incomplete line as-is, without any trailing newline or following
1993line. The side by side format normally represents incomplete lines as-is,
1994but in some cases uses a
1995.Li \e
1996or
1997.Li /
1998gutter marker;See Section
1999.Dq Side by Side .
2000The if-then-else line format preserves a line's incompleteness with
2001.Li %L ,
2002and discards the newline with
2003.Li %l
2004;See Section
2005.Dq Line Formats .
2006Finally, with the
2007.Xr ed
2008and forward
2009.Xr ed
2010output formats (see Section
2011.Dq Output Formats )
2012.Xr diff
2013cannot represent an incomplete line, so it pretends there was a newline and
2014reports an error.
2015.Pp
2016For example, suppose
2017.Pa F
2018and
2019.Pa G
2020are one-byte files that contain just
2021.Li f
2022and
2023.Li g ,
2024respectively. Then
2025.Li diff F G
2026outputs
2027.Pp
2028.Bd -literal -offset indent
20291c1
2030< f
2031\e No newline at end of file
2032---
2033> g
2034\e No newline at end of file
2035.Ed
2036.Pp
2037(The exact message may differ in non-English locales.)
2038.Li diff -n F G
2039outputs the following without a trailing newline:
2040.Pp
2041.Bd -literal -offset indent
2042d1 1
2043a1 1
2044g
2045.Ed
2046.Pp
2047.Li diff -e F G
2048reports two errors and outputs the following:
2049.Pp
2050.Bd -literal -offset indent
20511c
2052g
2053\&.
2054.Ed
2055.Pp
2056.Sh  Comparing Directories
2057You can use
2058.Xr diff
2059to compare some or all of the files in two directory trees. When both file
2060name arguments to
2061.Xr diff
2062are directories, it compares each file that is contained in both directories,
2063examining file names in alphabetical order as specified by the
2064.Ev LC_COLLATE
2065locale category. Normally
2066.Xr diff
2067is silent about pairs of files that contain no differences, but if you use
2068the
2069.Op -s
2070or
2071.Op --report-identical-files
2072option, it reports pairs of identical files. Normally
2073.Xr diff
2074reports subdirectories common to both directories without comparing subdirectories'
2075files, but if you use the
2076.Op -r
2077or
2078.Op --recursive
2079option, it compares every corresponding pair of files in the directory trees,
2080as many levels deep as they go.
2081.Pp
2082For file names that are in only one of the directories,
2083.Xr diff
2084normally does not show the contents of the file that exists; it reports only
2085that the file exists in that directory and not in the other. You can make
2086.Xr diff
2087act as though the file existed but was empty in the other directory, so that
2088it outputs the entire contents of the file that actually exists. (It is output
2089as either an insertion or a deletion, depending on whether it is in the first
2090or the second directory given.) To do this, use the
2091.Op -N
2092or
2093.Op --new-file
2094option.
2095.Pp
2096If the older directory contains one or more large files that are not in the
2097newer directory, you can make the patch smaller by using the
2098.Op --unidirectional-new-file
2099option instead of
2100.Op -N .
2101This option is like
2102.Op -N
2103except that it only inserts the contents of files that appear in the second
2104directory but not the first (that is, files that were added). At the top of
2105the patch, write instructions for the user applying the patch to remove the
2106files that were deleted before applying the patch.See Section
2107.Dq Making Patches ,
2108for more discussion of making patches for distribution.
2109.Pp
2110To ignore some files while comparing directories, use the
2111.Op -x Va pattern
2112or
2113.Op --exclude= Va pattern
2114option. This option ignores any files or subdirectories whose base names match
2115the shell pattern
2116.Va pattern .
2117Unlike in the shell, a period at the start of the base of a file name matches
2118a wildcard at the start of a pattern. You should enclose
2119.Va pattern
2120in quotes so that the shell does not expand it. For example, the option
2121.Op -x '*.[ao]'
2122ignores any file whose name ends with
2123.Li .a
2124or
2125.Li .o .
2126.Pp
2127This option accumulates if you specify it more than once. For example, using
2128the options
2129.Op -x 'RCS' -x '*,v'
2130ignores any file or subdirectory whose base name is
2131.Li RCS
2132or ends with
2133.Li ,v .
2134.Pp
2135If you need to give this option many times, you can instead put the patterns
2136in a file, one pattern per line, and use the
2137.Op -X Va file
2138or
2139.Op --exclude-from= Va file
2140option. Trailing white space and empty lines are ignored in the pattern file.
2141.Pp
2142If you have been comparing two directories and stopped partway through, later
2143you might want to continue where you left off. You can do this by using the
2144.Op -S Va file
2145or
2146.Op --starting-file= Va file
2147option. This compares only the file
2148.Va file
2149and all alphabetically later files in the topmost directory level.
2150.Pp
2151If two directories differ only in that file names are lower case in one directory
2152and upper case in the upper,
2153.Xr diff
2154normally reports many differences because it compares file names in a case
2155sensitive way. With the
2156.Op --ignore-file-name-case
2157option,
2158.Xr diff
2159ignores case differences in file names, so that for example the contents of
2160the file
2161.Pa Tao
2162in one directory are compared to the contents of the file
2163.Pa TAO
2164in the other. The
2165.Op --no-ignore-file-name-case
2166option cancels the effect of the
2167.Op --ignore-file-name-case
2168option, reverting to the default behavior.
2169.Pp
2170If an
2171.Op -x Va pattern
2172or
2173.Op --exclude= Va pattern
2174option, or an
2175.Op -X Va file
2176or
2177.Op --exclude-from= Va file
2178option, is specified while the
2179.Op --ignore-file-name-case
2180option is in effect, case is ignored when excluding file names matching the
2181specified patterns.
2182.Pp
2183.Sh  Making Xr diff Output Prettier
2184.Xr diff
2185provides several ways to adjust the appearance of its output. These adjustments
2186can be applied to any output format.
2187.Pp
2188.Ss  Preserving Tab Stop Alignment
2189The lines of text in some of the
2190.Xr diff
2191output formats are preceded by one or two characters that indicate whether
2192the text is inserted, deleted, or changed. The addition of those characters
2193can cause tabs to move to the next tab stop, throwing off the alignment of
2194columns in the line. GNU
2195.Xr diff
2196provides two ways to make tab-aligned columns line up correctly.
2197.Pp
2198The first way is to have
2199.Xr diff
2200convert all tabs into the correct number of spaces before outputting them;
2201select this method with the
2202.Op -t
2203or
2204.Op --expand-tabs
2205option. To use this form of output with
2206.Xr patch ,
2207you must give
2208.Xr patch
2209the
2210.Op -l
2211or
2212.Op --ignore-white-space
2213option (see Section
2214.Dq Changed White Space ,
2215for more information).
2216.Xr diff
2217normally assumes that tab stops are set every 8 print columns, but this can
2218be altered by the
2219.Op --tabsize= Va columns
2220option.
2221.Pp
2222The other method for making tabs line up correctly is to add a tab character
2223instead of a space after the indicator character at the beginning of the line.
2224This ensures that all following tab characters are in the same position relative
2225to tab stops that they were in the original files, so that the output is aligned
2226correctly. Its disadvantage is that it can make long lines too long to fit
2227on one line of the screen or the paper. It also does not work with the unified
2228output format, which does not have a space character after the change type
2229indicator character. Select this method with the
2230.Op -T
2231or
2232.Op --initial-tab
2233option.
2234.Pp
2235.Ss  Paginating Xr diff Output
2236It can be convenient to have long output page-numbered and time-stamped. The
2237.Op -l
2238or
2239.Op --paginate
2240option does this by sending the
2241.Xr diff
2242output through the
2243.Xr pr
2244program. Here is what the page header might look like for
2245.Li diff -lc lao tzu :
2246.Pp
2247.Bd -literal -offset indent
22482002-02-22 14:20                 diff -lc lao tzu                 Page 1
2249.Ed
2250.Pp
2251.Sh  Xr diff Performance Tradeoffs
2252GNU
2253.Xr diff
2254runs quite efficiently; however, in some circumstances you can cause it to
2255run faster or produce a more compact set of changes.
2256.Pp
2257One way to improve
2258.Xr diff
2259performance is to use hard or symbolic links to files instead of copies. This
2260improves performance because
2261.Xr diff
2262normally does not need to read two hard or symbolic links to the same file,
2263since their contents must be identical. For example, suppose you copy a large
2264directory hierarchy, make a few changes to the copy, and then often use
2265.Li diff -r
2266to compare the original to the copy. If the original files are read-only,
2267you can greatly improve performance by creating the copy using hard or symbolic
2268links (e.g., with GNU
2269.Li cp -lR
2270or
2271.Li cp -sR ) .
2272Before editing a file in the copy for the first time, you should break the
2273link and replace it with a regular copy.
2274.Pp
2275You can also affect the performance of GNU
2276.Xr diff
2277by giving it options that change the way it compares files. Performance has
2278more than one dimension. These options improve one aspect of performance at
2279the cost of another, or they improve performance in some cases while hurting
2280it in others.
2281.Pp
2282The way that GNU
2283.Xr diff
2284determines which lines have changed always comes up with a near-minimal set
2285of differences. Usually it is good enough for practical purposes. If the
2286.Xr diff
2287output is large, you might want
2288.Xr diff
2289to use a modified algorithm that sometimes produces a smaller set of differences.
2290The
2291.Op -d
2292or
2293.Op --minimal
2294option does this; however, it can also cause
2295.Xr diff
2296to run more slowly than usual, so it is not the default behavior.
2297.Pp
2298When the files you are comparing are large and have small groups of changes
2299scattered throughout them, you can use the
2300.Op --speed-large-files
2301option to make a different modification to the algorithm that
2302.Xr diff
2303uses. If the input files have a constant small density of changes, this option
2304speeds up the comparisons without changing the output. If not,
2305.Xr diff
2306might produce a larger set of differences; however, the output will still
2307be correct.
2308.Pp
2309Normally
2310.Xr diff
2311discards the prefix and suffix that is common to both files before it attempts
2312to find a minimal set of differences. This makes
2313.Xr diff
2314run faster, but occasionally it may produce non-minimal output. The
2315.Op --horizon-lines= Va lines
2316option prevents
2317.Xr diff
2318from discarding the last
2319.Va lines
2320lines of the prefix and the first
2321.Va lines
2322lines of the suffix. This gives
2323.Xr diff
2324further opportunities to find a minimal output.
2325.Pp
2326Suppose a run of changed lines includes a sequence of lines at one end and
2327there is an identical sequence of lines just outside the other end. The
2328.Xr diff
2329command is free to choose which identical sequence is included in the hunk.
2330In this case,
2331.Xr diff
2332normally shifts the hunk's boundaries when this merges adjacent hunks, or
2333shifts a hunk's lines towards the end of the file. Merging hunks can make
2334the output look nicer in some cases.
2335.Pp
2336.Sh  Comparing Three Files
2337Use the program
2338.Xr diff3
2339to compare three files and show any differences among them. (
2340.Xr diff3
2341can also merge files; see diff3 Merging).
2342.Pp
2343The \(lqnormal\(rq
2344.Xr diff3
2345output format shows each hunk of differences without surrounding context.
2346Hunks are labeled depending on whether they are two-way or three-way, and
2347lines are annotated by their location in the input files.
2348.Pp
2349See Section.Dq Invoking diff3 ,
2350for more information on how to run
2351.Xr diff3 .
2352.Pp
2353.Ss  A Third Sample Input File
2354Here is a third sample file that will be used in examples to illustrate the
2355output of
2356.Xr diff3
2357and how various options can change it. The first two files are the same that
2358we used for
2359.Xr diff
2360(see Section
2361.Dq Sample diff Input ) .
2362This is the third sample file, called
2363.Pa tao :
2364.Pp
2365.Bd -literal -offset indent
2366The Way that can be told of is not the eternal Way;
2367The name that can be named is not the eternal name.
2368The Nameless is the origin of Heaven and Earth;
2369The named is the mother of all things.
2370
2371Therefore let there always be non-being,
2372  so we may see their subtlety,
2373And let there always be being,
2374  so we may see their result.
2375The two are the same,
2376But after they are produced,
2377  they have different names.
2378
2379  -- The Way of Lao-Tzu, tr. Wing-tsit Chan
2380.Ed
2381.Pp
2382.Ss  An Example of Xr diff3 Normal Format
2383Here is the output of the command
2384.Li diff3 lao tzu tao
2385(see Section
2386.Dq Sample diff3 Input ,
2387for the complete contents of the files). Notice that it shows only the lines
2388that are different among the three files.
2389.Pp
2390.Bd -literal -offset indent
2391====2
23921:1,2c
23933:1,2c
2394  The Way that can be told of is not the eternal Way;
2395  The name that can be named is not the eternal name.
23962:0a
2397====1
23981:4c
2399  The Named is the mother of all things.
24002:2,3c
24013:4,5c
2402  The named is the mother of all things.
2403
2404====3
24051:8c
24062:7c
2407    so we may see their outcome.
24083:9c
2409    so we may see their result.
2410====
24111:11a
24122:11,13c
2413  They both may be called deep and profound.
2414  Deeper and more profound,
2415  The door of all subtleties!
24163:13,14c
2417
2418    -- The Way of Lao-Tzu, tr. Wing-tsit Chan
2419.Ed
2420.Pp
2421.Ss  Detailed Description of Xr diff3 Normal Format
2422Each hunk begins with a line marked
2423.Li ==== .
2424Three-way hunks have plain
2425.Li ====
2426lines, and two-way hunks have
2427.Li 1 ,
2428.Li 2 ,
2429or
2430.Li 3
2431appended to specify which of the three input files differ in that hunk. The
2432hunks contain copies of two or three sets of input lines each preceded by
2433one or two commands identifying where the lines came from.
2434.Pp
2435Normally, two spaces precede each copy of an input line to distinguish it
2436from the commands. But with the
2437.Op -T
2438or
2439.Op --initial-tab
2440option,
2441.Xr diff3
2442uses a tab instead of two spaces; this lines up tabs correctly.See Section
2443.Dq Tabs ,
2444for more information.
2445.Pp
2446Commands take the following forms:
2447.Pp
2448.Bl -tag -width Ds
2449.It  Va file: Va la
2450This hunk appears after line
2451.Va l
2452of file
2453.Va file ,
2454and contains no lines in that file. To edit this file to yield the other files,
2455one must append hunk lines taken from the other files. For example,
2456.Li 1:11a
2457means that the hunk follows line 11 in the first file and contains no lines
2458from that file.
2459.Pp
2460.It  Va file: Va rc
2461This hunk contains the lines in the range
2462.Va r
2463of file
2464.Va file .
2465The range
2466.Va r
2467is a comma-separated pair of line numbers, or just one number if the range
2468is a singleton. To edit this file to yield the other files, one must change
2469the specified lines to be the lines taken from the other files. For example,
2470.Li 2:11,13c
2471means that the hunk contains lines 11 through 13 from the second file.
2472.El
2473.Pp
2474If the last line in a set of input lines is incomplete (see Section
2475.Dq Incomplete Lines ) ,
2476it is distinguished on output from a full line by a following line that starts
2477with
2478.Li \e .
2479.Pp
2480.Ss  Xr diff3 Hunks
2481Groups of lines that differ in two or three of the input files are called
2482.Em diff3 hunks ,
2483by analogy with
2484.Xr diff
2485hunks (see Section
2486.Dq Hunks ) .
2487If all three input files differ in a
2488.Xr diff3
2489hunk, the hunk is called a
2490.Em three-way hunk
2491; if just two input files differ, it is a
2492.Em two-way hunk .
2493.Pp
2494As with
2495.Xr diff ,
2496several solutions are possible. When comparing the files
2497.Li A ,
2498.Li B ,
2499and
2500.Li C ,
2501.Xr diff3
2502normally finds
2503.Xr diff3
2504hunks by merging the two-way hunks output by the two commands
2505.Li diff A B
2506and
2507.Li diff A C .
2508This does not necessarily minimize the size of the output, but exceptions
2509should be rare.
2510.Pp
2511For example, suppose
2512.Pa F
2513contains the three lines
2514.Li a ,
2515.Li b ,
2516.Li f ,
2517.Pa G
2518contains the lines
2519.Li g ,
2520.Li b ,
2521.Li g ,
2522and
2523.Pa H
2524contains the lines
2525.Li a ,
2526.Li b ,
2527.Li h .
2528.Li diff3 F G H
2529might output the following:
2530.Pp
2531.Bd -literal -offset indent
2532====2
25331:1c
25343:1c
2535  a
25362:1c
2537  g
2538====
25391:3c
2540  f
25412:3c
2542  g
25433:3c
2544  h
2545.Ed
2546.Pp
2547because it found a two-way hunk containing
2548.Li a
2549in the first and third files and
2550.Li g
2551in the second file, then the single line
2552.Li b
2553common to all three files, then a three-way hunk containing the last line
2554of each file.
2555.Pp
2556.Sh  Merging From a Common Ancestor
2557When two people have made changes to copies of the same file,
2558.Xr diff3
2559can produce a merged output that contains both sets of changes together with
2560warnings about conflicts.
2561.Pp
2562One might imagine programs with names like
2563.Xr diff4
2564and
2565.Xr diff5
2566to compare more than three files simultaneously, but in practice the need
2567rarely arises. You can use
2568.Xr diff3
2569to merge three or more sets of changes to a file by merging two change sets
2570at a time.
2571.Pp
2572.Xr diff3
2573can incorporate changes from two modified versions into a common preceding
2574version. This lets you merge the sets of changes represented by the two newer
2575files. Specify the common ancestor version as the second argument and the
2576two newer versions as the first and third arguments, like this:
2577.Pp
2578.Bd -literal -offset indent
2579diff3 mine older yours
2580.Ed
2581.Pp
2582You can remember the order of the arguments by noting that they are in alphabetical
2583order.
2584.Pp
2585You can think of this as subtracting
2586.Va older
2587from
2588.Va yours
2589and adding the result to
2590.Va mine ,
2591or as merging into
2592.Va mine
2593the changes that would turn
2594.Va older
2595into
2596.Va yours .
2597This merging is well-defined as long as
2598.Va mine
2599and
2600.Va older
2601match in the neighborhood of each such change. This fails to be true when
2602all three input files differ or when only
2603.Va older
2604differs; we call this a
2605.Em conflict .
2606When all three input files differ, we call the conflict an
2607.Em overlap .
2608.Pp
2609.Xr diff3
2610gives you several ways to handle overlaps and conflicts. You can omit overlaps
2611or conflicts, or select only overlaps, or mark conflicts with special
2612.Li <<<<<<<
2613and
2614.Li >>>>>>>
2615lines.
2616.Pp
2617.Xr diff3
2618can output the merge results as an
2619.Xr ed
2620script that that can be applied to the first file to yield the merged output.
2621However, it is usually better to have
2622.Xr diff3
2623generate the merged output directly; this bypasses some problems with
2624.Xr ed .
2625.Pp
2626.Ss  Selecting Which Changes to Incorporate
2627You can select all unmerged changes from
2628.Va older
2629to
2630.Va yours
2631for merging into
2632.Va mine
2633with the
2634.Op -e
2635or
2636.Op --ed
2637option. You can select only the nonoverlapping unmerged changes with
2638.Op -3
2639or
2640.Op --easy-only ,
2641and you can select only the overlapping changes with
2642.Op -x
2643or
2644.Op --overlap-only .
2645.Pp
2646The
2647.Op -e ,
2648.Op -3
2649and
2650.Op -x
2651options select only
2652.Em unmerged changes ,
2653i.e. changes where
2654.Va mine
2655and
2656.Va yours
2657differ; they ignore changes from
2658.Va older
2659to
2660.Va yours
2661where
2662.Va mine
2663and
2664.Va yours
2665are identical, because they assume that such changes have already been merged.
2666If this assumption is not a safe one, you can use the
2667.Op -A
2668or
2669.Op --show-all
2670option (see Section
2671.Dq Marking Conflicts ) .
2672.Pp
2673Here is the output of the command
2674.Xr diff3
2675with each of these three options (see Section
2676.Dq Sample diff3 Input ,
2677for the complete contents of the files). Notice that
2678.Op -e
2679outputs the union of the disjoint sets of changes output by
2680.Op -3
2681and
2682.Op -x .
2683.Pp
2684Output of
2685.Li diff3 -e lao tzu tao :
2686.Bd -literal -offset indent
268711a
2688
2689  -- The Way of Lao-Tzu, tr. Wing-tsit Chan
2690\&.
26918c
2692  so we may see their result.
2693\&.
2694.Ed
2695.Pp
2696Output of
2697.Li diff3 -3 lao tzu tao :
2698.Bd -literal -offset indent
26998c
2700  so we may see their result.
2701\&.
2702.Ed
2703.Pp
2704Output of
2705.Li diff3 -x lao tzu tao :
2706.Bd -literal -offset indent
270711a
2708
2709  -- The Way of Lao-Tzu, tr. Wing-tsit Chan
2710\&.
2711.Ed
2712.Pp
2713.Ss  Marking Conflicts
2714.Xr diff3
2715can mark conflicts in the merged output by bracketing them with special marker
2716lines. A conflict that comes from two files
2717.Va A
2718and
2719.Va B
2720is marked as follows:
2721.Pp
2722.Bd -literal -offset indent
2723<<<<<<< A
2724lines from A
2725=======
2726lines from B
2727>>>>>>> B
2728.Ed
2729.Pp
2730A conflict that comes from three files
2731.Va A ,
2732.Va B
2733and
2734.Va C
2735is marked as follows:
2736.Pp
2737.Bd -literal -offset indent
2738<<<<<<< A
2739lines from A
2740||||||| B
2741lines from B
2742=======
2743lines from C
2744>>>>>>> C
2745.Ed
2746.Pp
2747The
2748.Op -A
2749or
2750.Op --show-all
2751option acts like the
2752.Op -e
2753option, except that it brackets conflicts, and it outputs all changes from
2754.Va older
2755to
2756.Va yours ,
2757not just the unmerged changes. Thus, given the sample input files (see Section
2758.Dq Sample diff3 Input ) ,
2759.Li diff3 -A lao tzu tao
2760puts brackets around the conflict where only
2761.Pa tzu
2762differs:
2763.Pp
2764.Bd -literal -offset indent
2765<<<<<<< tzu
2766=======
2767The Way that can be told of is not the eternal Way;
2768The name that can be named is not the eternal name.
2769>>>>>>> tao
2770.Ed
2771.Pp
2772And it outputs the three-way conflict as follows:
2773.Pp
2774.Bd -literal -offset indent
2775<<<<<<< lao
2776||||||| tzu
2777They both may be called deep and profound.
2778Deeper and more profound,
2779The door of all subtleties!
2780=======
2781
2782  -- The Way of Lao-Tzu, tr. Wing-tsit Chan
2783>>>>>>> tao
2784.Ed
2785.Pp
2786The
2787.Op -E
2788or
2789.Op --show-overlap
2790option outputs less information than the
2791.Op -A
2792or
2793.Op --show-all
2794option, because it outputs only unmerged changes, and it never outputs the
2795contents of the second file. Thus the
2796.Op -E
2797option acts like the
2798.Op -e
2799option, except that it brackets the first and third files from three-way overlapping
2800changes. Similarly,
2801.Op -X
2802acts like
2803.Op -x ,
2804except it brackets all its (necessarily overlapping) changes. For example,
2805for the three-way overlapping change above, the
2806.Op -E
2807and
2808.Op -X
2809options output the following:
2810.Pp
2811.Bd -literal -offset indent
2812<<<<<<< lao
2813=======
2814
2815  -- The Way of Lao-Tzu, tr. Wing-tsit Chan
2816>>>>>>> tao
2817.Ed
2818.Pp
2819If you are comparing files that have meaningless or uninformative names, you
2820can use the
2821.Op --label= Va label
2822option to show alternate names in the
2823.Li <<<<<<< ,
2824.Li |||||||
2825and
2826.Li >>>>>>>
2827brackets. This option can be given up to three times, once for each input
2828file. Thus
2829.Li diff3 -A --label X --label Y --label Z A B C
2830acts like
2831.Li diff3 -A A B C ,
2832except that the output looks like it came from files named
2833.Li X ,
2834.Li Y
2835and
2836.Li Z
2837rather than from files named
2838.Li A ,
2839.Li B
2840and
2841.Li C .
2842.Pp
2843.Ss  Generating the Merged Output Directly
2844With the
2845.Op -m
2846or
2847.Op --merge
2848option,
2849.Xr diff3
2850outputs the merged file directly. This is more efficient than using
2851.Xr ed
2852to generate it, and works even with non-text files that
2853.Xr ed
2854would reject. If you specify
2855.Op -m
2856without an
2857.Xr ed
2858script option,
2859.Op -A
2860is assumed.
2861.Pp
2862For example, the command
2863.Li diff3 -m lao tzu tao
2864(see Section
2865.Dq Sample diff3 Input
2866for a copy of the input files) would output the following:
2867.Pp
2868.Bd -literal -offset indent
2869<<<<<<< tzu
2870=======
2871The Way that can be told of is not the eternal Way;
2872The name that can be named is not the eternal name.
2873>>>>>>> tao
2874The Nameless is the origin of Heaven and Earth;
2875The Named is the mother of all things.
2876Therefore let there always be non-being,
2877  so we may see their subtlety,
2878And let there always be being,
2879  so we may see their result.
2880The two are the same,
2881But after they are produced,
2882  they have different names.
2883<<<<<<< lao
2884||||||| tzu
2885They both may be called deep and profound.
2886Deeper and more profound,
2887The door of all subtleties!
2888=======
2889
2890  -- The Way of Lao-Tzu, tr. Wing-tsit Chan
2891>>>>>>> tao
2892.Ed
2893.Pp
2894.Ss  How Xr diff3 Merges Incomplete Lines
2895With
2896.Op -m ,
2897incomplete lines (see Section
2898.Dq Incomplete Lines )
2899are simply copied to the output as they are found; if the merged output ends
2900in an conflict and one of the input files ends in an incomplete line, succeeding
2901.Li ||||||| ,
2902.Li =======
2903or
2904.Li >>>>>>>
2905brackets appear somewhere other than the start of a line because they are
2906appended to the incomplete line.
2907.Pp
2908Without
2909.Op -m ,
2910if an
2911.Xr ed
2912script option is specified and an incomplete line is found,
2913.Xr diff3
2914generates a warning and acts as if a newline had been present.
2915.Pp
2916.Ss  Saving the Changed File
2917Traditional Unix
2918.Xr diff3
2919generates an
2920.Xr ed
2921script without the trailing
2922.Li w
2923and
2924.Li q
2925commands that save the changes. System V
2926.Xr diff3
2927generates these extra commands. GNU
2928.Xr diff3
2929normally behaves like traditional Unix
2930.Xr diff3 ,
2931but with the
2932.Op -i
2933option it behaves like System V
2934.Xr diff3
2935and appends the
2936.Li w
2937and
2938.Li q
2939commands.
2940.Pp
2941The
2942.Op -i
2943option requires one of the
2944.Xr ed
2945script options
2946.Op -AeExX3 ,
2947and is incompatible with the merged output option
2948.Op -m .
2949.Pp
2950.Sh  Interactive Merging with Xr sdiff
2951With
2952.Xr sdiff ,
2953you can merge two files interactively based on a side-by-side
2954.Op -y
2955format comparison (see Section
2956.Dq Side by Side ) .
2957Use
2958.Op -o Va file
2959or
2960.Op --output= Va file
2961to specify where to put the merged text.See Section
2962.Dq Invoking sdiff ,
2963for more details on the options to
2964.Xr sdiff .
2965.Pp
2966Another way to merge files interactively is to use the Emacs Lisp package
2967.Xr emerge .
2968See Section.Dq emerge ,
2969for more information.
2970.Pp
2971.Ss  Specifying Xr diff Options to Xr sdiff
2972The following
2973.Xr sdiff
2974options have the same meaning as for
2975.Xr diff .
2976See Section.Dq diff Options ,
2977for the use of these options.
2978.Pp
2979.Bd -literal -offset indent
2980-a -b -d -i -t -v
2981-B -E -I regexp
2982
2983--expand-tabs
2984--ignore-blank-lines  --ignore-case
2985--ignore-matching-lines=regexp  --ignore-space-change
2986--ignore-tab-expansion
2987--left-column  --minimal  --speed-large-files
2988--strip-trailing-cr  --suppress-common-lines
2989--tabsize=columns  --text  --version  --width=columns
2990.Ed
2991.Pp
2992For historical reasons,
2993.Xr sdiff
2994has alternate names for some options. The
2995.Op -l
2996option is equivalent to the
2997.Op --left-column
2998option, and similarly
2999.Op -s
3000is equivalent to
3001.Op --suppress-common-lines .
3002The meaning of the
3003.Xr sdiff
3004.Op -w
3005and
3006.Op -W
3007options is interchanged from that of
3008.Xr diff :
3009with
3010.Xr sdiff ,
3011.Op -w Va columns
3012is equivalent to
3013.Op --width= Va columns ,
3014and
3015.Op -W
3016is equivalent to
3017.Op --ignore-all-space .
3018.Xr sdiff
3019without the
3020.Op -o
3021option is equivalent to
3022.Xr diff
3023with the
3024.Op -y
3025or
3026.Op --side-by-side
3027option (see Section
3028.Dq Side by Side ) .
3029.Pp
3030.Ss  Merge Commands
3031Groups of common lines, with a blank gutter, are copied from the first file
3032to the output. After each group of differing lines,
3033.Xr sdiff
3034prompts with
3035.Li %
3036and pauses, waiting for one of the following commands. Follow each command
3037with RET.
3038.Pp
3039.Bl -tag -width Ds
3040.It  e
3041Discard both versions. Invoke a text editor on an empty temporary file, then
3042copy the resulting file to the output.
3043.Pp
3044.It  eb
3045Concatenate the two versions, edit the result in a temporary file, then copy
3046the edited result to the output.
3047.Pp
3048.It  ed
3049Like
3050.Li eb ,
3051except precede each version with a header that shows what file and lines the
3052version came from.
3053.Pp
3054.It  el
3055.It  e1
3056Edit a copy of the left version, then copy the result to the output.
3057.Pp
3058.It  er
3059.It  e2
3060Edit a copy of the right version, then copy the result to the output.
3061.Pp
3062.It  l
3063.It  1
3064Copy the left version to the output.
3065.Pp
3066.It  q
3067Quit.
3068.Pp
3069.It  r
3070.It  2
3071Copy the right version to the output.
3072.Pp
3073.It  s
3074Silently copy common lines.
3075.Pp
3076.It  v
3077Verbosely copy common lines. This is the default.
3078.El
3079.Pp
3080The text editor invoked is specified by the
3081.Ev EDITOR
3082environment variable if it is set. The default is system-dependent.
3083.Pp
3084.Sh  Merging with Xr patch
3085.Xr patch
3086takes comparison output produced by
3087.Xr diff
3088and applies the differences to a copy of the original file, producing a patched
3089version. With
3090.Xr patch ,
3091you can distribute just the changes to a set of files instead of distributing
3092the entire file set; your correspondents can apply
3093.Xr patch
3094to update their copy of the files with your changes.
3095.Xr patch
3096automatically determines the diff format, skips any leading or trailing headers,
3097and uses the headers to determine which file to patch. This lets your correspondents
3098feed a mail message containing a difference listing directly to
3099.Xr patch .
3100.Pp
3101.Xr patch
3102detects and warns about common problems like forward patches. It saves any
3103patches that it could not apply. It can also maintain a
3104.Li patchlevel.h
3105file to ensure that your correspondents apply diffs in the proper order.
3106.Pp
3107.Xr patch
3108accepts a series of diffs in its standard input, usually separated by headers
3109that specify which file to patch. It applies
3110.Xr diff
3111hunks (see Section
3112.Dq Hunks )
3113one by one. If a hunk does not exactly match the original file,
3114.Xr patch
3115uses heuristics to try to patch the file as well as it can. If no approximate
3116match can be found,
3117.Xr patch
3118rejects the hunk and skips to the next hunk.
3119.Xr patch
3120normally replaces each file
3121.Va f
3122with its new version, putting reject hunks (if any) into
3123.Li  Va f.rej .
3124.Pp
3125See Section.Dq Invoking patch ,
3126for detailed information on the options to
3127.Xr patch .
3128.Pp
3129.Ss  Selecting the Xr patch Input Format
3130.Xr patch
3131normally determines which
3132.Xr diff
3133format the patch file uses by examining its contents. For patch files that
3134contain particularly confusing leading text, you might need to use one of
3135the following options to force
3136.Xr patch
3137to interpret the patch file as a certain format of diff. The output formats
3138listed here are the only ones that
3139.Xr patch
3140can understand.
3141.Pp
3142.Bl -tag -width Ds
3143.It  -c
3144.It  --context
3145context diff.
3146.Pp
3147.It  -e
3148.It  --ed
3149.Xr ed
3150script.
3151.Pp
3152.It  -n
3153.It  --normal
3154normal diff.
3155.Pp
3156.It  -u
3157.It  --unified
3158unified diff.
3159.El
3160.Pp
3161.Ss  Revision Control
3162If a nonexistent input file is under a revision control system supported by
3163.Xr patch ,
3164.Xr patch
3165normally asks the user whether to get (or check out) the file from the revision
3166control system. Patch currently supports RCS, ClearCase and SCCS. Under RCS
3167and SCCS,
3168.Xr patch
3169also asks when the input file is read-only and matches the default version
3170in the revision control system.
3171.Pp
3172The
3173.Op -g Va num
3174or
3175.Op --get= Va num
3176option affects access to files under supported revision control systems. If
3177.Va num
3178is positive,
3179.Xr patch
3180gets the file without asking the user; if zero,
3181.Xr patch
3182neither asks the user nor gets the file; and if negative,
3183.Xr patch
3184asks the user before getting the file. The default value of
3185.Va num
3186is given by the value of the
3187.Ev PATCH_GET
3188environment variable if it is set; if not, the default value is zero if
3189.Xr patch
3190is conforming to POSIX, negative otherwise.See Section
3191.Dq patch and POSIX .
3192.Pp
3193The choice of revision control system is unaffected by the
3194.Ev VERSION_CONTROL
3195environment variable (see Section
3196.Dq Backup Names ) .
3197.Pp
3198.Ss  Applying Imperfect Patches
3199.Xr patch
3200tries to skip any leading text in the patch file, apply the diff, and then
3201skip any trailing text. Thus you can feed a mail message directly to
3202.Xr patch ,
3203and it should work. If the entire diff is indented by a constant amount of
3204white space,
3205.Xr patch
3206automatically ignores the indentation. If a context diff contains trailing
3207carriage return on each line,
3208.Xr patch
3209automatically ignores the carriage return. If a context diff has been encapsulated
3210by prepending
3211.Li -
3212to lines beginning with
3213.Li -
3214as per
3215.Lk ftp://ftp.isi.edu/in-notes/rfc934.txt ,
3216.Xr patch
3217automatically unencapsulates the input.
3218.Pp
3219However, certain other types of imperfect input require user intervention
3220or testing.
3221.Pp
3222.Em  Applying Patches with Changed White Space
3223.Pp
3224Sometimes mailers, editors, or other programs change spaces into tabs, or
3225vice versa. If this happens to a patch file or an input file, the files might
3226look the same, but
3227.Xr patch
3228will not be able to match them properly. If this problem occurs, use the
3229.Op -l
3230or
3231.Op --ignore-white-space
3232option, which makes
3233.Xr patch
3234compare blank characters (i.e. spaces and tabs) loosely so that any nonempty
3235sequence of blanks in the patch file matches any nonempty sequence of blanks
3236in the input files. Non-blank characters must still match exactly. Each line
3237of the context must still match a line in the input file.
3238.Pp
3239.Em  Applying Reversed Patches
3240.Pp
3241Sometimes people run
3242.Xr diff
3243with the new file first instead of second. This creates a diff that is \(lqreversed\(rq.
3244To apply such patches, give
3245.Xr patch
3246the
3247.Op -R
3248or
3249.Op --reverse
3250option.
3251.Xr patch
3252then attempts to swap each hunk around before applying it. Rejects come out
3253in the swapped format.
3254.Pp
3255Often
3256.Xr patch
3257can guess that the patch is reversed. If the first hunk of a patch fails,
3258.Xr patch
3259reverses the hunk to see if it can apply it that way. If it can,
3260.Xr patch
3261asks you if you want to have the
3262.Op -R
3263option set; if it can't,
3264.Xr patch
3265continues to apply the patch normally. This method cannot detect a reversed
3266patch if it is a normal diff and the first command is an append (which should
3267have been a delete) since appends always succeed, because a null context matches
3268anywhere. But most patches add or change lines rather than delete them, so
3269most reversed normal diffs begin with a delete, which fails, and
3270.Xr patch
3271notices.
3272.Pp
3273If you apply a patch that you have already applied,
3274.Xr patch
3275thinks it is a reversed patch and offers to un-apply the patch. This could
3276be construed as a feature. If you did this inadvertently and you don't want
3277to un-apply the patch, just answer
3278.Li n
3279to this offer and to the subsequent \(lqapply anyway\(rq question---or type
3280.Li C-c
3281to kill the
3282.Xr patch
3283process.
3284.Pp
3285.Em  Helping Xr patch Find Inexact Matches
3286.Pp
3287For context diffs, and to a lesser extent normal diffs,
3288.Xr patch
3289can detect when the line numbers mentioned in the patch are incorrect, and
3290it attempts to find the correct place to apply each hunk of the patch. As
3291a first guess, it takes the line number mentioned in the hunk, plus or minus
3292any offset used in applying the previous hunk. If that is not the correct
3293place,
3294.Xr patch
3295scans both forward and backward for a set of lines matching the context given
3296in the hunk.
3297.Pp
3298First
3299.Xr patch
3300looks for a place where all lines of the context match. If it cannot find
3301such a place, and it is reading a context or unified diff, and the maximum
3302fuzz factor is set to 1 or more, then
3303.Xr patch
3304makes another scan, ignoring the first and last line of context. If that fails,
3305and the maximum fuzz factor is set to 2 or more, it makes another scan, ignoring
3306the first two and last two lines of context are ignored. It continues similarly
3307if the maximum fuzz factor is larger.
3308.Pp
3309The
3310.Op -F Va lines
3311or
3312.Op --fuzz= Va lines
3313option sets the maximum fuzz factor to
3314.Va lines .
3315This option only applies to context and unified diffs; it ignores up to
3316.Va lines
3317lines while looking for the place to install a hunk. Note that a larger fuzz
3318factor increases the odds of making a faulty patch. The default fuzz factor
3319is 2; there is no point to setting it to more than the number of lines of
3320context in the diff, ordinarily 3.
3321.Pp
3322If
3323.Xr patch
3324cannot find a place to install a hunk of the patch, it writes the hunk out
3325to a reject file (see Section
3326.Dq Reject Names ,
3327for information on how reject files are named). It writes out rejected hunks
3328in context format no matter what form the input patch is in. If the input
3329is a normal or
3330.Xr ed
3331diff, many of the contexts are simply null. The line numbers on the hunks
3332in the reject file may be different from those in the patch file: they show
3333the approximate location where
3334.Xr patch
3335thinks the failed hunks belong in the new file rather than in the old one.
3336.Pp
3337If the
3338.Op --verbose
3339option is given, then as it completes each hunk
3340.Xr patch
3341tells you whether the hunk succeeded or failed, and if it failed, on which
3342line (in the new file)
3343.Xr patch
3344thinks the hunk should go. If this is different from the line number specified
3345in the diff, it tells you the offset. A single large offset
3346.Em may
3347indicate that
3348.Xr patch
3349installed a hunk in the wrong place.
3350.Xr patch
3351also tells you if it used a fuzz factor to make the match, in which case you
3352should also be slightly suspicious.
3353.Pp
3354.Xr patch
3355cannot tell if the line numbers are off in an
3356.Xr ed
3357script, and can only detect wrong line numbers in a normal diff when it finds
3358a change or delete command. It may have the same problem with a context diff
3359using a fuzz factor equal to or greater than the number of lines of context
3360shown in the diff (typically 3). In these cases, you should probably look
3361at a context diff between your original and patched input files to see if
3362the changes make sense. Compiling without errors is a pretty good indication
3363that the patch worked, but not a guarantee.
3364.Pp
3365A patch against an empty file applies to a nonexistent file, and vice versa.See Section
3366.Dq Creating and Removing .
3367.Pp
3368.Xr patch
3369usually produces the correct results, even when it must make many guesses.
3370However, the results are guaranteed only when the patch is applied to an exact
3371copy of the file that the patch was generated from.
3372.Pp
3373.Em  Predicting what Xr patch will do
3374.Pp
3375It may not be obvious in advance what
3376.Xr patch
3377will do with a complicated or poorly formatted patch. If you are concerned
3378that the input might cause
3379.Xr patch
3380to modify the wrong files, you can use the
3381.Op --dry-run
3382option, which causes
3383.Xr patch
3384to print the results of applying patches without actually changing any files.
3385You can then inspect the diagnostics generated by the dry run to see whether
3386.Xr patch
3387will modify the files that you expect. If the patch does not do what you want,
3388you can modify the patch (or the other options to
3389.Xr patch )
3390and try another dry run. Once you are satisfied with the proposed patch you
3391can apply it by invoking
3392.Xr patch
3393as before, but this time without the
3394.Op --dry-run
3395option.
3396.Pp
3397.Ss  Creating and Removing Files
3398Sometimes when comparing two directories, a file may exist in one directory
3399but not the other. If you give
3400.Xr diff
3401the
3402.Op -N
3403or
3404.Op --new-file
3405option, or if you supply an old or new file that is named
3406.Pa /dev/null
3407or is empty and is dated the Epoch (1970-01-01 00:00:00 UTC),
3408.Xr diff
3409outputs a patch that adds or deletes the contents of this file. When given
3410such a patch,
3411.Xr patch
3412normally creates a new file or removes the old file. However, when conforming
3413to POSIX (see Section
3414.Dq patch and POSIX ) ,
3415.Xr patch
3416does not remove the old file, but leaves it empty. The
3417.Op -E
3418or
3419.Op --remove-empty-files
3420option causes
3421.Xr patch
3422to remove output files that are empty after applying a patch, even if the
3423patch does not appear to be one that removed the file.
3424.Pp
3425If the patch appears to create a file that already exists,
3426.Xr patch
3427asks for confirmation before applying the patch.
3428.Pp
3429.Ss  Updating Time Stamps on Patched Files
3430When
3431.Xr patch
3432updates a file, it normally sets the file's last-modified time stamp to the
3433current time of day. If you are using
3434.Xr patch
3435to track a software distribution, this can cause
3436.Xr make
3437to incorrectly conclude that a patched file is out of date. For example, if
3438.Pa syntax.c
3439depends on
3440.Pa syntax.y ,
3441and
3442.Xr patch
3443updates
3444.Pa syntax.c
3445and then
3446.Pa syntax.y ,
3447then
3448.Pa syntax.c
3449will normally appear to be out of date with respect to
3450.Pa syntax.y
3451even though its contents are actually up to date.
3452.Pp
3453The
3454.Op -Z
3455or
3456.Op --set-utc
3457option causes
3458.Xr patch
3459to set a patched file's modification and access times to the time stamps given
3460in context diff headers. If the context diff headers do not specify a time
3461zone, they are assumed to use Coordinated Universal Time (UTC, often known
3462as GMT).
3463.Pp
3464The
3465.Op -T
3466or
3467.Op --set-time
3468option acts like
3469.Op -Z
3470or
3471.Op --set-utc ,
3472except that it assumes that the context diff headers' time stamps use local
3473time instead of UTC. This option is not recommended, because patches using
3474local time cannot easily be used by people in other time zones, and because
3475local time stamps are ambiguous when local clocks move backwards during daylight-saving
3476time adjustments. If the context diff headers specify a time zone, this option
3477is equivalent to
3478.Op -Z
3479or
3480.Op --set-utc .
3481.Pp
3482.Xr patch
3483normally refrains from setting a file's time stamps if the file's original
3484last-modified time stamp does not match the time given in the diff header,
3485of if the file's contents do not exactly match the patch. However, if the
3486.Op -f
3487or
3488.Op --force
3489option is given, the file's time stamps are set regardless.
3490.Pp
3491Due to the limitations of the current
3492.Xr diff
3493format,
3494.Xr patch
3495cannot update the times of files whose contents have not changed. Also, if
3496you set file time stamps to values other than the current time of day, you
3497should also remove (e.g., with
3498.Li make clean )
3499all files that depend on the patched files, so that later invocations of
3500.Xr make
3501do not get confused by the patched files' times.
3502.Pp
3503.Ss  Multiple Patches in a File
3504If the patch file contains more than one patch, and if you do not specify
3505an input file on the command line,
3506.Xr patch
3507tries to apply each patch as if they came from separate patch files. This
3508means that it determines the name of the file to patch for each patch, and
3509that it examines the leading text before each patch for file names and prerequisite
3510revision level (see Section
3511.Dq Making Patches ,
3512for more on that topic).
3513.Pp
3514.Xr patch
3515uses the following rules to intuit a file name from the leading text before
3516a patch. First,
3517.Xr patch
3518takes an ordered list of candidate file names as follows:
3519.Pp
3520.Bl -bullet
3521.It
3522If the header is that of a context diff,
3523.Xr patch
3524takes the old and new file names in the header. A name is ignored if it does
3525not have enough slashes to satisfy the
3526.Op -p Va num
3527or
3528.Op --strip= Va num
3529option. The name
3530.Pa /dev/null
3531is also ignored.
3532.Pp
3533.It
3534If there is an
3535.Li Index:
3536line in the leading garbage and if either the old and new names are both absent
3537or if
3538.Xr patch
3539is conforming to POSIX,
3540.Xr patch
3541takes the name in the
3542.Li Index:
3543line.
3544.Pp
3545.It
3546For the purpose of the following rules, the candidate file names are considered
3547to be in the order (old, new, index), regardless of the order that they appear
3548in the header.
3549.El
3550.Pp
3551Then
3552.Xr patch
3553selects a file name from the candidate list as follows:
3554.Pp
3555.Bl -bullet
3556.It
3557If some of the named files exist,
3558.Xr patch
3559selects the first name if conforming to POSIX, and the best name otherwise.
3560.Pp
3561.It
3562If
3563.Xr patch
3564is not ignoring RCS, ClearCase, and SCCS (see Section
3565.Dq Revision Control ) ,
3566and no named files exist but an RCS, ClearCase, or SCCS master is found,
3567.Xr patch
3568selects the first named file with an RCS, ClearCase, or SCCS master.
3569.Pp
3570.It
3571If no named files exist, no RCS, ClearCase, or SCCS master was found, some
3572names are given,
3573.Xr patch
3574is not conforming to POSIX, and the patch appears to create a file,
3575.Xr patch
3576selects the best name requiring the creation of the fewest directories.
3577.Pp
3578.It
3579If no file name results from the above heuristics, you are asked for the name
3580of the file to patch, and
3581.Xr patch
3582selects that name.
3583.El
3584.Pp
3585To determine the
3586.Em best
3587of a nonempty list of file names,
3588.Xr patch
3589first takes all the names with the fewest path name components; of those,
3590it then takes all the names with the shortest basename; of those, it then
3591takes all the shortest names; finally, it takes the first remaining name.
3592.Pp
3593See Section.Dq patch and POSIX ,
3594to see whether
3595.Xr patch
3596is conforming to POSIX.
3597.Pp
3598.Ss  Applying Patches in Other Directories
3599The
3600.Op -d Va directory
3601or
3602.Op --directory= Va directory
3603option to
3604.Xr patch
3605makes directory
3606.Va directory
3607the current directory for interpreting both file names in the patch file,
3608and file names given as arguments to other options (such as
3609.Op -B
3610and
3611.Op -o ) .
3612For example, while in a mail reading program, you can patch a file in the
3613.Pa /usr/src/emacs
3614directory directly from a message containing the patch like this:
3615.Pp
3616.Bd -literal -offset indent
3617| patch -d /usr/src/emacs
3618.Ed
3619.Pp
3620Sometimes the file names given in a patch contain leading directories, but
3621you keep your files in a directory different from the one given in the patch.
3622In those cases, you can use the
3623.Op -p Va number
3624or
3625.Op --strip= Va number
3626option to set the file name strip count to
3627.Va number .
3628The strip count tells
3629.Xr patch
3630how many slashes, along with the directory names between them, to strip from
3631the front of file names. A sequence of one or more adjacent slashes is counted
3632as a single slash. By default,
3633.Xr patch
3634strips off all leading directories, leaving just the base file names.
3635.Pp
3636For example, suppose the file name in the patch file is
3637.Pa /gnu/src/emacs/etc/NEWS .
3638Using
3639.Op -p0
3640gives the entire file name unmodified,
3641.Op -p1
3642gives
3643.Pa gnu/src/emacs/etc/NEWS
3644(no leading slash),
3645.Op -p4
3646gives
3647.Pa etc/NEWS ,
3648and not specifying
3649.Op -p
3650at all gives
3651.Pa NEWS .
3652.Pp
3653.Xr patch
3654looks for each file (after any slashes have been stripped) in the current
3655directory, or if you used the
3656.Op -d Va directory
3657option, in that directory.
3658.Pp
3659.Ss  Backup Files
3660Normally,
3661.Xr patch
3662creates a backup file if the patch does not exactly match the original input
3663file, because in that case the original data might not be recovered if you
3664undo the patch with
3665.Li patch -R
3666(see Section
3667.Dq Reversed Patches ) .
3668However, when conforming to POSIX,
3669.Xr patch
3670does not create backup files by default.See Section
3671.Dq patch and POSIX .
3672.Pp
3673The
3674.Op -b
3675or
3676.Op --backup
3677option causes
3678.Xr patch
3679to make a backup file regardless of whether the patch matches the original
3680input. The
3681.Op --backup-if-mismatch
3682option causes
3683.Xr patch
3684to create backup files for mismatches files; this is the default when not
3685conforming to POSIX. The
3686.Op --no-backup-if-mismatch
3687option causes
3688.Xr patch
3689to not create backup files, even for mismatched patches; this is the default
3690when conforming to POSIX.
3691.Pp
3692When backing up a file that does not exist, an empty, unreadable backup file
3693is created as a placeholder to represent the nonexistent file.
3694.Pp
3695.Ss  Backup File Names
3696Normally,
3697.Xr patch
3698renames an original input file into a backup file by appending to its name
3699the extension
3700.Li .orig ,
3701or
3702.Li ~
3703if using
3704.Li .orig
3705would make the backup file name too long. The
3706.Op -z Va backup-suffix
3707or
3708.Op --suffix= Va backup-suffix
3709option causes
3710.Xr patch
3711to use
3712.Va backup-suffix
3713as the backup extension instead.
3714.Pp
3715Alternately, you can specify the extension for backup files with the
3716.Ev SIMPLE_BACKUP_SUFFIX
3717environment variable, which the options override.
3718.Pp
3719.Xr patch
3720can also create numbered backup files the way GNU Emacs does. With this method,
3721instead of having a single backup of each file,
3722.Xr patch
3723makes a new backup file name each time it patches a file. For example, the
3724backups of a file named
3725.Pa sink
3726would be called, successively,
3727.Pa sink.~1~ ,
3728.Pa sink.~2~ ,
3729.Pa sink.~3~ ,
3730etc.
3731.Pp
3732The
3733.Op -V Va backup-style
3734or
3735.Op --version-control= Va backup-style
3736option takes as an argument a method for creating backup file names. You can
3737alternately control the type of backups that
3738.Xr patch
3739makes with the
3740.Ev PATCH_VERSION_CONTROL
3741environment variable, which the
3742.Op -V
3743option overrides. If
3744.Ev PATCH_VERSION_CONTROL
3745is not set, the
3746.Ev VERSION_CONTROL
3747environment variable is used instead. Please note that these options and variables
3748control backup file names; they do not affect the choice of revision control
3749system (see Section
3750.Dq Revision Control ) .
3751.Pp
3752The values of these environment variables and the argument to the
3753.Op -V
3754option are like the GNU Emacs
3755.Li version-control
3756variable (see Section
3757.Dq Backup Names ,
3758for more information on backup versions in Emacs). They also recognize synonyms
3759that are more descriptive. The valid values are listed below; unique abbreviations
3760are acceptable.
3761.Pp
3762.Bl -tag -width Ds
3763.It  t
3764.It  numbered
3765Always make numbered backups.
3766.Pp
3767.It  nil
3768.It  existing
3769Make numbered backups of files that already have them, simple backups of the
3770others. This is the default.
3771.Pp
3772.It  never
3773.It  simple
3774Always make simple backups.
3775.El
3776.Pp
3777You can also tell
3778.Xr patch
3779to prepend a prefix, such as a directory name, to produce backup file names.
3780The
3781.Op -B Va prefix
3782or
3783.Op --prefix= Va prefix
3784option makes backup files by prepending
3785.Va prefix
3786to them. The
3787.Op -Y Va prefix
3788or
3789.Op --basename-prefix= Va prefix
3790prepends
3791.Va prefix
3792to the last file name component of backup file names instead; for example,
3793.Op -Y ~
3794causes the backup name for
3795.Pa dir/file.c
3796to be
3797.Pa dir/~file.c .
3798If you use either of these prefix options, the suffix-based options are ignored.
3799.Pp
3800If you specify the output file with the
3801.Op -o
3802option, that file is the one that is backed up, not the input file.
3803.Pp
3804Options that affect the names of backup files do not affect whether backups
3805are made. For example, if you specify the
3806.Op --no-backup-if-mismatch
3807option, none of the options described in this section have any affect, because
3808no backups are made.
3809.Pp
3810.Ss  Reject File Names
3811The names for reject files (files containing patches that
3812.Xr patch
3813could not find a place to apply) are normally the name of the output file
3814with
3815.Li .rej
3816appended (or
3817.Li #
3818if using
3819.Li .rej
3820would make the backup file name too long).
3821.Pp
3822Alternatively, you can tell
3823.Xr patch
3824to place all of the rejected patches in a single file. The
3825.Op -r Va reject-file
3826or
3827.Op --reject-file= Va reject-file
3828option uses
3829.Va reject-file
3830as the reject file name.
3831.Pp
3832.Ss  Messages and Questions from Xr patch
3833.Xr patch
3834can produce a variety of messages, especially if it has trouble decoding its
3835input. In a few situations where it's not sure how to proceed,
3836.Xr patch
3837normally prompts you for more information from the keyboard. There are options
3838to produce more or fewer messages, to have it not ask for keyboard input,
3839and to affect the way that file names are quoted in messages.
3840.Pp
3841.Xr patch
3842exits with status 0 if all hunks are applied successfully, 1 if some hunks
3843cannot be applied, and 2 if there is more serious trouble. When applying a
3844set of patches in a loop, you should check the exit status, so you don't apply
3845a later patch to a partially patched file.
3846.Pp
3847.Em  Controlling the Verbosity of Xr patch
3848.Pp
3849You can cause
3850.Xr patch
3851to produce more messages by using the
3852.Op --verbose
3853option. For example, when you give this option, the message
3854.Li Hmm...
3855indicates that
3856.Xr patch
3857is reading text in the patch file, attempting to determine whether there is
3858a patch in that text, and if so, what kind of patch it is.
3859.Pp
3860You can inhibit all terminal output from
3861.Xr patch ,
3862unless an error occurs, by using the
3863.Op -s ,
3864.Op --quiet ,
3865or
3866.Op --silent
3867option.
3868.Pp
3869.Em  Inhibiting Keyboard Input
3870.Pp
3871There are two ways you can prevent
3872.Xr patch
3873from asking you any questions. The
3874.Op -f
3875or
3876.Op --force
3877option assumes that you know what you are doing. It causes
3878.Xr patch
3879to do the following:
3880.Pp
3881.Bl -bullet
3882.It
3883Skip patches that do not contain file names in their headers.
3884.Pp
3885.It
3886Patch files even though they have the wrong version for the
3887.Li Prereq:
3888line in the patch;
3889.Pp
3890.It
3891Assume that patches are not reversed even if they look like they are.
3892.El
3893.Pp
3894The
3895.Op -t
3896or
3897.Op --batch
3898option is similar to
3899.Op -f ,
3900in that it suppresses questions, but it makes somewhat different assumptions:
3901.Pp
3902.Bl -bullet
3903.It
3904Skip patches that do not contain file names in their headers (the same as
3905.Op -f ) .
3906.Pp
3907.It
3908Skip patches for which the file has the wrong version for the
3909.Li Prereq:
3910line in the patch;
3911.Pp
3912.It
3913Assume that patches are reversed if they look like they are.
3914.El
3915.Pp
3916.Em  Xr patch Quoting Style
3917.Pp
3918When
3919.Xr patch
3920outputs a file name in a diagnostic message, it can format the name in any
3921of several ways. This can be useful to output file names unambiguously, even
3922if they contain punctuation or special characters like newlines. The
3923.Op --quoting-style= Va word
3924option controls how names are output. The
3925.Va word
3926should be one of the following:
3927.Pp
3928.Bl -tag -width Ds
3929.It  literal
3930Output names as-is.
3931.It  shell
3932Quote names for the shell if they contain shell metacharacters or would cause
3933ambiguous output.
3934.It  shell-always
3935Quote names for the shell, even if they would normally not require quoting.
3936.It  c
3937Quote names as for a C language string.
3938.It  escape
3939Quote as with
3940.Li c
3941except omit the surrounding double-quote characters.
3942.El
3943.Pp
3944You can specify the default value of the
3945.Op --quoting-style
3946option with the environment variable
3947.Ev QUOTING_STYLE .
3948If that environment variable is not set, the default value is
3949.Li shell ,
3950but this default may change in a future version of
3951.Xr patch .
3952.Pp
3953.Ss  Xr patch and the POSIX Standard
3954If you specify the
3955.Op --posix
3956option, or set the
3957.Ev POSIXLY_CORRECT
3958environment variable,
3959.Xr patch
3960conforms more strictly to the POSIX standard, as follows:
3961.Pp
3962.Bl -bullet
3963.It
3964Take the first existing file from the list (old, new, index) when intuiting
3965file names from diff headers.See Section
3966.Dq Multiple Patches .
3967.Pp
3968.It
3969Do not remove files that are removed by a diff.See Section
3970.Dq Creating and Removing .
3971.Pp
3972.It
3973Do not ask whether to get files from RCS, ClearCase, or SCCS.See Section
3974.Dq Revision Control .
3975.Pp
3976.It
3977Require that all options precede the files in the command line.
3978.Pp
3979.It
3980Do not backup files, even when there is a mismatch.See Section
3981.Dq Backups .
3982.Pp
3983.El
3984.Ss  GNU Xr patch and Traditional Xr patch
3985The current version of GNU
3986.Xr patch
3987normally follows the POSIX standard.See Section
3988.Dq patch and POSIX ,
3989for the few exceptions to this general rule.
3990.Pp
3991Unfortunately, POSIX redefined the behavior of
3992.Xr patch
3993in several important ways. You should be aware of the following differences
3994if you must interoperate with traditional
3995.Xr patch ,
3996or with GNU
3997.Xr patch
3998version 2.1 and earlier.
3999.Pp
4000.Bl -bullet
4001.It
4002In traditional
4003.Xr patch ,
4004the
4005.Op -p
4006option's operand was optional, and a bare
4007.Op -p
4008was equivalent to
4009.Op -p0 .
4010The
4011.Op -p
4012option now requires an operand, and
4013.Op -p 0
4014is now equivalent to
4015.Op -p0 .
4016For maximum compatibility, use options like
4017.Op -p0
4018and
4019.Op -p1 .
4020.Pp
4021Also, traditional
4022.Xr patch
4023simply counted slashes when stripping path prefixes;
4024.Xr patch
4025now counts pathname components. That is, a sequence of one or more adjacent
4026slashes now counts as a single slash. For maximum portability, avoid sending
4027patches containing
4028.Pa //
4029in file names.
4030.Pp
4031.It
4032In traditional
4033.Xr patch ,
4034backups were enabled by default. This behavior is now enabled with the
4035.Op -b
4036or
4037.Op --backup
4038option.
4039.Pp
4040Conversely, in POSIX
4041.Xr patch ,
4042backups are never made, even when there is a mismatch. In GNU
4043.Xr patch ,
4044this behavior is enabled with the
4045.Op --no-backup-if-mismatch
4046option, or by conforming to POSIX.
4047.Pp
4048The
4049.Op -b  Va suffix
4050option of traditional
4051.Xr patch
4052is equivalent to the
4053.Li -b -z  Va suffix
4054options of GNU
4055.Xr patch .
4056.Pp
4057.It
4058Traditional
4059.Xr patch
4060used a complicated (and incompletely documented) method to intuit the name
4061of the file to be patched from the patch header. This method did not conform
4062to POSIX, and had a few gotchas. Now
4063.Xr patch
4064uses a different, equally complicated (but better documented) method that
4065is optionally POSIX-conforming; we hope it has fewer gotchas. The two methods
4066are compatible if the file names in the context diff header and the
4067.Li Index:
4068line are all identical after prefix-stripping. Your patch is normally compatible
4069if each header's file names all contain the same number of slashes.
4070.Pp
4071.It
4072When traditional
4073.Xr patch
4074asked the user a question, it sent the question to standard error and looked
4075for an answer from the first file in the following list that was a terminal:
4076standard error, standard output,
4077.Pa /dev/tty ,
4078and standard input. Now
4079.Xr patch
4080sends questions to standard output and gets answers from
4081.Pa /dev/tty .
4082Defaults for some answers have been changed so that
4083.Xr patch
4084never goes into an infinite loop when using default answers.
4085.Pp
4086.It
4087Traditional
4088.Xr patch
4089exited with a status value that counted the number of bad hunks, or with status
40901 if there was real trouble. Now
4091.Xr patch
4092exits with status 1 if some hunks failed, or with 2 if there was real trouble.
4093.Pp
4094.It
4095Limit yourself to the following options when sending instructions meant to
4096be executed by anyone running GNU
4097.Xr patch ,
4098traditional
4099.Xr patch ,
4100or a
4101.Xr patch
4102that conforms to POSIX. Spaces are significant in the following list, and
4103operands are required.
4104.Pp
4105.Bd -literal -offset indent
4106-c
4107-d dir
4108-D define
4109-e
4110-l
4111-n
4112-N
4113-o outfile
4114-pnum
4115-R
4116-r rejectfile
4117.Ed
4118.Pp
4119.El
4120.Sh  Tips for Making and Using Patches
4121Use some common sense when making and using patches. For example, when sending
4122bug fixes to a program's maintainer, send several small patches, one per independent
4123subject, instead of one large, harder-to-digest patch that covers all the
4124subjects.
4125.Pp
4126Here are some other things you should keep in mind if you are going to distribute
4127patches for updating a software package.
4128.Pp
4129.Ss  Tips for Patch Producers
4130To create a patch that changes an older version of a package into a newer
4131version, first make a copy of the older and newer versions in adjacent subdirectories.
4132It is common to do that by unpacking
4133.Xr tar
4134archives of the two versions.
4135.Pp
4136To generate the patch, use the command
4137.Li diff -Naur Va old Va new
4138where
4139.Va old
4140and
4141.Va new
4142identify the old and new directories. The names
4143.Va old
4144and
4145.Va new
4146should not contain any slashes. The
4147.Op -N
4148option lets the patch create and remove files;
4149.Op -a
4150lets the patch update non-text files;
4151.Op -u
4152generates useful time stamps and enough context; and
4153.Op -r
4154lets the patch update subdirectories. Here is an example command, using Bourne
4155shell syntax:
4156.Pp
4157.Bd -literal -offset indent
4158diff -Naur gcc-3.0.3 gcc-3.0.4
4159.Ed
4160.Pp
4161Tell your recipients how to apply the patches. This should include which working
4162directory to use, and which
4163.Xr patch
4164options to use; the option
4165.Li -p1
4166is recommended. Test your procedure by pretending to be a recipient and applying
4167your patches to a copy of the original files.
4168.Pp
4169See Section.Dq Avoiding Common Mistakes ,
4170for how to avoid common mistakes when generating a patch.
4171.Pp
4172.Ss  Tips for Patch Consumers
4173A patch producer should tell recipients how to apply the patches, so the first
4174rule of thumb for a patch consumer is to follow the instructions supplied
4175with the patch.
4176.Pp
4177GNU
4178.Xr diff
4179can analyze files with arbitrarily long lines and files that end in incomplete
4180lines. However, older versions of
4181.Xr patch
4182cannot patch such files. If you are having trouble applying such patches,
4183try upgrading to a recent version of GNU
4184.Xr patch .
4185.Pp
4186.Ss  Avoiding Common Mistakes
4187When producing a patch for multiple files, apply
4188.Xr diff
4189to directories whose names do not have slashes. This reduces confusion when
4190the patch consumer specifies the
4191.Op -p Va number
4192option, since this option can have surprising results when the old and new
4193file names have different numbers of slashes. For example, do not send a patch
4194with a header that looks like this:
4195.Pp
4196.Bd -literal -offset indent
4197diff -Naur v2.0.29/prog/README prog/README
4198--- v2.0.29/prog/README	2002-03-10 23:30:39.942229878 -0800
4199+++ prog/README	2002-03-17 20:49:32.442260588 -0800
4200.Ed
4201.Pp
4202because the two file names have different numbers of slashes, and different
4203versions of
4204.Xr patch
4205interpret the file names differently. To avoid confusion, send output that
4206looks like this instead:
4207.Pp
4208.Bd -literal -offset indent
4209diff -Naur v2.0.29/prog/README v2.0.30/prog/README
4210--- v2.0.29/prog/README	2002-03-10 23:30:39.942229878 -0800
4211+++ v2.0.30/prog/README	2002-03-17 20:49:32.442260588 -0800
4212.Ed
4213.Pp
4214Make sure you have specified the file names correctly, either in a context
4215diff header or with an
4216.Li Index:
4217line. Take care to not send out reversed patches, since these make people
4218wonder whether they have already applied the patch.
4219.Pp
4220Avoid sending patches that compare backup file names like
4221.Pa README.orig
4222or
4223.Pa README~ ,
4224since this might confuse
4225.Xr patch
4226into patching a backup file instead of the real file. Instead, send patches
4227that compare the same base file names in different directories, e.g.
4228.Pa old/README
4229and
4230.Pa new/README .
4231.Pp
4232To save people from partially applying a patch before other patches that should
4233have gone before it, you can make the first patch in the patch file update
4234a file with a name like
4235.Pa patchlevel.h
4236or
4237.Pa version.c ,
4238which contains a patch level or version number. If the input file contains
4239the wrong version number,
4240.Xr patch
4241will complain immediately.
4242.Pp
4243An even clearer way to prevent this problem is to put a
4244.Li Prereq:
4245line before the patch. If the leading text in the patch file contains a line
4246that starts with
4247.Li Prereq: ,
4248.Xr patch
4249takes the next word from that line (normally a version number) and checks
4250whether the next input file contains that word, preceded and followed by either
4251white space or a newline. If not,
4252.Xr patch
4253prompts you for confirmation before proceeding. This makes it difficult to
4254accidentally apply patches in the wrong order.
4255.Pp
4256.Ss  Generating Smaller Patches
4257The simplest way to generate a patch is to use
4258.Li diff -Naur
4259(see Section
4260.Dq Tips for Patch Producers ) ,
4261but you might be able to reduce the size of the patch by renaming or removing
4262some files before making the patch. If the older version of the package contains
4263any files that the newer version does not, or if any files have been renamed
4264between the two versions, make a list of
4265.Xr rm
4266and
4267.Xr mv
4268commands for the user to execute in the old version directory before applying
4269the patch. Then run those commands yourself in the scratch directory.
4270.Pp
4271If there are any files that you don't need to include in the patch because
4272they can easily be rebuilt from other files (for example,
4273.Pa TAGS
4274and output from
4275.Xr yacc
4276and
4277.Xr makeinfo ) ,
4278exclude them from the patch by giving
4279.Xr diff
4280the
4281.Op -x Va pattern
4282option (see Section
4283.Dq Comparing Directories ) .
4284If you want your patch to modify a derived file because your recipients lack
4285tools to build it, make sure that the patch for the derived file follows any
4286patches for files that it depends on, so that the recipients' time stamps
4287will not confuse
4288.Xr make .
4289.Pp
4290Now you can create the patch using
4291.Li diff -Naur .
4292Make sure to specify the scratch directory first and the newer directory second.
4293.Pp
4294Add to the top of the patch a note telling the user any
4295.Xr rm
4296and
4297.Xr mv
4298commands to run before applying the patch. Then you can remove the scratch
4299directory.
4300.Pp
4301You can also shrink the patch size by using fewer lines of context, but bear
4302in mind that
4303.Xr patch
4304typically needs at least two lines for proper operation when patches do not
4305exactly match the input files.
4306.Pp
4307.Sh  Invoking Xr cmp
4308The
4309.Xr cmp
4310command compares two files, and if they differ, tells the first byte and line
4311number where they differ or reports that one file is a prefix of the other.
4312Bytes and lines are numbered starting with 1. The arguments of
4313.Xr cmp
4314are as follows:
4315.Pp
4316.Bd -literal -offset indent
4317cmp options... from-file [to-file [from-skip [to-skip]]]
4318.Ed
4319.Pp
4320The file name
4321.Pa -
4322is always the standard input.
4323.Xr cmp
4324also uses the standard input if one file name is omitted. The
4325.Va from-skip
4326and
4327.Va to-skip
4328operands specify how many bytes to ignore at the start of each file; they
4329are equivalent to the
4330.Op --ignore-initial= Va from-skip: Va to-skip
4331option.
4332.Pp
4333By default,
4334.Xr cmp
4335outputs nothing if the two files have the same contents. If one file is a
4336prefix of the other,
4337.Xr cmp
4338prints to standard error a message of the following form:
4339.Pp
4340.Bd -literal -offset indent
4341cmp: EOF on shorter-file
4342.Ed
4343.Pp
4344Otherwise,
4345.Xr cmp
4346prints to standard output a message of the following form:
4347.Pp
4348.Bd -literal -offset indent
4349from-file to-file differ: char byte-number, line line-number
4350.Ed
4351.Pp
4352The message formats can differ outside the POSIX locale. Also, POSIX allows
4353the EOF message to be followed by a blank and some additional information.
4354.Pp
4355An exit status of 0 means no differences were found, 1 means some differences
4356were found, and 2 means trouble.
4357.Pp
4358.Ss  Options to Xr cmp
4359Below is a summary of all of the options that GNU
4360.Xr cmp
4361accepts. Most options have two equivalent names, one of which is a single
4362letter preceded by
4363.Li - ,
4364and the other of which is a long name preceded by
4365.Li -- .
4366Multiple single letter options (unless they take an argument) can be combined
4367into a single command line word:
4368.Op -bl
4369is equivalent to
4370.Op -b -l .
4371.Pp
4372.Bl -tag -width Ds
4373.It  -b
4374.It  --print-bytes
4375Print the differing bytes. Display control bytes as a
4376.Li ^
4377followed by a letter of the alphabet and precede bytes that have the high
4378bit set with
4379.Li M-
4380(which stands for \(lqmeta\(rq).
4381.Pp
4382.It  --help
4383Output a summary of usage and then exit.
4384.Pp
4385.It  -i Va skip
4386.It  --ignore-initial= Va skip
4387Ignore any differences in the first
4388.Va skip
4389bytes of the input files. Treat files with fewer than
4390.Va skip
4391bytes as if they are empty. If
4392.Va skip
4393is of the form
4394.Op  Va from-skip: Va to-skip ,
4395skip the first
4396.Va from-skip
4397bytes of the first input file and the first
4398.Va to-skip
4399bytes of the second.
4400.Pp
4401.It  -l
4402.It  --verbose
4403Output the (decimal) byte numbers and (octal) values of all differing bytes,
4404instead of the default standard output.
4405.Pp
4406.It  -n Va count
4407.It  --bytes= Va count
4408Compare at most
4409.Va count
4410input bytes.
4411.Pp
4412.It  -s
4413.It  --quiet
4414.It  --silent
4415Do not print anything; only return an exit status indicating whether the files
4416differ.
4417.Pp
4418.It  -v
4419.It  --version
4420Output version information and then exit.
4421.El
4422.Pp
4423In the above table, operands that are byte counts are normally decimal, but
4424may be preceded by
4425.Li 0
4426for octal and
4427.Li 0x
4428for hexadecimal.
4429.Pp
4430A byte count can be followed by a suffix to specify a multiple of that count;
4431in this case an omitted integer is understood to be 1. A bare size letter,
4432or one followed by
4433.Li iB ,
4434specifies a multiple using powers of 1024. A size letter followed by
4435.Li B
4436specifies powers of 1000 instead. For example,
4437.Op -n 4M
4438and
4439.Op -n 4MiB
4440are equivalent to
4441.Op -n 4194304 ,
4442whereas
4443.Op -n 4MB
4444is equivalent to
4445.Op -n 4000000 .
4446This notation is upward compatible with the
4447.Lk http://www.bipm.fr/enus/3_SI/si-prefixes.html
4448for decimal multiples and with the
4449.Lk http://physics.nist.gov/cuu/Units/binary.html .
4450.Pp
4451The following suffixes are defined. Large sizes like
4452.Li 1Y
4453may be rejected by your computer due to limitations of its arithmetic.
4454.Pp
4455.Bl -tag -width Ds
4456.It  kB
4457kilobyte: 10^3 = 1000.
4458.It  k
4459.It  K
4460.It  KiB
4461kibibyte: 2^10 = 1024.
4462.Li K
4463is special: the SI prefix is
4464.Li k
4465and the IEC 60027-2 prefix is
4466.Li Ki ,
4467but tradition and POSIX use
4468.Li k
4469to mean
4470.Li KiB .
4471.It  MB
4472megabyte: 10^6 = 1,000,000.
4473.It  M
4474.It  MiB
4475mebibyte: 2^20 = 1,048,576.
4476.It  GB
4477gigabyte: 10^9 = 1,000,000,000.
4478.It  G
4479.It  GiB
4480gibibyte: 2^30 = 1,073,741,824.
4481.It  TB
4482terabyte: 10^12 = 1,000,000,000,000.
4483.It  T
4484.It  TiB
4485tebibyte: 2^40 = 1,099,511,627,776.
4486.It  PB
4487petabyte: 10^15 = 1,000,000,000,000,000.
4488.It  P
4489.It  PiB
4490pebibyte: 2^50 = 1,125,899,906,842,624.
4491.It  EB
4492exabyte: 10^18 = 1,000,000,000,000,000,000.
4493.It  E
4494.It  EiB
4495exbibyte: 2^60 = 1,152,921,504,606,846,976.
4496.It  ZB
4497zettabyte: 10^21 = 1,000,000,000,000,000,000,000
4498.It  Z
4499.It  ZiB
45002^70 = 1,180,591,620,717,411,303,424. (
4501.Li Zi
4502is a GNU extension to IEC 60027-2.)
4503.It  YB
4504yottabyte: 10^24 = 1,000,000,000,000,000,000,000,000.
4505.It  Y
4506.It  YiB
45072^80 = 1,208,925,819,614,629,174,706,176. (
4508.Li Yi
4509is a GNU extension to IEC 60027-2.)
4510.El
4511.Pp
4512.Sh  Invoking Xr diff
4513The format for running the
4514.Xr diff
4515command is:
4516.Pp
4517.Bd -literal -offset indent
4518diff options... files...
4519.Ed
4520.Pp
4521In the simplest case, two file names
4522.Va from-file
4523and
4524.Va to-file
4525are given, and
4526.Xr diff
4527compares the contents of
4528.Va from-file
4529and
4530.Va to-file .
4531A file name of
4532.Pa -
4533stands for text read from the standard input. As a special case,
4534.Li diff - -
4535compares a copy of standard input to itself.
4536.Pp
4537If one file is a directory and the other is not,
4538.Xr diff
4539compares the file in the directory whose name is that of the non-directory.
4540The non-directory file must not be
4541.Pa - .
4542.Pp
4543If two file names are given and both are directories,
4544.Xr diff
4545compares corresponding files in both directories, in alphabetical order; this
4546comparison is not recursive unless the
4547.Op -r
4548or
4549.Op --recursive
4550option is given.
4551.Xr diff
4552never compares the actual contents of a directory as if it were a file. The
4553file that is fully specified may not be standard input, because standard input
4554is nameless and the notion of \(lqfile with the same name\(rq does not apply.
4555.Pp
4556If the
4557.Op --from-file= Va file
4558option is given, the number of file names is arbitrary, and
4559.Va file
4560is compared to each named file. Similarly, if the
4561.Op --to-file= Va file
4562option is given, each named file is compared to
4563.Va file .
4564.Pp
4565.Xr diff
4566options begin with
4567.Li - ,
4568so normally file names may not begin with
4569.Li - .
4570However,
4571.Op --
4572as an argument by itself treats the remaining arguments as file names even
4573if they begin with
4574.Li - .
4575.Pp
4576An exit status of 0 means no differences were found, 1 means some differences
4577were found, and 2 means trouble. Normally, differing binary files count as
4578trouble, but this can be altered by using the
4579.Op -a
4580or
4581.Op --text
4582option, or the
4583.Op -q
4584or
4585.Op --brief
4586option.
4587.Pp
4588.Ss  Options to Xr diff
4589Below is a summary of all of the options that GNU
4590.Xr diff
4591accepts. Most options have two equivalent names, one of which is a single
4592letter preceded by
4593.Li - ,
4594and the other of which is a long name preceded by
4595.Li -- .
4596Multiple single letter options (unless they take an argument) can be combined
4597into a single command line word:
4598.Op -ac
4599is equivalent to
4600.Op -a -c .
4601Long named options can be abbreviated to any unique prefix of their name.
4602Brackets ([ and ]) indicate that an option takes an optional argument.
4603.Pp
4604.Bl -tag -width Ds
4605.It  -a
4606.It  --text
4607Treat all files as text and compare them line-by-line, even if they do not
4608seem to be text.See Section
4609.Dq Binary .
4610.Pp
4611.It  -b
4612.It  --ignore-space-change
4613Ignore changes in amount of white space.See Section
4614.Dq White Space .
4615.Pp
4616.It  -B
4617.It  --ignore-blank-lines
4618Ignore changes that just insert or delete blank lines.See Section
4619.Dq Blank Lines .
4620.Pp
4621.It  --binary
4622Read and write data in binary mode.See Section
4623.Dq Binary .
4624.Pp
4625.It  -c
4626Use the context output format, showing three lines of context.See Section
4627.Dq Context Format .
4628.Pp
4629.It  -C Va lines
4630.It  --context[= Va lines]
4631Use the context output format, showing
4632.Va lines
4633(an integer) lines of context, or three if
4634.Va lines
4635is not given.See Section
4636.Dq Context Format .
4637For proper operation,
4638.Xr patch
4639typically needs at least two lines of context.
4640.Pp
4641On older systems,
4642.Xr diff
4643supports an obsolete option
4644.Op - Va lines
4645that has effect when combined with
4646.Op -c
4647or
4648.Op -p .
4649POSIX 1003.1-2001 (see Section
4650.Dq Standards conformance )
4651does not allow this; use
4652.Op -C Va lines
4653instead.
4654.Pp
4655.It  --changed-group-format= Va format
4656Use
4657.Va format
4658to output a line group containing differing lines from both files in if-then-else
4659format.See Section
4660.Dq Line Group Formats .
4661.Pp
4662.It  -d
4663.It  --minimal
4664Change the algorithm perhaps find a smaller set of changes. This makes
4665.Xr diff
4666slower (sometimes much slower).See Section
4667.Dq diff Performance .
4668.Pp
4669.It  -D Va name
4670.It  --ifdef= Va name
4671Make merged
4672.Li #ifdef
4673format output, conditional on the preprocessor macro
4674.Va name .
4675See Section.Dq If-then-else .
4676.Pp
4677.It  -e
4678.It  --ed
4679Make output that is a valid
4680.Xr ed
4681script.See Section
4682.Dq ed Scripts .
4683.Pp
4684.It  -E
4685.It  --ignore-tab-expansion
4686Ignore changes due to tab expansion.See Section
4687.Dq White Space .
4688.Pp
4689.It  -f
4690.It  --forward-ed
4691Make output that looks vaguely like an
4692.Xr ed
4693script but has changes in the order they appear in the file.See Section
4694.Dq Forward ed .
4695.Pp
4696.It  -F Va regexp
4697.It  --show-function-line= Va regexp
4698In context and unified format, for each hunk of differences, show some of
4699the last preceding line that matches
4700.Va regexp .
4701See Section.Dq Specified Headings .
4702.Pp
4703.It  --from-file= Va file
4704Compare
4705.Va file
4706to each operand;
4707.Va file
4708may be a directory.
4709.Pp
4710.It  --help
4711Output a summary of usage and then exit.
4712.Pp
4713.It  --horizon-lines= Va lines
4714Do not discard the last
4715.Va lines
4716lines of the common prefix and the first
4717.Va lines
4718lines of the common suffix.See Section
4719.Dq diff Performance .
4720.Pp
4721.It  -i
4722.It  --ignore-case
4723Ignore changes in case; consider upper- and lower-case letters equivalent.See Section
4724.Dq Case Folding .
4725.Pp
4726.It  -I Va regexp
4727.It  --ignore-matching-lines= Va regexp
4728Ignore changes that just insert or delete lines that match
4729.Va regexp .
4730See Section.Dq Specified Lines .
4731.Pp
4732.It  --ignore-file-name-case
4733Ignore case when comparing file names during recursive comparison.See Section
4734.Dq Comparing Directories .
4735.Pp
4736.It  -l
4737.It  --paginate
4738Pass the output through
4739.Xr pr
4740to paginate it.See Section
4741.Dq Pagination .
4742.Pp
4743.It  --label= Va label
4744Use
4745.Va label
4746instead of the file name in the context format (see Section
4747.Dq Context Format )
4748and unified format (see Section
4749.Dq Unified Format )
4750headers.See Section
4751.Dq RCS .
4752.Pp
4753.It  --left-column
4754Print only the left column of two common lines in side by side format.See Section
4755.Dq Side by Side Format .
4756.Pp
4757.It  --line-format= Va format
4758Use
4759.Va format
4760to output all input lines in if-then-else format.See Section
4761.Dq Line Formats .
4762.Pp
4763.It  -n
4764.It  --rcs
4765Output RCS-format diffs; like
4766.Op -f
4767except that each command specifies the number of lines affected.See Section
4768.Dq RCS .
4769.Pp
4770.It  -N
4771.It  --new-file
4772In directory comparison, if a file is found in only one directory, treat it
4773as present but empty in the other directory.See Section
4774.Dq Comparing Directories .
4775.Pp
4776.It  --new-group-format= Va format
4777Use
4778.Va format
4779to output a group of lines taken from just the second file in if-then-else
4780format.See Section
4781.Dq Line Group Formats .
4782.Pp
4783.It  --new-line-format= Va format
4784Use
4785.Va format
4786to output a line taken from just the second file in if-then-else format.See Section
4787.Dq Line Formats .
4788.Pp
4789.It  --old-group-format= Va format
4790Use
4791.Va format
4792to output a group of lines taken from just the first file in if-then-else
4793format.See Section
4794.Dq Line Group Formats .
4795.Pp
4796.It  --old-line-format= Va format
4797Use
4798.Va format
4799to output a line taken from just the first file in if-then-else format.See Section
4800.Dq Line Formats .
4801.Pp
4802.It  -p
4803.It  --show-c-function
4804Show which C function each change is in.See Section
4805.Dq C Function Headings .
4806.Pp
4807.It  -q
4808.It  --brief
4809Report only whether the files differ, not the details of the differences.See Section
4810.Dq Brief .
4811.Pp
4812.It  -r
4813.It  --recursive
4814When comparing directories, recursively compare any subdirectories found.See Section
4815.Dq Comparing Directories .
4816.Pp
4817.It  -s
4818.It  --report-identical-files
4819Report when two files are the same.See Section
4820.Dq Comparing Directories .
4821.Pp
4822.It  -S Va file
4823.It  --starting-file= Va file
4824When comparing directories, start with the file
4825.Va file .
4826This is used for resuming an aborted comparison.See Section
4827.Dq Comparing Directories .
4828.Pp
4829.It  --speed-large-files
4830Use heuristics to speed handling of large files that have numerous scattered
4831small changes.See Section
4832.Dq diff Performance .
4833.Pp
4834.It  --strip-trailing-cr
4835Strip any trailing carriage return at the end of an input line.See Section
4836.Dq Binary .
4837.Pp
4838.It  --suppress-common-lines
4839Do not print common lines in side by side format.See Section
4840.Dq Side by Side Format .
4841.Pp
4842.It  -t
4843.It  --expand-tabs
4844Expand tabs to spaces in the output, to preserve the alignment of tabs in
4845the input files.See Section
4846.Dq Tabs .
4847.Pp
4848.It  -T
4849.It  --initial-tab
4850Output a tab rather than a space before the text of a line in normal or context
4851format. This causes the alignment of tabs in the line to look normal.See Section
4852.Dq Tabs .
4853.Pp
4854.It  --tabsize= Va columns
4855Assume that tab stops are set every
4856.Va columns
4857(default 8) print columns.See Section
4858.Dq Tabs .
4859.Pp
4860.It  --to-file= Va file
4861Compare each operand to
4862.Va file
4863;
4864.Va file
4865may be a directory.
4866.Pp
4867.It  -u
4868Use the unified output format, showing three lines of context.See Section
4869.Dq Unified Format .
4870.Pp
4871.It  --unchanged-group-format= Va format
4872Use
4873.Va format
4874to output a group of common lines taken from both files in if-then-else format.See Section
4875.Dq Line Group Formats .
4876.Pp
4877.It  --unchanged-line-format= Va format
4878Use
4879.Va format
4880to output a line common to both files in if-then-else format.See Section
4881.Dq Line Formats .
4882.Pp
4883.It  --unidirectional-new-file
4884When comparing directories, if a file appears only in the second directory
4885of the two, treat it as present but empty in the other.See Section
4886.Dq Comparing Directories .
4887.Pp
4888.It  -U Va lines
4889.It  --unified[= Va lines]
4890Use the unified output format, showing
4891.Va lines
4892(an integer) lines of context, or three if
4893.Va lines
4894is not given.See Section
4895.Dq Unified Format .
4896For proper operation,
4897.Xr patch
4898typically needs at least two lines of context.
4899.Pp
4900On older systems,
4901.Xr diff
4902supports an obsolete option
4903.Op - Va lines
4904that has effect when combined with
4905.Op -u .
4906POSIX 1003.1-2001 (see Section
4907.Dq Standards conformance )
4908does not allow this; use
4909.Op -U Va lines
4910instead.
4911.Pp
4912.It  -v
4913.It  --version
4914Output version information and then exit.
4915.Pp
4916.It  -w
4917.It  --ignore-all-space
4918Ignore white space when comparing lines.See Section
4919.Dq White Space .
4920.Pp
4921.It  -W Va columns
4922.It  --width= Va columns
4923Output at most
4924.Va columns
4925(default 130) print columns per line in side by side format.See Section
4926.Dq Side by Side Format .
4927.Pp
4928.It  -x Va pattern
4929.It  --exclude= Va pattern
4930When comparing directories, ignore files and subdirectories whose basenames
4931match
4932.Va pattern .
4933See Section.Dq Comparing Directories .
4934.Pp
4935.It  -X Va file
4936.It  --exclude-from= Va file
4937When comparing directories, ignore files and subdirectories whose basenames
4938match any pattern contained in
4939.Va file .
4940See Section.Dq Comparing Directories .
4941.Pp
4942.It  -y
4943.It  --side-by-side
4944Use the side by side output format.See Section
4945.Dq Side by Side Format .
4946.El
4947.Pp
4948.Sh  Invoking Xr diff3
4949The
4950.Xr diff3
4951command compares three files and outputs descriptions of their differences.
4952Its arguments are as follows:
4953.Pp
4954.Bd -literal -offset indent
4955diff3 options... mine older yours
4956.Ed
4957.Pp
4958The files to compare are
4959.Va mine ,
4960.Va older ,
4961and
4962.Va yours .
4963At most one of these three file names may be
4964.Pa - ,
4965which tells
4966.Xr diff3
4967to read the standard input for that file.
4968.Pp
4969An exit status of 0 means
4970.Xr diff3
4971was successful, 1 means some conflicts were found, and 2 means trouble.
4972.Pp
4973.Ss  Options to Xr diff3
4974Below is a summary of all of the options that GNU
4975.Xr diff3
4976accepts. Multiple single letter options (unless they take an argument) can
4977be combined into a single command line argument.
4978.Pp
4979.Bl -tag -width Ds
4980.It  -a
4981.It  --text
4982Treat all files as text and compare them line-by-line, even if they do not
4983appear to be text.See Section
4984.Dq Binary .
4985.Pp
4986.It  -A
4987.It  --show-all
4988Incorporate all unmerged changes from
4989.Va older
4990to
4991.Va yours
4992into
4993.Va mine ,
4994surrounding conflicts with bracket lines.See Section
4995.Dq Marking Conflicts .
4996.Pp
4997.It  --diff-program= Va program
4998Use the compatible comparison program
4999.Va program
5000to compare files instead of
5001.Xr diff .
5002.Pp
5003.It  -e
5004.It  --ed
5005Generate an
5006.Xr ed
5007script that incorporates all the changes from
5008.Va older
5009to
5010.Va yours
5011into
5012.Va mine .
5013See Section.Dq Which Changes .
5014.Pp
5015.It  -E
5016.It  --show-overlap
5017Like
5018.Op -e ,
5019except bracket lines from overlapping changes' first and third files.See Section
5020.Dq Marking Conflicts .
5021With
5022.Op -E ,
5023an overlapping change looks like this:
5024.Pp
5025.Bd -literal -offset indent
5026<<<<<<< mine
5027lines from mine
5028=======
5029lines from yours
5030>>>>>>> yours
5031.Ed
5032.Pp
5033.It  --help
5034Output a summary of usage and then exit.
5035.Pp
5036.It  -i
5037Generate
5038.Li w
5039and
5040.Li q
5041commands at the end of the
5042.Xr ed
5043script for System V compatibility. This option must be combined with one of
5044the
5045.Op -AeExX3
5046options, and may not be combined with
5047.Op -m .
5048See Section.Dq Saving the Changed File .
5049.Pp
5050.It  --label= Va label
5051Use the label
5052.Va label
5053for the brackets output by the
5054.Op -A ,
5055.Op -E
5056and
5057.Op -X
5058options. This option may be given up to three times, one for each input file.
5059The default labels are the names of the input files. Thus
5060.Li diff3 --label X --label Y --label Z -m A B C
5061acts like
5062.Li diff3 -m A B C ,
5063except that the output looks like it came from files named
5064.Li X ,
5065.Li Y
5066and
5067.Li Z
5068rather than from files named
5069.Li A ,
5070.Li B
5071and
5072.Li C .
5073See Section.Dq Marking Conflicts .
5074.Pp
5075.It  -m
5076.It  --merge
5077Apply the edit script to the first file and send the result to standard output.
5078Unlike piping the output from
5079.Xr diff3
5080to
5081.Xr ed ,
5082this works even for binary files and incomplete lines.
5083.Op -A
5084is assumed if no edit script option is specified.See Section
5085.Dq Bypassing ed .
5086.Pp
5087.It  --strip-trailing-cr
5088Strip any trailing carriage return at the end of an input line.See Section
5089.Dq Binary .
5090.Pp
5091.It  -T
5092.It  --initial-tab
5093Output a tab rather than two spaces before the text of a line in normal format.
5094This causes the alignment of tabs in the line to look normal.See Section
5095.Dq Tabs .
5096.Pp
5097.It  -v
5098.It  --version
5099Output version information and then exit.
5100.Pp
5101.It  -x
5102.It  --overlap-only
5103Like
5104.Op -e ,
5105except output only the overlapping changes.See Section
5106.Dq Which Changes .
5107.Pp
5108.It  -X
5109Like
5110.Op -E ,
5111except output only the overlapping changes. In other words, like
5112.Op -x ,
5113except bracket changes as in
5114.Op -E .
5115See Section.Dq Marking Conflicts .
5116.Pp
5117.It  -3
5118.It  --easy-only
5119Like
5120.Op -e ,
5121except output only the nonoverlapping changes.See Section
5122.Dq Which Changes .
5123.El
5124.Pp
5125.Sh  Invoking Xr patch
5126Normally
5127.Xr patch
5128is invoked like this:
5129.Pp
5130.Bd -literal -offset indent
5131patch <patchfile
5132.Ed
5133.Pp
5134The full format for invoking
5135.Xr patch
5136is:
5137.Pp
5138.Bd -literal -offset indent
5139patch options... [origfile [patchfile]]
5140.Ed
5141.Pp
5142You can also specify where to read the patch from with the
5143.Op -i Va patchfile
5144or
5145.Op --input= Va patchfile
5146option. If you do not specify
5147.Va patchfile ,
5148or if
5149.Va patchfile
5150is
5151.Pa - ,
5152.Xr patch
5153reads the patch (that is, the
5154.Xr diff
5155output) from the standard input.
5156.Pp
5157If you do not specify an input file on the command line,
5158.Xr patch
5159tries to intuit from the
5160.Em leading text
5161(any text in the patch that comes before the
5162.Xr diff
5163output) which file to edit.See Section
5164.Dq Multiple Patches .
5165.Pp
5166By default,
5167.Xr patch
5168replaces the original input file with the patched version, possibly after
5169renaming the original file into a backup file (see Section
5170.Dq Backup Names ,
5171for a description of how
5172.Xr patch
5173names backup files). You can also specify where to put the output with the
5174.Op -o Va file
5175or
5176.Op --output= Va file
5177option; however, do not use this option if
5178.Va file
5179is one of the input files.
5180.Pp
5181.Ss  Options to Xr patch
5182Here is a summary of all of the options that GNU
5183.Xr patch
5184accepts.See Section
5185.Dq patch and Tradition ,
5186for which of these options are safe to use in older versions of
5187.Xr patch .
5188.Pp
5189Multiple single-letter options that do not take an argument can be combined
5190into a single command line argument with only one dash.
5191.Pp
5192.Bl -tag -width Ds
5193.It  -b
5194.It  --backup
5195Back up the original contents of each file, even if backups would normally
5196not be made.See Section
5197.Dq Backups .
5198.Pp
5199.It  -B Va prefix
5200.It  --prefix= Va prefix
5201Prepend
5202.Va prefix
5203to backup file names.See Section
5204.Dq Backup Names .
5205.Pp
5206.It  --backup-if-mismatch
5207Back up the original contents of each file if the patch does not exactly match
5208the file. This is the default behavior when not conforming to POSIX.See Section
5209.Dq Backups .
5210.Pp
5211.It  --binary
5212Read and write all files in binary mode, except for standard output and
5213.Pa /dev/tty .
5214This option has no effect on POSIX-conforming systems like GNU/Linux. On systems
5215where this option makes a difference, the patch should be generated by
5216.Li diff -a --binary .
5217See Section.Dq Binary .
5218.Pp
5219.It  -c
5220.It  --context
5221Interpret the patch file as a context diff.See Section
5222.Dq patch Input .
5223.Pp
5224.It  -d Va directory
5225.It  --directory= Va directory
5226Make directory
5227.Va directory
5228the current directory for interpreting both file names in the patch file,
5229and file names given as arguments to other options.See Section
5230.Dq patch Directories .
5231.Pp
5232.It  -D Va name
5233.It  --ifdef= Va name
5234Make merged if-then-else output using
5235.Va name .
5236See Section.Dq If-then-else .
5237.Pp
5238.It  --dry-run
5239Print the results of applying the patches without actually changing any files.See Section
5240.Dq Dry Runs .
5241.Pp
5242.It  -e
5243.It  --ed
5244Interpret the patch file as an
5245.Xr ed
5246script.See Section
5247.Dq patch Input .
5248.Pp
5249.It  -E
5250.It  --remove-empty-files
5251Remove output files that are empty after the patches have been applied.See Section
5252.Dq Creating and Removing .
5253.Pp
5254.It  -f
5255.It  --force
5256Assume that the user knows exactly what he or she is doing, and do not ask
5257any questions.See Section
5258.Dq patch Messages .
5259.Pp
5260.It  -F Va lines
5261.It  --fuzz= Va lines
5262Set the maximum fuzz factor to
5263.Va lines .
5264See Section.Dq Inexact .
5265.Pp
5266.It  -g Va num
5267.It  --get= Va num
5268If
5269.Va num
5270is positive, get input files from a revision control system as necessary;
5271if zero, do not get the files; if negative, ask the user whether to get the
5272files.See Section
5273.Dq Revision Control .
5274.Pp
5275.It  --help
5276Output a summary of usage and then exit.
5277.Pp
5278.It  -i Va patchfile
5279.It  --input= Va patchfile
5280Read the patch from
5281.Va patchfile
5282rather than from standard input.See Section
5283.Dq patch Options .
5284.Pp
5285.It  -l
5286.It  --ignore-white-space
5287Let any sequence of blanks (spaces or tabs) in the patch file match any sequence
5288of blanks in the input file.See Section
5289.Dq Changed White Space .
5290.Pp
5291.It  -n
5292.It  --normal
5293Interpret the patch file as a normal diff.See Section
5294.Dq patch Input .
5295.Pp
5296.It  -N
5297.It  --forward
5298Ignore patches that
5299.Xr patch
5300thinks are reversed or already applied. See also
5301.Op -R .
5302See Section.Dq Reversed Patches .
5303.Pp
5304.It  --no-backup-if-mismatch
5305Do not back up the original contents of files. This is the default behavior
5306when conforming to POSIX.See Section
5307.Dq Backups .
5308.Pp
5309.It  -o Va file
5310.It  --output= Va file
5311Use
5312.Va file
5313as the output file name.See Section
5314.Dq patch Options .
5315.Pp
5316.It  -p Va number
5317.It  --strip= Va number
5318Set the file name strip count to
5319.Va number .
5320See Section.Dq patch Directories .
5321.Pp
5322.It  --posix
5323Conform to POSIX, as if the
5324.Ev POSIXLY_CORRECT
5325environment variable had been set.See Section
5326.Dq patch and POSIX .
5327.Pp
5328.It  --quoting-style= Va word
5329Use style
5330.Va word
5331to quote names in diagnostics, as if the
5332.Ev QUOTING_STYLE
5333environment variable had been set to
5334.Va word .
5335See Section.Dq patch Quoting Style .
5336.Pp
5337.It  -r Va reject-file
5338.It  --reject-file= Va reject-file
5339Use
5340.Va reject-file
5341as the reject file name.See Section
5342.Dq Reject Names .
5343.Pp
5344.It  -R
5345.It  --reverse
5346Assume that this patch was created with the old and new files swapped.See Section
5347.Dq Reversed Patches .
5348.Pp
5349.It  -s
5350.It  --quiet
5351.It  --silent
5352Work silently unless an error occurs.See Section
5353.Dq patch Messages .
5354.Pp
5355.It  -t
5356.It  --batch
5357Do not ask any questions.See Section
5358.Dq patch Messages .
5359.Pp
5360.It  -T
5361.It  --set-time
5362Set the modification and access times of patched files from time stamps given
5363in context diff headers, assuming that the context diff headers use local
5364time.See Section
5365.Dq Patching Time Stamps .
5366.Pp
5367.It  -u
5368.It  --unified
5369Interpret the patch file as a unified diff.See Section
5370.Dq patch Input .
5371.Pp
5372.It  -v
5373.It  --version
5374Output version information and then exit.
5375.Pp
5376.It  -V Va backup-style
5377.It  --version=control= Va backup-style
5378Select the naming convention for backup file names.See Section
5379.Dq Backup Names .
5380.Pp
5381.It  --verbose
5382Print more diagnostics than usual.See Section
5383.Dq patch Messages .
5384.Pp
5385.It  -x Va number
5386.It  --debug= Va number
5387Set internal debugging flags. Of interest only to
5388.Xr patch
5389patchers.
5390.Pp
5391.It  -Y Va prefix
5392.It  --basename-prefix= Va prefix
5393Prepend
5394.Va prefix
5395to base names of backup files.See Section
5396.Dq Backup Names .
5397.Pp
5398.It  -z Va suffix
5399.It  --suffix= Va suffix
5400Use
5401.Va suffix
5402as the backup extension instead of
5403.Li .orig
5404or
5405.Li ~ .
5406See Section.Dq Backup Names .
5407.Pp
5408.It  -Z
5409.It  --set-utc
5410Set the modification and access times of patched files from time stamps given
5411in context diff headers, assuming that the context diff headers use UTC.See Section
5412.Dq Patching Time Stamps .
5413.Pp
5414.El
5415.Sh  Invoking Xr sdiff
5416The
5417.Xr sdiff
5418command merges two files and interactively outputs the results. Its arguments
5419are as follows:
5420.Pp
5421.Bd -literal -offset indent
5422sdiff -o outfile options... from-file to-file
5423.Ed
5424.Pp
5425This merges
5426.Va from-file
5427with
5428.Va to-file ,
5429with output to
5430.Va outfile .
5431If
5432.Va from-file
5433is a directory and
5434.Va to-file
5435is not,
5436.Xr sdiff
5437compares the file in
5438.Va from-file
5439whose file name is that of
5440.Va to-file ,
5441and vice versa.
5442.Va from-file
5443and
5444.Va to-file
5445may not both be directories.
5446.Pp
5447.Xr sdiff
5448options begin with
5449.Li - ,
5450so normally
5451.Va from-file
5452and
5453.Va to-file
5454may not begin with
5455.Li - .
5456However,
5457.Op --
5458as an argument by itself treats the remaining arguments as file names even
5459if they begin with
5460.Li - .
5461You may not use
5462.Pa -
5463as an input file.
5464.Pp
5465.Xr sdiff
5466without
5467.Op -o
5468(or
5469.Op --output )
5470produces a side-by-side difference. This usage is obsolete; use the
5471.Op -y
5472or
5473.Op --side-by-side
5474option of
5475.Xr diff
5476instead.
5477.Pp
5478An exit status of 0 means no differences were found, 1 means some differences
5479were found, and 2 means trouble.
5480.Pp
5481.Ss  Options to Xr sdiff
5482Below is a summary of all of the options that GNU
5483.Xr sdiff
5484accepts. Each option has two equivalent names, one of which is a single letter
5485preceded by
5486.Li - ,
5487and the other of which is a long name preceded by
5488.Li -- .
5489Multiple single letter options (unless they take an argument) can be combined
5490into a single command line argument. Long named options can be abbreviated
5491to any unique prefix of their name.
5492.Pp
5493.Bl -tag -width Ds
5494.It  -a
5495.It  --text
5496Treat all files as text and compare them line-by-line, even if they do not
5497appear to be text.See Section
5498.Dq Binary .
5499.Pp
5500.It  -b
5501.It  --ignore-space-change
5502Ignore changes in amount of white space.See Section
5503.Dq White Space .
5504.Pp
5505.It  -B
5506.It  --ignore-blank-lines
5507Ignore changes that just insert or delete blank lines.See Section
5508.Dq Blank Lines .
5509.Pp
5510.It  -d
5511.It  --minimal
5512Change the algorithm to perhaps find a smaller set of changes. This makes
5513.Xr sdiff
5514slower (sometimes much slower).See Section
5515.Dq diff Performance .
5516.Pp
5517.It  --diff-program= Va program
5518Use the compatible comparison program
5519.Va program
5520to compare files instead of
5521.Xr diff .
5522.Pp
5523.It  -E
5524.It  --ignore-tab-expansion
5525Ignore changes due to tab expansion.See Section
5526.Dq White Space .
5527.Pp
5528.It  --help
5529Output a summary of usage and then exit.
5530.Pp
5531.It  -i
5532.It  --ignore-case
5533Ignore changes in case; consider upper- and lower-case to be the same.See Section
5534.Dq Case Folding .
5535.Pp
5536.It  -I Va regexp
5537.It  --ignore-matching-lines= Va regexp
5538Ignore changes that just insert or delete lines that match
5539.Va regexp .
5540See Section.Dq Specified Lines .
5541.Pp
5542.It  -l
5543.It  --left-column
5544Print only the left column of two common lines.See Section
5545.Dq Side by Side Format .
5546.Pp
5547.It  -o Va file
5548.It  --output= Va file
5549Put merged output into
5550.Va file .
5551This option is required for merging.
5552.Pp
5553.It  -s
5554.It  --suppress-common-lines
5555Do not print common lines.See Section
5556.Dq Side by Side Format .
5557.Pp
5558.It  --speed-large-files
5559Use heuristics to speed handling of large files that have numerous scattered
5560small changes.See Section
5561.Dq diff Performance .
5562.Pp
5563.It  --strip-trailing-cr
5564Strip any trailing carriage return at the end of an input line.See Section
5565.Dq Binary .
5566.Pp
5567.It  -t
5568.It  --expand-tabs
5569Expand tabs to spaces in the output, to preserve the alignment of tabs in
5570the input files.See Section
5571.Dq Tabs .
5572.Pp
5573.It  --tabsize= Va columns
5574Assume that tab stops are set every
5575.Va columns
5576(default 8) print columns.See Section
5577.Dq Tabs .
5578.Pp
5579.It  -v
5580.It  --version
5581Output version information and then exit.
5582.Pp
5583.It  -w Va columns
5584.It  --width= Va columns
5585Output at most
5586.Va columns
5587(default 130) print columns per line.See Section
5588.Dq Side by Side Format .
5589Note that for historical reasons, this option is
5590.Op -W
5591in
5592.Xr diff ,
5593.Op -w
5594in
5595.Xr sdiff .
5596.Pp
5597.It  -W
5598.It  --ignore-all-space
5599Ignore white space when comparing lines.See Section
5600.Dq White Space .
5601Note that for historical reasons, this option is
5602.Op -w
5603in
5604.Xr diff ,
5605.Op -W
5606in
5607.Xr sdiff .
5608.El
5609.Pp
5610.Sh  Standards conformance
5611In a few cases, the GNU utilities' default behavior is incompatible with the
5612POSIX standard. To suppress these incompatibilities, define the
5613.Ev POSIXLY_CORRECT
5614environment variable. Unless you are checking for POSIX conformance, you probably
5615do not need to define
5616.Ev POSIXLY_CORRECT .
5617.Pp
5618Normally options and operands can appear in any order, and programs act as
5619if all the options appear before any operands. For example,
5620.Li diff lao tzu -C 2
5621acts like
5622.Li diff -C 2 lao tzu ,
5623since
5624.Li 2
5625is an option-argument of
5626.Op -C .
5627However, if the
5628.Ev POSIXLY_CORRECT
5629environment variable is set, options must appear before operands, unless otherwise
5630specified for a particular command.
5631.Pp
5632Newer versions of POSIX are occasionally incompatible with older versions.
5633For example, older versions of POSIX allowed the command
5634.Li diff -c -10
5635to have the same meaning as
5636.Li diff -C 10 ,
5637but POSIX 1003.1-2001
5638.Li diff
5639no longer allows digit-string options like
5640.Op -10 .
5641.Pp
5642The GNU utilities normally conform to the version of POSIX that is standard
5643for your system. To cause them to conform to a different version of POSIX,
5644define the
5645.Ev _POSIX2_VERSION
5646environment variable to a value of the form
5647.Va yyyymm
5648specifying the year and month the standard was adopted. Two values are currently
5649supported for
5650.Ev _POSIX2_VERSION :
5651.Li 199209
5652stands for POSIX 1003.2-1992, and
5653.Li 200112
5654stands for POSIX 1003.1-2001. For example, if you are running older software
5655that assumes an older version of POSIX and uses
5656.Li diff -c -10 ,
5657you can work around the compatibility problems by setting
5658.Li _POSIX2_VERSION=199209
5659in your environment.
5660.Pp
5661.Sh  Future Projects
5662Here are some ideas for improving GNU
5663.Xr diff
5664and
5665.Xr patch .
5666The GNU project has identified some improvements as potential programming
5667projects for volunteers. You can also help by reporting any bugs that you
5668find.
5669.Pp
5670If you are a programmer and would like to contribute something to the GNU
5671project, please consider volunteering for one of these projects. If you are
5672seriously contemplating work, please write to
5673.Mt gvc@gnu.org
5674to coordinate with other volunteers.
5675.Pp
5676.Ss  Suggested Projects for Improving GNU Xr diff and Xr patch
5677One should be able to use GNU
5678.Xr diff
5679to generate a patch from any pair of directory trees, and given the patch
5680and a copy of one such tree, use
5681.Xr patch
5682to generate a faithful copy of the other. Unfortunately, some changes to directory
5683trees cannot be expressed using current patch formats; also,
5684.Xr patch
5685does not handle some of the existing formats. These shortcomings motivate
5686the following suggested projects.
5687.Pp
5688.Em  Handling Multibyte and Varying-Width Characters
5689.Pp
5690.Xr diff ,
5691.Xr diff3
5692and
5693.Xr sdiff
5694treat each line of input as a string of unibyte characters. This can mishandle
5695multibyte characters in some cases. For example, when asked to ignore spaces,
5696.Xr diff
5697does not properly ignore a multibyte space character.
5698.Pp
5699Also,
5700.Xr diff
5701currently assumes that each byte is one column wide, and this assumption is
5702incorrect in some locales, e.g., locales that use UTF-8 encoding. This causes
5703problems with the
5704.Op -y
5705or
5706.Op --side-by-side
5707option of
5708.Xr diff .
5709.Pp
5710These problems need to be fixed without unduly affecting the performance of
5711the utilities in unibyte environments.
5712.Pp
5713The IBM GNU/Linux Technology Center Internationalization Team has proposed
5714.Lk http://oss.software.ibm.com/developer/opensource/linux/patches/i18n/diffutils-2.7.2-i18n-0.1.patch.gz .
5715Unfortunately, these patches are incomplete and are to an older version of
5716.Xr diff ,
5717so more work needs to be done in this area.
5718.Pp
5719.Em  Handling Changes to the Directory Structure
5720.Pp
5721.Xr diff
5722and
5723.Xr patch
5724do not handle some changes to directory structure. For example, suppose one
5725directory tree contains a directory named
5726.Li D
5727with some subsidiary files, and another contains a file with the same name
5728.Li D .
5729.Li diff -r
5730does not output enough information for
5731.Xr patch
5732to transform the directory subtree into the file.
5733.Pp
5734There should be a way to specify that a file has been removed without having
5735to include its entire contents in the patch file. There should also be a way
5736to tell
5737.Xr patch
5738that a file was renamed, even if there is no way for
5739.Xr diff
5740to generate such information. There should be a way to tell
5741.Xr patch
5742that a file's time stamp has changed, even if its contents have not changed.
5743.Pp
5744These problems can be fixed by extending the
5745.Xr diff
5746output format to represent changes in directory structure, and extending
5747.Xr patch
5748to understand these extensions.
5749.Pp
5750.Em  Files that are Neither Directories Nor Regular Files
5751.Pp
5752Some files are neither directories nor regular files: they are unusual files
5753like symbolic links, device special files, named pipes, and sockets. Currently,
5754.Xr diff
5755treats symbolic links as if they were the pointed-to files, except that a
5756recursive
5757.Xr diff
5758reports an error if it detects infinite loops of symbolic links (e.g., symbolic
5759links to
5760.Pa .. ) .
5761.Xr diff
5762treats other special files like regular files if they are specified at the
5763top level, but simply reports their presence when comparing directories. This
5764means that
5765.Xr patch
5766cannot represent changes to such files. For example, if you change which file
5767a symbolic link points to,
5768.Xr diff
5769outputs the difference between the two files, instead of the change to the
5770symbolic link.
5771.Pp
5772.Xr diff
5773should optionally report changes to special files specially, and
5774.Xr patch
5775should be extended to understand these extensions.
5776.Pp
5777.Em  File Names that Contain Unusual Characters
5778.Pp
5779When a file name contains an unusual character like a newline or white space,
5780.Li diff -r
5781generates a patch that
5782.Xr patch
5783cannot parse. The problem is with format of
5784.Xr diff
5785output, not just with
5786.Xr patch ,
5787because with odd enough file names one can cause
5788.Xr diff
5789to generate a patch that is syntactically correct but patches the wrong files.
5790The format of
5791.Xr diff
5792output should be extended to handle all possible file names.
5793.Pp
5794.Em  Outputting Diffs in Time Stamp Order
5795.Pp
5796Applying
5797.Xr patch
5798to a multiple-file diff can result in files whose time stamps are out of order.
5799GNU
5800.Xr patch
5801has options to restore the time stamps of the updated files (see Section
5802.Dq Patching Time Stamps ) ,
5803but sometimes it is useful to generate a patch that works even if the recipient
5804does not have GNU patch, or does not use these options. One way to do this
5805would be to implement a
5806.Xr diff
5807option to output diffs in time stamp order.
5808.Pp
5809.Em  Ignoring Certain Changes
5810.Pp
5811It would be nice to have a feature for specifying two strings, one in
5812.Va from-file
5813and one in
5814.Va to-file ,
5815which should be considered to match. Thus, if the two strings are
5816.Li foo
5817and
5818.Li bar ,
5819then if two lines differ only in that
5820.Li foo
5821in file 1 corresponds to
5822.Li bar
5823in file 2, the lines are treated as identical.
5824.Pp
5825It is not clear how general this feature can or should be, or what syntax
5826should be used for it.
5827.Pp
5828A partial substitute is to filter one or both files before comparing, e.g.:
5829.Pp
5830.Bd -literal -offset indent
5831sed 's/foo/bar/g' file1 | diff - file2
5832.Ed
5833.Pp
5834However, this outputs the filtered text, not the original.
5835.Pp
5836.Em  Improving Performance
5837.Pp
5838When comparing two large directory structures, one of which was originally
5839copied from the other with time stamps preserved (e.g., with
5840.Li cp -pR ) ,
5841it would greatly improve performance if an option told
5842.Xr diff
5843to assume that two files with the same size and time stamps have the same
5844content.See Section
5845.Dq diff Performance .
5846.Pp
5847.Ss  Reporting Bugs
5848If you think you have found a bug in GNU
5849.Xr cmp ,
5850.Xr diff ,
5851.Xr diff3 ,
5852or
5853.Xr sdiff ,
5854please report it by electronic mail to the
5855.Lk http://mail.gnu.org/mailman/listinfo/bug-gnu-utils
5856.Mt bug-gnu-utils@gnu.org .
5857Please send bug reports for GNU
5858.Xr patch
5859to
5860.Mt bug-patch@gnu.org .
5861Send as precise a description of the problem as you can, including the output
5862of the
5863.Op --version
5864option and sample input files that produce the bug, if applicable. If you
5865have a nontrivial fix for the bug, please send it as well. If you have a patch,
5866please send it too. It may simplify the maintainer's job if the patch is relative
5867to a recent test release, which you can find in the directory
5868.Lk ftp://alpha.gnu.org/gnu/diffutils/ .
5869.Pp
5870.Sh  Copying This Manual
5871.Ss  GNU Free Documentation License
5872.Bd -filled -offset indent
5873Copyright \(co 2000,2001,2002 Free Software Foundation, Inc. 59 Temple Place,
5874Suite 330, Boston, MA 02111-1307, USA
5875.Pp
5876Everyone is permitted to copy and distribute verbatim copies of this license
5877document, but changing it is not allowed.
5878.Ed
5879.Pp
5880.Bl -enum
5881.It
5882PREAMBLE
5883.Pp
5884The purpose of this License is to make a manual, textbook, or other functional
5885and useful document
5886.Em free
5887in the sense of freedom: to assure everyone the effective freedom to copy
5888and redistribute it, with or without modifying it, either commercially or
5889noncommercially. Secondarily, this License preserves for the author and publisher
5890a way to get credit for their work, while not being considered responsible
5891for modifications made by others.
5892.Pp
5893This License is a kind of \(lqcopyleft\(rq, which means that derivative works of the
5894document must themselves be free in the same sense. It complements the GNU
5895General Public License, which is a copyleft license designed for free software.
5896.Pp
5897We have designed this License in order to use it for manuals for free software,
5898because free software needs free documentation: a free program should come
5899with manuals providing the same freedoms that the software does. But this
5900License is not limited to software manuals; it can be used for any textual
5901work, regardless of subject matter or whether it is published as a printed
5902book. We recommend this License principally for works whose purpose is instruction
5903or reference.
5904.Pp
5905.It
5906APPLICABILITY AND DEFINITIONS
5907.Pp
5908This License applies to any manual or other work, in any medium, that contains
5909a notice placed by the copyright holder saying it can be distributed under
5910the terms of this License. Such a notice grants a world-wide, royalty-free
5911license, unlimited in duration, to use that work under the conditions stated
5912herein. The \(lqDocument\(rq, below, refers to any such manual or work. Any member
5913of the public is a licensee, and is addressed as \(lqyou\(rq. You accept the license
5914if you copy, modify or distribute the work in a way requiring permission under
5915copyright law.
5916.Pp
5917A \(lqModified Version\(rq of the Document means any work containing the Document
5918or a portion of it, either copied verbatim, or with modifications and/or translated
5919into another language.
5920.Pp
5921A \(lqSecondary Section\(rq is a named appendix or a front-matter section of the Document
5922that deals exclusively with the relationship of the publishers or authors
5923of the Document to the Document's overall subject (or to related matters)
5924and contains nothing that could fall directly within that overall subject.
5925(Thus, if the Document is in part a textbook of mathematics, a Secondary Section
5926may not explain any mathematics.) The relationship could be a matter of historical
5927connection with the subject or with related matters, or of legal, commercial,
5928philosophical, ethical or political position regarding them.
5929.Pp
5930The \(lqInvariant Sections\(rq are certain Secondary Sections whose titles are designated,
5931as being those of Invariant Sections, in the notice that says that the Document
5932is released under this License. If a section does not fit the above definition
5933of Secondary then it is not allowed to be designated as Invariant. The Document
5934may contain zero Invariant Sections. If the Document does not identify any
5935Invariant Sections then there are none.
5936.Pp
5937The \(lqCover Texts\(rq are certain short passages of text that are listed, as Front-Cover
5938Texts or Back-Cover Texts, in the notice that says that the Document is released
5939under this License. A Front-Cover Text may be at most 5 words, and a Back-Cover
5940Text may be at most 25 words.
5941.Pp
5942A \(lqTransparent\(rq copy of the Document means a machine-readable copy, represented
5943in a format whose specification is available to the general public, that is
5944suitable for revising the document straightforwardly with generic text editors
5945or (for images composed of pixels) generic paint programs or (for drawings)
5946some widely available drawing editor, and that is suitable for input to text
5947formatters or for automatic translation to a variety of formats suitable for
5948input to text formatters. A copy made in an otherwise Transparent file format
5949whose markup, or absence of markup, has been arranged to thwart or discourage
5950subsequent modification by readers is not Transparent. An image format is
5951not Transparent if used for any substantial amount of text. A copy that is
5952not \(lqTransparent\(rq is called \(lqOpaque\(rq.
5953.Pp
5954Examples of suitable formats for Transparent copies include plain ascii without
5955markup, Texinfo input format, LaTeX input format, SGML or XML using a publicly
5956available DTD, and standard-conforming simple HTML, PostScript or PDF designed
5957for human modification. Examples of transparent image formats include PNG,
5958XCF and JPG. Opaque formats include proprietary formats that can be read and
5959edited only by proprietary word processors, SGML or XML for which the DTD
5960and/or processing tools are not generally available, and the machine-generated
5961HTML, PostScript or PDF produced by some word processors for output purposes
5962only.
5963.Pp
5964The \(lqTitle Page\(rq means, for a printed book, the title page itself, plus such
5965following pages as are needed to hold, legibly, the material this License
5966requires to appear in the title page. For works in formats which do not have
5967any title page as such, \(lqTitle Page\(rq means the text near the most prominent
5968appearance of the work's title, preceding the beginning of the body of the
5969text.
5970.Pp
5971A section \(lqEntitled XYZ\(rq means a named subunit of the Document whose title either
5972is precisely XYZ or contains XYZ in parentheses following text that translates
5973XYZ in another language. (Here XYZ stands for a specific section name mentioned
5974below, such as \(lqAcknowledgements\(rq, \(lqDedications\(rq, \(lqEndorsements\(rq, or \(lqHistory\(rq.) To
5975\(lqPreserve the Title\(rq of such a section when you modify the Document means that
5976it remains a section \(lqEntitled XYZ\(rq according to this definition.
5977.Pp
5978The Document may include Warranty Disclaimers next to the notice which states
5979that this License applies to the Document. These Warranty Disclaimers are
5980considered to be included by reference in this License, but only as regards
5981disclaiming warranties: any other implication that these Warranty Disclaimers
5982may have is void and has no effect on the meaning of this License.
5983.Pp
5984.It
5985VERBATIM COPYING
5986.Pp
5987You may copy and distribute the Document in any medium, either commercially
5988or noncommercially, provided that this License, the copyright notices, and
5989the license notice saying this License applies to the Document are reproduced
5990in all copies, and that you add no other conditions whatsoever to those of
5991this License. You may not use technical measures to obstruct or control the
5992reading or further copying of the copies you make or distribute. However,
5993you may accept compensation in exchange for copies. If you distribute a large
5994enough number of copies you must also follow the conditions in section 3.
5995.Pp
5996You may also lend copies, under the same conditions stated above, and you
5997may publicly display copies.
5998.Pp
5999.It
6000COPYING IN QUANTITY
6001.Pp
6002If you publish printed copies (or copies in media that commonly have printed
6003covers) of the Document, numbering more than 100, and the Document's license
6004notice requires Cover Texts, you must enclose the copies in covers that carry,
6005clearly and legibly, all these Cover Texts: Front-Cover Texts on the front
6006cover, and Back-Cover Texts on the back cover. Both covers must also clearly
6007and legibly identify you as the publisher of these copies. The front cover
6008must present the full title with all words of the title equally prominent
6009and visible. You may add other material on the covers in addition. Copying
6010with changes limited to the covers, as long as they preserve the title of
6011the Document and satisfy these conditions, can be treated as verbatim copying
6012in other respects.
6013.Pp
6014If the required texts for either cover are too voluminous to fit legibly,
6015you should put the first ones listed (as many as fit reasonably) on the actual
6016cover, and continue the rest onto adjacent pages.
6017.Pp
6018If you publish or distribute Opaque copies of the Document numbering more
6019than 100, you must either include a machine-readable Transparent copy along
6020with each Opaque copy, or state in or with each Opaque copy a computer-network
6021location from which the general network-using public has access to download
6022using public-standard network protocols a complete Transparent copy of the
6023Document, free of added material. If you use the latter option, you must take
6024reasonably prudent steps, when you begin distribution of Opaque copies in
6025quantity, to ensure that this Transparent copy will remain thus accessible
6026at the stated location until at least one year after the last time you distribute
6027an Opaque copy (directly or through your agents or retailers) of that edition
6028to the public.
6029.Pp
6030It is requested, but not required, that you contact the authors of the Document
6031well before redistributing any large number of copies, to give them a chance
6032to provide you with an updated version of the Document.
6033.Pp
6034.It
6035MODIFICATIONS
6036.Pp
6037You may copy and distribute a Modified Version of the Document under the conditions
6038of sections 2 and 3 above, provided that you release the Modified Version
6039under precisely this License, with the Modified Version filling the role of
6040the Document, thus licensing distribution and modification of the Modified
6041Version to whoever possesses a copy of it. In addition, you must do these
6042things in the Modified Version:
6043.Pp
6044.Bl -enum
6045.It
6046Use in the Title Page (and on the covers, if any) a title distinct from that
6047of the Document, and from those of previous versions (which should, if there
6048were any, be listed in the History section of the Document). You may use the
6049same title as a previous version if the original publisher of that version
6050gives permission.
6051.Pp
6052.It
6053List on the Title Page, as authors, one or more persons or entities responsible
6054for authorship of the modifications in the Modified Version, together with
6055at least five of the principal authors of the Document (all of its principal
6056authors, if it has fewer than five), unless they release you from this requirement.
6057.Pp
6058.It
6059State on the Title page the name of the publisher of the Modified Version,
6060as the publisher.
6061.Pp
6062.It
6063Preserve all the copyright notices of the Document.
6064.Pp
6065.It
6066Add an appropriate copyright notice for your modifications adjacent to the
6067other copyright notices.
6068.Pp
6069.It
6070Include, immediately after the copyright notices, a license notice giving
6071the public permission to use the Modified Version under the terms of this
6072License, in the form shown in the Addendum below.
6073.Pp
6074.It
6075Preserve in that license notice the full lists of Invariant Sections and required
6076Cover Texts given in the Document's license notice.
6077.Pp
6078.It
6079Include an unaltered copy of this License.
6080.Pp
6081.It
6082Preserve the section Entitled \(lqHistory\(rq, Preserve its Title, and add to it an
6083item stating at least the title, year, new authors, and publisher of the Modified
6084Version as given on the Title Page. If there is no section Entitled \(lqHistory\(rq
6085in the Document, create one stating the title, year, authors, and publisher
6086of the Document as given on its Title Page, then add an item describing the
6087Modified Version as stated in the previous sentence.
6088.Pp
6089.It
6090Preserve the network location, if any, given in the Document for public access
6091to a Transparent copy of the Document, and likewise the network locations
6092given in the Document for previous versions it was based on. These may be
6093placed in the \(lqHistory\(rq section. You may omit a network location for a work
6094that was published at least four years before the Document itself, or if the
6095original publisher of the version it refers to gives permission.
6096.Pp
6097.It
6098For any section Entitled \(lqAcknowledgements\(rq or \(lqDedications\(rq, Preserve the Title
6099of the section, and preserve in the section all the substance and tone of
6100each of the contributor acknowledgements and/or dedications given therein.
6101.Pp
6102.It
6103Preserve all the Invariant Sections of the Document, unaltered in their text
6104and in their titles. Section numbers or the equivalent are not considered
6105part of the section titles.
6106.Pp
6107.It
6108Delete any section Entitled \(lqEndorsements\(rq. Such a section may not be included
6109in the Modified Version.
6110.Pp
6111.It
6112Do not retitle any existing section to be Entitled \(lqEndorsements\(rq or to conflict
6113in title with any Invariant Section.
6114.Pp
6115.It
6116Preserve any Warranty Disclaimers.
6117.El
6118.Pp
6119If the Modified Version includes new front-matter sections or appendices that
6120qualify as Secondary Sections and contain no material copied from the Document,
6121you may at your option designate some or all of these sections as invariant.
6122To do this, add their titles to the list of Invariant Sections in the Modified
6123Version's license notice. These titles must be distinct from any other section
6124titles.
6125.Pp
6126You may add a section Entitled \(lqEndorsements\(rq, provided it contains nothing
6127but endorsements of your Modified Version by various parties---for example,
6128statements of peer review or that the text has been approved by an organization
6129as the authoritative definition of a standard.
6130.Pp
6131You may add a passage of up to five words as a Front-Cover Text, and a passage
6132of up to 25 words as a Back-Cover Text, to the end of the list of Cover Texts
6133in the Modified Version. Only one passage of Front-Cover Text and one of Back-Cover
6134Text may be added by (or through arrangements made by) any one entity. If
6135the Document already includes a cover text for the same cover, previously
6136added by you or by arrangement made by the same entity you are acting on behalf
6137of, you may not add another; but you may replace the old one, on explicit
6138permission from the previous publisher that added the old one.
6139.Pp
6140The author(s) and publisher(s) of the Document do not by this License give
6141permission to use their names for publicity for or to assert or imply endorsement
6142of any Modified Version.
6143.Pp
6144.It
6145COMBINING DOCUMENTS
6146.Pp
6147You may combine the Document with other documents released under this License,
6148under the terms defined in section 4 above for modified versions, provided
6149that you include in the combination all of the Invariant Sections of all of
6150the original documents, unmodified, and list them all as Invariant Sections
6151of your combined work in its license notice, and that you preserve all their
6152Warranty Disclaimers.
6153.Pp
6154The combined work need only contain one copy of this License, and multiple
6155identical Invariant Sections may be replaced with a single copy. If there
6156are multiple Invariant Sections with the same name but different contents,
6157make the title of each such section unique by adding at the end of it, in
6158parentheses, the name of the original author or publisher of that section
6159if known, or else a unique number. Make the same adjustment to the section
6160titles in the list of Invariant Sections in the license notice of the combined
6161work.
6162.Pp
6163In the combination, you must combine any sections Entitled \(lqHistory\(rq in the
6164various original documents, forming one section Entitled \(lqHistory\(rq; likewise
6165combine any sections Entitled \(lqAcknowledgements\(rq, and any sections Entitled
6166\(lqDedications\(rq. You must delete all sections Entitled \(lqEndorsements.\(rq
6167.Pp
6168.It
6169COLLECTIONS OF DOCUMENTS
6170.Pp
6171You may make a collection consisting of the Document and other documents released
6172under this License, and replace the individual copies of this License in the
6173various documents with a single copy that is included in the collection, provided
6174that you follow the rules of this License for verbatim copying of each of
6175the documents in all other respects.
6176.Pp
6177You may extract a single document from such a collection, and distribute it
6178individually under this License, provided you insert a copy of this License
6179into the extracted document, and follow this License in all other respects
6180regarding verbatim copying of that document.
6181.Pp
6182.It
6183AGGREGATION WITH INDEPENDENT WORKS
6184.Pp
6185A compilation of the Document or its derivatives with other separate and independent
6186documents or works, in or on a volume of a storage or distribution medium,
6187is called an \(lqaggregate\(rq if the copyright resulting from the compilation is
6188not used to limit the legal rights of the compilation's users beyond what
6189the individual works permit. When the Document is included in an aggregate,
6190this License does not apply to the other works in the aggregate which are
6191not themselves derivative works of the Document.
6192.Pp
6193If the Cover Text requirement of section 3 is applicable to these copies of
6194the Document, then if the Document is less than one half of the entire aggregate,
6195the Document's Cover Texts may be placed on covers that bracket the Document
6196within the aggregate, or the electronic equivalent of covers if the Document
6197is in electronic form. Otherwise they must appear on printed covers that bracket
6198the whole aggregate.
6199.Pp
6200.It
6201TRANSLATION
6202.Pp
6203Translation is considered a kind of modification, so you may distribute translations
6204of the Document under the terms of section 4. Replacing Invariant Sections
6205with translations requires special permission from their copyright holders,
6206but you may include translations of some or all Invariant Sections in addition
6207to the original versions of these Invariant Sections. You may include a translation
6208of this License, and all the license notices in the Document, and any Warranty
6209Disclaimers, provided that you also include the original English version of
6210this License and the original versions of those notices and disclaimers. In
6211case of a disagreement between the translation and the original version of
6212this License or a notice or disclaimer, the original version will prevail.
6213.Pp
6214If a section in the Document is Entitled \(lqAcknowledgements\(rq, \(lqDedications\(rq, or
6215\(lqHistory\(rq, the requirement (section 4) to Preserve its Title (section 1) will
6216typically require changing the actual title.
6217.Pp
6218.It
6219TERMINATION
6220.Pp
6221You may not copy, modify, sublicense, or distribute the Document except as
6222expressly provided for under this License. Any other attempt to copy, modify,
6223sublicense or distribute the Document is void, and will automatically terminate
6224your rights under this License. However, parties who have received copies,
6225or rights, from you under this License will not have their licenses terminated
6226so long as such parties remain in full compliance.
6227.Pp
6228.It
6229FUTURE REVISIONS OF THIS LICENSE
6230.Pp
6231The Free Software Foundation may publish new, revised versions of the GNU
6232Free Documentation License from time to time. Such new versions will be similar
6233in spirit to the present version, but may differ in detail to address new
6234problems or concerns. See
6235.Lk http://www.gnu.org/copyleft/ .
6236.Pp
6237Each version of the License is given a distinguishing version number. If the
6238Document specifies that a particular numbered version of this License \(lqor any
6239later version\(rq applies to it, you have the option of following the terms and
6240conditions either of that specified version or of any later version that has
6241been published (not as a draft) by the Free Software Foundation. If the Document
6242does not specify a version number of this License, you may choose any version
6243ever published (not as a draft) by the Free Software Foundation.
6244.El
6245.Pp
6246.Em  ADDENDUM: How to use this License for your documents
6247.Pp
6248To use this License in a document you have written, include a copy of the
6249License in the document and put the following copyright and license notices
6250just after the title page:
6251.Pp
6252.Bd -literal -offset indent
6253
6254  Copyright (C)  year  your name.
6255  Permission is granted to copy, distribute and/or modify this document
6256  under the terms of the GNU Free Documentation License, Version 1.2
6257  or any later version published by the Free Software Foundation;
6258  with no Invariant Sections, no Front-Cover Texts, and no Back-Cover
6259  Texts.  A copy of the license is included in the section entitled \(lqGNU
6260  Free Documentation License\(rq.
6261
6262.Ed
6263.Pp
6264If you have Invariant Sections, Front-Cover Texts and Back-Cover Texts, replace
6265the \(lqwith...Texts.\(rq line with this:
6266.Pp
6267.Bd -literal -offset indent
6268
6269    with the Invariant Sections being list their titles, with
6270    the Front-Cover Texts being list, and with the Back-Cover Texts
6271    being list.
6272
6273.Ed
6274.Pp
6275If you have Invariant Sections without Cover Texts, or some other combination
6276of the three, merge those two alternatives to suit the situation.
6277.Pp
6278If your document contains nontrivial examples of program code, we recommend
6279releasing these examples in parallel under your choice of free software license,
6280such as the GNU General Public License, to permit their use in free software.
6281.Pp
6282.Sh  Translations of This Manual
6283Nishio Futoshi of the GNUjdoc project has prepared a Japanese translation
6284of this manual. Its most recent version can be found at
6285.Lk http://openlab.ring.gr.jp/gnujdoc/cvsweb/cvsweb.cgi/gnujdoc/ .
6286.Pp
6287.Sh  Index
6288