1 /* od -- dump files in octal and other formats
2 Copyright (C) 1992-2012 Free Software Foundation, Inc.
4 This program is free software: you can redistribute it and/or modify
5 it under the terms of the GNU General Public License as published by
6 the Free Software Foundation, either version 3 of the License, or
7 (at your option) any later version.
9 This program is distributed in the hope that it will be useful,
10 but WITHOUT ANY WARRANTY; without even the implied warranty of
11 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 GNU General Public License for more details.
14 You should have received a copy of the GNU General Public License
15 along with this program. If not, see <http://www.gnu.org/licenses/>. */
17 /* Written by Jim Meyering. */
24 #include <sys/types.h>
33 /* The official name of this program (e.g., no 'g' prefix). */
34 #define PROGRAM_NAME "od"
36 #define AUTHORS proper_name ("Jim Meyering")
38 /* The default number of input bytes per output line. */
39 #define DEFAULT_BYTES_PER_BLOCK 16
41 #if HAVE_UNSIGNED_LONG_LONG_INT
42 typedef unsigned long long int unsigned_long_long_int
;
44 /* This is just a place-holder to avoid a few '#if' directives.
45 In this case, the type isn't actually used. */
46 typedef unsigned long int unsigned_long_long_int
;
57 /* FIXME: add INTMAX support, too */
75 #define MAX_INTEGRAL_TYPE_SIZE sizeof (unsigned_long_long_int)
77 /* The maximum number of bytes needed for a format string, including
78 the trailing nul. Each format string expects a variable amount of
79 padding (guaranteed to be at least 1 plus the field width), then an
80 element that will be formatted in the field. */
92 /* Ensure that our choice for FMT_BYTES_ALLOCATED is reasonable. */
93 verify (MAX_INTEGRAL_TYPE_SIZE
* CHAR_BIT
/ 3 <= 99);
95 /* Each output format specification (from '-t spec' or from
96 old-style options) is represented by one of these structures. */
99 enum output_format fmt
;
100 enum size_spec size
; /* Type of input object. */
101 /* FIELDS is the number of fields per line, BLANK is the number of
102 fields to leave blank. WIDTH is width of one field, excluding
103 leading space, and PAD is total pad to divide among FIELDS.
104 PAD is at least as large as FIELDS. */
105 void (*print_function
) (size_t fields
, size_t blank
, void const *data
,
106 char const *fmt
, int width
, int pad
);
107 char fmt_string
[FMT_BYTES_ALLOCATED
]; /* Of the style "%*d". */
108 bool hexl_mode_trailer
;
109 int field_width
; /* Minimum width of a field, excluding leading space. */
110 int pad_width
; /* Total padding to be divided among fields. */
113 /* Convert the number of 8-bit bytes of a binary representation to
114 the number of characters (digits + sign if the type is signed)
115 required to represent the same quantity in the specified base/type.
116 For example, a 32-bit (4-byte) quantity may require a field width
117 as wide as the following for these types:
121 8 unsigned hexadecimal */
123 static unsigned int const bytes_to_oct_digits
[] =
124 {0, 3, 6, 8, 11, 14, 16, 19, 22, 25, 27, 30, 32, 35, 38, 41, 43};
126 static unsigned int const bytes_to_signed_dec_digits
[] =
127 {1, 4, 6, 8, 11, 13, 16, 18, 20, 23, 25, 28, 30, 33, 35, 37, 40};
129 static unsigned int const bytes_to_unsigned_dec_digits
[] =
130 {0, 3, 5, 8, 10, 13, 15, 17, 20, 22, 25, 27, 29, 32, 34, 37, 39};
132 static unsigned int const bytes_to_hex_digits
[] =
133 {0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32};
135 /* It'll be a while before we see integral types wider than 16 bytes,
136 but if/when it happens, this check will catch it. Without this check,
137 a wider type would provoke a buffer overrun. */
138 verify (MAX_INTEGRAL_TYPE_SIZE
< ARRAY_CARDINALITY (bytes_to_hex_digits
));
140 /* Make sure the other arrays have the same length. */
141 verify (sizeof bytes_to_oct_digits
== sizeof bytes_to_signed_dec_digits
);
142 verify (sizeof bytes_to_oct_digits
== sizeof bytes_to_unsigned_dec_digits
);
143 verify (sizeof bytes_to_oct_digits
== sizeof bytes_to_hex_digits
);
145 /* Convert enum size_spec to the size of the named type. */
146 static const int width_bytes
[] =
153 sizeof (unsigned_long_long_int
),
159 /* Ensure that for each member of 'enum size_spec' there is an
160 initializer in the width_bytes array. */
161 verify (ARRAY_CARDINALITY (width_bytes
) == N_SIZE_SPECS
);
163 /* Names for some non-printing characters. */
164 static char const charname
[33][4] =
166 "nul", "soh", "stx", "etx", "eot", "enq", "ack", "bel",
167 "bs", "ht", "nl", "vt", "ff", "cr", "so", "si",
168 "dle", "dc1", "dc2", "dc3", "dc4", "nak", "syn", "etb",
169 "can", "em", "sub", "esc", "fs", "gs", "rs", "us",
173 /* Address base (8, 10 or 16). */
174 static int address_base
;
176 /* The number of octal digits required to represent the largest
178 #define MAX_ADDRESS_LENGTH \
179 ((sizeof (uintmax_t) * CHAR_BIT + CHAR_BIT - 1) / 3)
181 /* Width of a normal address. */
182 static int address_pad_len
;
184 /* Minimum length when detecting --strings. */
185 static size_t string_min
;
187 /* True when in --strings mode. */
188 static bool flag_dump_strings
;
190 /* True if we should recognize the older non-option arguments
191 that specified at most one file and optional arguments specifying
192 offset and pseudo-start address. */
193 static bool traditional
;
195 /* True if an old-style 'pseudo-address' was specified. */
196 static bool flag_pseudo_start
;
198 /* The difference between the old-style pseudo starting address and
199 the number of bytes to skip. */
200 static uintmax_t pseudo_offset
;
202 /* Function that accepts an address and an optional following char,
203 and prints the address and char to stdout. */
204 static void (*format_address
) (uintmax_t, char);
206 /* The number of input bytes to skip before formatting and writing. */
207 static uintmax_t n_bytes_to_skip
= 0;
209 /* When false, MAX_BYTES_TO_FORMAT and END_OFFSET are ignored, and all
210 input is formatted. */
211 static bool limit_bytes_to_format
= false;
213 /* The maximum number of bytes that will be formatted. */
214 static uintmax_t max_bytes_to_format
;
216 /* The offset of the first byte after the last byte to be formatted. */
217 static uintmax_t end_offset
;
219 /* When true and two or more consecutive blocks are equal, format
220 only the first block and output an asterisk alone on the following
221 line to indicate that identical blocks have been elided. */
222 static bool abbreviate_duplicate_blocks
= true;
224 /* An array of specs describing how to format each input block. */
225 static struct tspec
*spec
;
227 /* The number of format specs. */
228 static size_t n_specs
;
230 /* The allocated length of SPEC. */
231 static size_t n_specs_allocated
;
233 /* The number of input bytes formatted per output line. It must be
234 a multiple of the least common multiple of the sizes associated with
235 the specified output types. It should be as large as possible, but
236 no larger than 16 -- unless specified with the -w option. */
237 static size_t bytes_per_block
;
239 /* Human-readable representation of *file_list (for error messages).
240 It differs from file_list[-1] only when file_list[-1] is "-". */
241 static char const *input_filename
;
243 /* A NULL-terminated list of the file-arguments from the command line. */
244 static char const *const *file_list
;
246 /* Initializer for file_list if no file-arguments
247 were specified on the command line. */
248 static char const *const default_file_list
[] = {"-", NULL
};
250 /* The input stream associated with the current file. */
251 static FILE *in_stream
;
253 /* If true, at least one of the files we read was standard input. */
254 static bool have_read_stdin
;
256 /* Map the size in bytes to a type identifier. */
257 static enum size_spec integral_type_size
[MAX_INTEGRAL_TYPE_SIZE
+ 1];
259 #define MAX_FP_TYPE_SIZE sizeof (long double)
260 static enum size_spec fp_type_size
[MAX_FP_TYPE_SIZE
+ 1];
262 static char const short_options
[] = "A:aBbcDdeFfHhIij:LlN:OoS:st:vw::Xx";
264 /* For long options that have no equivalent short option, use a
265 non-character as a pseudo short option, starting with CHAR_MAX + 1. */
268 TRADITIONAL_OPTION
= CHAR_MAX
+ 1
271 static struct option
const long_options
[] =
273 {"skip-bytes", required_argument
, NULL
, 'j'},
274 {"address-radix", required_argument
, NULL
, 'A'},
275 {"read-bytes", required_argument
, NULL
, 'N'},
276 {"format", required_argument
, NULL
, 't'},
277 {"output-duplicates", no_argument
, NULL
, 'v'},
278 {"strings", optional_argument
, NULL
, 'S'},
279 {"traditional", no_argument
, NULL
, TRADITIONAL_OPTION
},
280 {"width", optional_argument
, NULL
, 'w'},
282 {GETOPT_HELP_OPTION_DECL
},
283 {GETOPT_VERSION_OPTION_DECL
},
290 if (status
!= EXIT_SUCCESS
)
295 Usage: %s [OPTION]... [FILE]...\n\
296 or: %s [-abcdfilosx]... [FILE] [[+]OFFSET[.][b]]\n\
297 or: %s --traditional [OPTION]... [FILE] [[+]OFFSET[.][b] [+][LABEL][.][b]]\n\
299 program_name
, program_name
, program_name
);
301 Write an unambiguous representation, octal bytes by default,\n\
302 of FILE to standard output. With more than one FILE argument,\n\
303 concatenate them in the listed order to form the input.\n\
304 With no FILE, or when FILE is -, read standard input.\n\
308 All arguments to long options are mandatory for short options.\n\
311 -A, --address-radix=RADIX decide how file offsets are printed\n\
312 -j, --skip-bytes=BYTES skip BYTES input bytes first\n\
315 -N, --read-bytes=BYTES limit dump to BYTES input bytes\n\
316 -S BYTES, --strings[=BYTES] output strings of at least BYTES graphic chars\n\
317 -t, --format=TYPE select output format or formats\n\
318 -v, --output-duplicates do not use * to mark line suppression\n\
319 -w[BYTES], --width[=BYTES] output BYTES bytes per output line\n\
320 --traditional accept arguments in traditional form\n\
322 fputs (HELP_OPTION_DESCRIPTION
, stdout
);
323 fputs (VERSION_OPTION_DESCRIPTION
, stdout
);
326 Traditional format specifications may be intermixed; they accumulate:\n\
327 -a same as -t a, select named characters, ignoring high-order bit\n\
328 -b same as -t o1, select octal bytes\n\
329 -c same as -t c, select ASCII characters or backslash escapes\n\
330 -d same as -t u2, select unsigned decimal 2-byte units\n\
333 -f same as -t fF, select floats\n\
334 -i same as -t dI, select decimal ints\n\
335 -l same as -t dL, select decimal longs\n\
336 -o same as -t o2, select octal 2-byte units\n\
337 -s same as -t d2, select decimal 2-byte units\n\
338 -x same as -t x2, select hexadecimal 2-byte units\n\
342 If first and second call formats both apply, the second format is assumed\n\
343 if the last operand begins with + or (if there are 2 operands) a digit.\n\
344 An OFFSET operand means -j OFFSET. LABEL is the pseudo-address\n\
345 at first byte printed, incremented when dump is progressing.\n\
346 For OFFSET and LABEL, a 0x or 0X prefix indicates hexadecimal;\n\
347 suffixes may be . for octal and b for multiply by 512.\n\
351 TYPE is made up of one or more of these specifications:\n\
353 a named character, ignoring high-order bit\n\
354 c ASCII character or backslash escape\n\
357 d[SIZE] signed decimal, SIZE bytes per integer\n\
358 f[SIZE] floating point, SIZE bytes per integer\n\
359 o[SIZE] octal, SIZE bytes per integer\n\
360 u[SIZE] unsigned decimal, SIZE bytes per integer\n\
361 x[SIZE] hexadecimal, SIZE bytes per integer\n\
365 SIZE is a number. For TYPE in doux, SIZE may also be C for\n\
366 sizeof(char), S for sizeof(short), I for sizeof(int) or L for\n\
367 sizeof(long). If TYPE is f, SIZE may also be F for sizeof(float), D\n\
368 for sizeof(double) or L for sizeof(long double).\n\
372 RADIX is d for decimal, o for octal, x for hexadecimal or n for none.\n\
373 BYTES is hexadecimal with 0x or 0X prefix, and may have a multiplier suffix:\n\
374 b 512, kB 1000, K 1024, MB 1000*1000, M 1024*1024,\n\
375 GB 1000*1000*1000, G 1024*1024*1024, and so on for T, P, E, Z, Y.\n\
376 Adding a z suffix to any type displays printable characters at the end of each\
381 Option --string without a number implies 3; option --width without a number\n\
382 implies 32. By default, od uses -A o -t oS -w16.\n\
384 emit_ancillary_info ();
389 /* Define the print functions. */
391 #define PRINT_FIELDS(N, T, FMT_STRING, ACTION) \
393 N (size_t fields, size_t blank, void const *block, \
394 char const *FMT_STRING, int width, int pad) \
396 T const *p = block; \
398 int pad_remaining = pad; \
399 for (i = fields; blank < i; i--) \
401 int next_pad = pad * (i - 1) / fields; \
402 int adjusted_width = pad_remaining - next_pad + width; \
405 pad_remaining = next_pad; \
409 #define PRINT_TYPE(N, T) \
410 PRINT_FIELDS (N, T, fmt_string, xprintf (fmt_string, adjusted_width, x))
412 #define PRINT_FLOATTYPE(N, T, FTOASTR, BUFSIZE) \
413 PRINT_FIELDS (N, T, fmt_string ATTRIBUTE_UNUSED, \
415 FTOASTR (buf, sizeof buf, 0, 0, x); \
416 xprintf ("%*s", adjusted_width, buf))
418 PRINT_TYPE (print_s_char
, signed char)
419 PRINT_TYPE (print_char
, unsigned char)
420 PRINT_TYPE (print_s_short
, short int)
421 PRINT_TYPE (print_short
, unsigned short int)
422 PRINT_TYPE (print_int
, unsigned int)
423 PRINT_TYPE (print_long
, unsigned long int)
424 PRINT_TYPE (print_long_long
, unsigned_long_long_int
)
426 PRINT_FLOATTYPE (print_float
, float, ftoastr
, FLT_BUFSIZE_BOUND
)
427 PRINT_FLOATTYPE (print_double
, double, dtoastr
, DBL_BUFSIZE_BOUND
)
428 PRINT_FLOATTYPE (print_long_double
, long double, ldtoastr
, LDBL_BUFSIZE_BOUND
)
431 #undef PRINT_FLOATTYPE
434 dump_hexl_mode_trailer (size_t n_bytes
, const char *block
)
437 fputs (" >", stdout
);
438 for (i
= n_bytes
; i
> 0; i
--)
440 unsigned char c
= *block
++;
441 unsigned char c2
= (isprint (c
) ? c
: '.');
448 print_named_ascii (size_t fields
, size_t blank
, void const *block
,
449 const char *unused_fmt_string ATTRIBUTE_UNUSED
,
452 unsigned char const *p
= block
;
454 int pad_remaining
= pad
;
455 for (i
= fields
; blank
< i
; i
--)
457 int next_pad
= pad
* (i
- 1) / fields
;
458 int masked_c
= *p
++ & 0x7f;
464 else if (masked_c
<= 040)
465 s
= charname
[masked_c
];
473 xprintf ("%*s", pad_remaining
- next_pad
+ width
, s
);
474 pad_remaining
= next_pad
;
479 print_ascii (size_t fields
, size_t blank
, void const *block
,
480 const char *unused_fmt_string ATTRIBUTE_UNUSED
, int width
,
483 unsigned char const *p
= block
;
485 int pad_remaining
= pad
;
486 for (i
= fields
; blank
< i
; i
--)
488 int next_pad
= pad
* (i
- 1) / fields
;
489 unsigned char c
= *p
++;
528 sprintf (buf
, (isprint (c
) ? "%c" : "%03o"), c
);
532 xprintf ("%*s", pad_remaining
- next_pad
+ width
, s
);
533 pad_remaining
= next_pad
;
537 /* Convert a null-terminated (possibly zero-length) string S to an
538 unsigned long integer value. If S points to a non-digit set *P to S,
539 *VAL to 0, and return true. Otherwise, accumulate the integer value of
540 the string of digits. If the string of digits represents a value
541 larger than ULONG_MAX, don't modify *VAL or *P and return false.
542 Otherwise, advance *P to the first non-digit after S, set *VAL to
543 the result of the conversion and return true. */
546 simple_strtoul (const char *s
, const char **p
, unsigned long int *val
)
548 unsigned long int sum
;
554 if (sum
> (ULONG_MAX
- c
) / 10)
563 /* If S points to a single valid modern od format string, put
564 a description of that format in *TSPEC, make *NEXT point at the
565 character following the just-decoded format (if *NEXT is non-NULL),
566 and return true. If S is not valid, don't modify *NEXT or *TSPEC,
567 give a diagnostic, and return false. For example, if S were
568 "d4afL" *NEXT would be set to "afL" and *TSPEC would be
570 fmt = SIGNED_DECIMAL;
571 size = INT or LONG; (whichever integral_type_size[4] resolves to)
572 print_function = print_int; (assuming size == INT)
576 pad_width is determined later, but is at least as large as the
577 number of fields printed per row.
578 S_ORIG is solely for reporting errors. It should be the full format
583 decode_one_format (const char *s_orig
, const char *s
, const char **next
,
586 enum size_spec size_spec
;
587 unsigned long int size
;
588 enum output_format fmt
;
589 void (*print_function
) (size_t, size_t, void const *, char const *,
595 assert (tspec
!= NULL
);
609 size
= sizeof (char);
614 size
= sizeof (short int);
624 size
= sizeof (long int);
628 if (! simple_strtoul (s
, &p
, &size
))
630 /* The integer at P in S would overflow an unsigned long int.
631 A digit string that long is sufficiently odd looking
632 that the following diagnostic is sufficient. */
633 error (0, 0, _("invalid type string %s"), quote (s_orig
));
640 if (MAX_INTEGRAL_TYPE_SIZE
< size
641 || integral_type_size
[size
] == NO_SIZE
)
643 error (0, 0, _("invalid type string %s;\nthis system"
644 " doesn't provide a %lu-byte integral type"),
645 quote (s_orig
), size
);
653 #define ISPEC_TO_FORMAT(Spec, Min_format, Long_format, Max_format) \
654 ((Spec) == LONG_LONG ? (Max_format) \
655 : ((Spec) == LONG ? (Long_format) \
658 size_spec = integral_type_size[size];
663 fmt
= SIGNED_DECIMAL
;
664 field_width
= bytes_to_signed_dec_digits
[size
];
665 sprintf (tspec
->fmt_string
, "%%*%s",
666 ISPEC_TO_FORMAT (size_spec
, "d", "ld", PRIdMAX
));
671 sprintf (tspec
->fmt_string
, "%%*.%d%s",
672 (field_width
= bytes_to_oct_digits
[size
]),
673 ISPEC_TO_FORMAT (size_spec
, "o", "lo", PRIoMAX
));
677 fmt
= UNSIGNED_DECIMAL
;
678 field_width
= bytes_to_unsigned_dec_digits
[size
];
679 sprintf (tspec
->fmt_string
, "%%*%s",
680 ISPEC_TO_FORMAT (size_spec
, "u", "lu", PRIuMAX
));
685 sprintf (tspec
->fmt_string
, "%%*.%d%s",
686 (field_width
= bytes_to_hex_digits
[size
]),
687 ISPEC_TO_FORMAT (size_spec
, "x", "lx", PRIxMAX
));
694 assert (strlen (tspec
->fmt_string
) < FMT_BYTES_ALLOCATED
);
699 print_function
= (fmt
== SIGNED_DECIMAL
705 print_function
= (fmt
== SIGNED_DECIMAL
711 print_function
= print_int
;
715 print_function
= print_long
;
719 print_function
= print_long_long
;
728 fmt
= FLOATING_POINT
;
734 size
= sizeof (float);
739 size
= sizeof (double);
744 size
= sizeof (long double);
748 if (! simple_strtoul (s
, &p
, &size
))
750 /* The integer at P in S would overflow an unsigned long int.
751 A digit string that long is sufficiently odd looking
752 that the following diagnostic is sufficient. */
753 error (0, 0, _("invalid type string %s"), quote (s_orig
));
757 size
= sizeof (double);
760 if (size
> MAX_FP_TYPE_SIZE
761 || fp_type_size
[size
] == NO_SIZE
)
764 _("invalid type string %s;\n"
765 "this system doesn't provide a %lu-byte"
766 " floating point type"),
767 quote (s_orig
), size
);
774 size_spec
= fp_type_size
[size
];
777 struct lconv
const *locale
= localeconv ();
778 size_t decimal_point_len
=
779 (locale
->decimal_point
[0] ? strlen (locale
->decimal_point
) : 1);
784 print_function
= print_float
;
785 field_width
= FLT_STRLEN_BOUND_L (decimal_point_len
);
789 print_function
= print_double
;
790 field_width
= DBL_STRLEN_BOUND_L (decimal_point_len
);
793 case FLOAT_LONG_DOUBLE
:
794 print_function
= print_long_double
;
795 field_width
= LDBL_STRLEN_BOUND_L (decimal_point_len
);
807 fmt
= NAMED_CHARACTER
;
809 print_function
= print_named_ascii
;
817 print_function
= print_ascii
;
822 error (0, 0, _("invalid character '%c' in type string %s"),
827 tspec
->size
= size_spec
;
829 tspec
->print_function
= print_function
;
831 tspec
->field_width
= field_width
;
832 tspec
->hexl_mode_trailer
= (*s
== 'z');
833 if (tspec
->hexl_mode_trailer
)
842 /* Given a list of one or more input filenames FILE_LIST, set the global
843 file pointer IN_STREAM and the global string INPUT_FILENAME to the
844 first one that can be successfully opened. Modify FILE_LIST to
845 reference the next filename in the list. A file name of "-" is
846 interpreted as standard input. If any file open fails, give an error
847 message and return false. */
850 open_next_file (void)
856 input_filename
= *file_list
;
857 if (input_filename
== NULL
)
861 if (STREQ (input_filename
, "-"))
863 input_filename
= _("standard input");
865 have_read_stdin
= true;
866 if (O_BINARY
&& ! isatty (STDIN_FILENO
))
867 xfreopen (NULL
, "rb", stdin
);
871 in_stream
= fopen (input_filename
, (O_BINARY
? "rb" : "r"));
872 if (in_stream
== NULL
)
874 error (0, errno
, "%s", input_filename
);
879 while (in_stream
== NULL
);
881 if (limit_bytes_to_format
&& !flag_dump_strings
)
882 setvbuf (in_stream
, NULL
, _IONBF
, 0);
887 /* Test whether there have been errors on in_stream, and close it if
888 it is not standard input. Return false if there has been an error
889 on in_stream or stdout; return true otherwise. This function will
890 report more than one error only if both a read and a write error
891 have occurred. IN_ERRNO, if nonzero, is the error number
892 corresponding to the most recent action for IN_STREAM. */
895 check_and_close (int in_errno
)
899 if (in_stream
!= NULL
)
901 if (ferror (in_stream
))
903 error (0, in_errno
, _("%s: read error"), input_filename
);
904 if (! STREQ (file_list
[-1], "-"))
908 else if (! STREQ (file_list
[-1], "-") && fclose (in_stream
) != 0)
910 error (0, errno
, "%s", input_filename
);
919 error (0, 0, _("write error"));
926 /* Decode the modern od format string S. Append the decoded
927 representation to the global array SPEC, reallocating SPEC if
928 necessary. Return true if S is valid. */
931 decode_format_string (const char *s
)
933 const char *s_orig
= s
;
940 if (n_specs_allocated
<= n_specs
)
941 spec
= X2NREALLOC (spec
, &n_specs_allocated
);
943 if (! decode_one_format (s_orig
, s
, &next
, &spec
[n_specs
]))
954 /* Given a list of one or more input filenames FILE_LIST, set the global
955 file pointer IN_STREAM to position N_SKIP in the concatenation of
956 those files. If any file operation fails or if there are fewer than
957 N_SKIP bytes in the combined input, give an error message and return
958 false. When possible, use seek rather than read operations to
959 advance IN_STREAM. */
962 skip (uintmax_t n_skip
)
970 while (in_stream
!= NULL
) /* EOF. */
972 struct stat file_stats
;
974 /* First try seeking. For large offsets, this extra work is
975 worthwhile. If the offset is below some threshold it may be
976 more efficient to move the pointer by reading. There are two
977 issues when trying to seek:
978 - the file must be seekable.
979 - before seeking to the specified position, make sure
980 that the new position is in the current file.
981 Try to do that by getting file's size using fstat.
982 But that will work only for regular files. */
984 if (fstat (fileno (in_stream
), &file_stats
) == 0)
986 /* The st_size field is valid for regular files.
987 If the number of bytes left to skip is larger than
988 the size of the current file, we can decrement n_skip
989 and go on to the next file. Skip this optimization also
990 when st_size is 0, because some kernels report that
991 nonempty files in /proc have st_size == 0. */
992 if (S_ISREG (file_stats
.st_mode
) && 0 < file_stats
.st_size
)
994 if ((uintmax_t) file_stats
.st_size
< n_skip
)
995 n_skip
-= file_stats
.st_size
;
998 if (fseeko (in_stream
, n_skip
, SEEK_CUR
) != 0)
1007 /* If it's not a regular file with nonnegative size,
1008 position the file pointer by reading. */
1013 size_t n_bytes_read
, n_bytes_to_read
= BUFSIZ
;
1017 if (n_skip
< n_bytes_to_read
)
1018 n_bytes_to_read
= n_skip
;
1019 n_bytes_read
= fread (buf
, 1, n_bytes_to_read
, in_stream
);
1020 n_skip
-= n_bytes_read
;
1021 if (n_bytes_read
!= n_bytes_to_read
)
1035 else /* cannot fstat() file */
1037 error (0, errno
, "%s", input_filename
);
1041 ok
&= check_and_close (in_errno
);
1043 ok
&= open_next_file ();
1047 error (EXIT_FAILURE
, 0, _("cannot skip past end of combined input"));
1053 format_address_none (uintmax_t address ATTRIBUTE_UNUSED
,
1054 char c ATTRIBUTE_UNUSED
)
1059 format_address_std (uintmax_t address
, char c
)
1061 char buf
[MAX_ADDRESS_LENGTH
+ 2];
1062 char *p
= buf
+ sizeof buf
;
1067 pbound
= p
- address_pad_len
;
1069 /* Use a special case of the code for each base. This is measurably
1070 faster than generic code. */
1071 switch (address_base
)
1075 *--p
= '0' + (address
& 7);
1076 while ((address
>>= 3) != 0);
1081 *--p
= '0' + (address
% 10);
1082 while ((address
/= 10) != 0);
1087 *--p
= "0123456789abcdef"[address
& 15];
1088 while ((address
>>= 4) != 0);
1099 format_address_paren (uintmax_t address
, char c
)
1102 format_address_std (address
, ')');
1108 format_address_label (uintmax_t address
, char c
)
1110 format_address_std (address
, ' ');
1111 format_address_paren (address
+ pseudo_offset
, c
);
1114 /* Write N_BYTES bytes from CURR_BLOCK to standard output once for each
1115 of the N_SPEC format specs. CURRENT_OFFSET is the byte address of
1116 CURR_BLOCK in the concatenation of input files, and it is printed
1117 (optionally) only before the output line associated with the first
1118 format spec. When duplicate blocks are being abbreviated, the output
1119 for a sequence of identical input blocks is the output for the first
1120 block followed by an asterisk alone on a line. It is valid to compare
1121 the blocks PREV_BLOCK and CURR_BLOCK only when N_BYTES == BYTES_PER_BLOCK.
1122 That condition may be false only for the last input block. */
1125 write_block (uintmax_t current_offset
, size_t n_bytes
,
1126 const char *prev_block
, const char *curr_block
)
1128 static bool first
= true;
1129 static bool prev_pair_equal
= false;
1131 #define EQUAL_BLOCKS(b1, b2) (memcmp (b1, b2, bytes_per_block) == 0)
1133 if (abbreviate_duplicate_blocks
1134 && !first
&& n_bytes
== bytes_per_block
1135 && EQUAL_BLOCKS (prev_block
, curr_block
))
1137 if (prev_pair_equal
)
1139 /* The two preceding blocks were equal, and the current
1140 block is the same as the last one, so print nothing. */
1145 prev_pair_equal
= true;
1152 prev_pair_equal
= false;
1153 for (i
= 0; i
< n_specs
; i
++)
1155 int datum_width
= width_bytes
[spec
[i
].size
];
1156 int fields_per_block
= bytes_per_block
/ datum_width
;
1157 int blank_fields
= (bytes_per_block
- n_bytes
) / datum_width
;
1159 format_address (current_offset
, '\0');
1161 printf ("%*s", address_pad_len
, "");
1162 (*spec
[i
].print_function
) (fields_per_block
, blank_fields
,
1163 curr_block
, spec
[i
].fmt_string
,
1164 spec
[i
].field_width
, spec
[i
].pad_width
);
1165 if (spec
[i
].hexl_mode_trailer
)
1167 /* space-pad out to full line width, then dump the trailer */
1168 int field_width
= spec
[i
].field_width
;
1169 int pad_width
= (spec
[i
].pad_width
* blank_fields
1170 / fields_per_block
);
1171 printf ("%*s", blank_fields
* field_width
+ pad_width
, "");
1172 dump_hexl_mode_trailer (n_bytes
, curr_block
);
1180 /* Read a single byte into *C from the concatenation of the input files
1181 named in the global array FILE_LIST. On the first call to this
1182 function, the global variable IN_STREAM is expected to be an open
1183 stream associated with the input file INPUT_FILENAME. If IN_STREAM
1184 is at end-of-file, close it and update the global variables IN_STREAM
1185 and INPUT_FILENAME so they correspond to the next file in the list.
1186 Then try to read a byte from the newly opened file. Repeat if
1187 necessary until EOF is reached for the last file in FILE_LIST, then
1188 set *C to EOF and return. Subsequent calls do likewise. Return
1189 true if successful. */
1198 while (in_stream
!= NULL
) /* EOF. */
1200 *c
= fgetc (in_stream
);
1205 ok
&= check_and_close (errno
);
1207 ok
&= open_next_file ();
1213 /* Read N bytes into BLOCK from the concatenation of the input files
1214 named in the global array FILE_LIST. On the first call to this
1215 function, the global variable IN_STREAM is expected to be an open
1216 stream associated with the input file INPUT_FILENAME. If all N
1217 bytes cannot be read from IN_STREAM, close IN_STREAM and update
1218 the global variables IN_STREAM and INPUT_FILENAME. Then try to
1219 read the remaining bytes from the newly opened file. Repeat if
1220 necessary until EOF is reached for the last file in FILE_LIST.
1221 On subsequent calls, don't modify BLOCK and return true. Set
1222 *N_BYTES_IN_BUFFER to the number of bytes read. If an error occurs,
1223 it will be detected through ferror when the stream is about to be
1224 closed. If there is an error, give a message but continue reading
1225 as usual and return false. Otherwise return true. */
1228 read_block (size_t n
, char *block
, size_t *n_bytes_in_buffer
)
1232 assert (0 < n
&& n
<= bytes_per_block
);
1234 *n_bytes_in_buffer
= 0;
1239 while (in_stream
!= NULL
) /* EOF. */
1244 n_needed
= n
- *n_bytes_in_buffer
;
1245 n_read
= fread (block
+ *n_bytes_in_buffer
, 1, n_needed
, in_stream
);
1247 *n_bytes_in_buffer
+= n_read
;
1249 if (n_read
== n_needed
)
1252 ok
&= check_and_close (errno
);
1254 ok
&= open_next_file ();
1260 /* Return the least common multiple of the sizes associated
1261 with the format specs. */
1263 static int _GL_ATTRIBUTE_PURE
1269 for (i
= 0; i
< n_specs
; i
++)
1270 l_c_m
= lcm (l_c_m
, width_bytes
[spec
[i
].size
]);
1274 /* If S is a valid traditional offset specification with an optional
1275 leading '+' return true and set *OFFSET to the offset it denotes. */
1278 parse_old_offset (const char *s
, uintmax_t *offset
)
1285 /* Skip over any leading '+'. */
1289 /* Determine the radix we'll use to interpret S. If there is a '.',
1290 it's decimal, otherwise, if the string begins with '0X'or '0x',
1291 it's hexadecimal, else octal. */
1292 if (strchr (s
, '.') != NULL
)
1296 if (s
[0] == '0' && (s
[1] == 'x' || s
[1] == 'X'))
1302 return xstrtoumax (s
, NULL
, radix
, offset
, "Bb") == LONGINT_OK
;
1305 /* Read a chunk of size BYTES_PER_BLOCK from the input files, write the
1306 formatted block to standard output, and repeat until the specified
1307 maximum number of bytes has been read or until all input has been
1308 processed. If the last block read is smaller than BYTES_PER_BLOCK
1309 and its size is not a multiple of the size associated with a format
1310 spec, extend the input block with zero bytes until its length is a
1311 multiple of all format spec sizes. Write the final block. Finally,
1312 write on a line by itself the offset of the byte after the last byte
1313 read. Accumulate return values from calls to read_block and
1314 check_and_close, and if any was false, return false.
1315 Otherwise, return true. */
1321 uintmax_t current_offset
;
1324 size_t n_bytes_read
;
1326 block
[0] = xnmalloc (2, bytes_per_block
);
1327 block
[1] = block
[0] + bytes_per_block
;
1329 current_offset
= n_bytes_to_skip
;
1331 if (limit_bytes_to_format
)
1336 if (current_offset
>= end_offset
)
1341 n_needed
= MIN (end_offset
- current_offset
,
1342 (uintmax_t) bytes_per_block
);
1343 ok
&= read_block (n_needed
, block
[idx
], &n_bytes_read
);
1344 if (n_bytes_read
< bytes_per_block
)
1346 assert (n_bytes_read
== bytes_per_block
);
1347 write_block (current_offset
, n_bytes_read
,
1348 block
[!idx
], block
[idx
]);
1349 current_offset
+= n_bytes_read
;
1357 ok
&= read_block (bytes_per_block
, block
[idx
], &n_bytes_read
);
1358 if (n_bytes_read
< bytes_per_block
)
1360 assert (n_bytes_read
== bytes_per_block
);
1361 write_block (current_offset
, n_bytes_read
,
1362 block
[!idx
], block
[idx
]);
1363 current_offset
+= n_bytes_read
;
1368 if (n_bytes_read
> 0)
1371 size_t bytes_to_write
;
1375 /* Ensure zero-byte padding up to the smallest multiple of l_c_m that
1376 is at least as large as n_bytes_read. */
1377 bytes_to_write
= l_c_m
* ((n_bytes_read
+ l_c_m
- 1) / l_c_m
);
1379 memset (block
[idx
] + n_bytes_read
, 0, bytes_to_write
- n_bytes_read
);
1380 write_block (current_offset
, n_bytes_read
, block
[!idx
], block
[idx
]);
1381 current_offset
+= n_bytes_read
;
1384 format_address (current_offset
, '\n');
1386 if (limit_bytes_to_format
&& current_offset
>= end_offset
)
1387 ok
&= check_and_close (0);
1394 /* STRINGS mode. Find each "string constant" in the input.
1395 A string constant is a run of at least 'string_min' ASCII
1396 graphic (or formatting) characters terminated by a null.
1397 Based on a function written by Richard Stallman for a
1398 traditional version of od. Return true if successful. */
1403 size_t bufsize
= MAX (100, string_min
);
1404 char *buf
= xmalloc (bufsize
);
1405 uintmax_t address
= n_bytes_to_skip
;
1413 /* See if the next 'string_min' chars are all printing chars. */
1416 if (limit_bytes_to_format
1417 && (end_offset
< string_min
|| end_offset
- string_min
<= address
))
1420 for (i
= 0; i
< string_min
; i
++)
1422 ok
&= read_char (&c
);
1430 /* Found a non-printing. Try again starting with next char. */
1435 /* We found a run of 'string_min' printable characters.
1436 Now see if it is terminated with a null byte. */
1437 while (!limit_bytes_to_format
|| address
< end_offset
)
1441 buf
= X2REALLOC (buf
, &bufsize
);
1443 ok
&= read_char (&c
);
1451 break; /* It is; print this string. */
1453 goto tryline
; /* It isn't; give up on this string. */
1454 buf
[i
++] = c
; /* String continues; store it all. */
1457 /* If we get here, the string is all printable and null-terminated,
1458 so print it. It is all in 'buf' and 'i' is its length. */
1460 format_address (address
- i
- 1, ' ');
1462 for (i
= 0; (c
= buf
[i
]); i
++)
1467 fputs ("\\a", stdout
);
1471 fputs ("\\b", stdout
);
1475 fputs ("\\f", stdout
);
1479 fputs ("\\n", stdout
);
1483 fputs ("\\r", stdout
);
1487 fputs ("\\t", stdout
);
1491 fputs ("\\v", stdout
);
1501 /* We reach this point only if we search through
1502 (max_bytes_to_format - string_min) bytes before reaching EOF. */
1506 ok
&= check_and_close (0);
1511 main (int argc
, char **argv
)
1516 size_t desired_width
IF_LINT ( = 0);
1517 bool modern
= false;
1518 bool width_specified
= false;
1520 size_t width_per_block
= 0;
1521 static char const multipliers
[] = "bEGKkMmPTYZ0";
1523 /* The old-style 'pseudo starting address' to be printed in parentheses
1524 after any true address. */
1525 uintmax_t pseudo_start
IF_LINT ( = 0);
1527 initialize_main (&argc
, &argv
);
1528 set_program_name (argv
[0]);
1529 setlocale (LC_ALL
, "");
1530 bindtextdomain (PACKAGE
, LOCALEDIR
);
1531 textdomain (PACKAGE
);
1533 atexit (close_stdout
);
1535 for (i
= 0; i
<= MAX_INTEGRAL_TYPE_SIZE
; i
++)
1536 integral_type_size
[i
] = NO_SIZE
;
1538 integral_type_size
[sizeof (char)] = CHAR
;
1539 integral_type_size
[sizeof (short int)] = SHORT
;
1540 integral_type_size
[sizeof (int)] = INT
;
1541 integral_type_size
[sizeof (long int)] = LONG
;
1542 #if HAVE_UNSIGNED_LONG_LONG_INT
1543 /* If 'long int' and 'long long int' have the same size, it's fine
1544 to overwrite the entry for 'long' with this one. */
1545 integral_type_size
[sizeof (unsigned_long_long_int
)] = LONG_LONG
;
1548 for (i
= 0; i
<= MAX_FP_TYPE_SIZE
; i
++)
1549 fp_type_size
[i
] = NO_SIZE
;
1551 fp_type_size
[sizeof (float)] = FLOAT_SINGLE
;
1552 /* The array entry for 'double' is filled in after that for 'long double'
1553 so that if they are the same size, we avoid any overhead of
1554 long double computation in libc. */
1555 fp_type_size
[sizeof (long double)] = FLOAT_LONG_DOUBLE
;
1556 fp_type_size
[sizeof (double)] = FLOAT_DOUBLE
;
1559 n_specs_allocated
= 0;
1562 format_address
= format_address_std
;
1564 address_pad_len
= 7;
1565 flag_dump_strings
= false;
1570 enum strtol_error s_err
;
1572 int c
= getopt_long (argc
, argv
, short_options
, long_options
, &oi
);
1583 format_address
= format_address_std
;
1585 address_pad_len
= 7;
1588 format_address
= format_address_std
;
1590 address_pad_len
= 7;
1593 format_address
= format_address_std
;
1595 address_pad_len
= 6;
1598 format_address
= format_address_none
;
1599 address_pad_len
= 0;
1602 error (EXIT_FAILURE
, 0,
1603 _("invalid output address radix '%c';\
1604 it must be one character from [doxn]"),
1612 s_err
= xstrtoumax (optarg
, NULL
, 0, &n_bytes_to_skip
, multipliers
);
1613 if (s_err
!= LONGINT_OK
)
1614 xstrtol_fatal (s_err
, oi
, c
, long_options
, optarg
);
1619 limit_bytes_to_format
= true;
1621 s_err
= xstrtoumax (optarg
, NULL
, 0, &max_bytes_to_format
,
1623 if (s_err
!= LONGINT_OK
)
1624 xstrtol_fatal (s_err
, oi
, c
, long_options
, optarg
);
1633 s_err
= xstrtoumax (optarg
, NULL
, 0, &tmp
, multipliers
);
1634 if (s_err
!= LONGINT_OK
)
1635 xstrtol_fatal (s_err
, oi
, c
, long_options
, optarg
);
1637 /* The minimum string length may be no larger than SIZE_MAX,
1638 since we may allocate a buffer of this size. */
1640 error (EXIT_FAILURE
, 0, _("%s is too large"), optarg
);
1644 flag_dump_strings
= true;
1649 ok
&= decode_format_string (optarg
);
1654 abbreviate_duplicate_blocks
= false;
1657 case TRADITIONAL_OPTION
:
1661 /* The next several cases map the traditional format
1662 specification options to the corresponding modern format
1663 specs. GNU od accepts any combination of old- and
1664 new-style options. Format specification options accumulate.
1665 The obsolescent and undocumented formats are compatible
1666 with FreeBSD 4.10 od. */
1668 #define CASE_OLD_ARG(old_char,new_string) \
1670 ok &= decode_format_string (new_string); \
1673 CASE_OLD_ARG ('a', "a");
1674 CASE_OLD_ARG ('b', "o1");
1675 CASE_OLD_ARG ('c', "c");
1676 CASE_OLD_ARG ('D', "u4"); /* obsolescent and undocumented */
1677 CASE_OLD_ARG ('d', "u2");
1678 case 'F': /* obsolescent and undocumented alias */
1679 CASE_OLD_ARG ('e', "fD"); /* obsolescent and undocumented */
1680 CASE_OLD_ARG ('f', "fF");
1681 case 'X': /* obsolescent and undocumented alias */
1682 CASE_OLD_ARG ('H', "x4"); /* obsolescent and undocumented */
1683 CASE_OLD_ARG ('i', "dI");
1684 case 'I': case 'L': /* obsolescent and undocumented aliases */
1685 CASE_OLD_ARG ('l', "dL");
1686 CASE_OLD_ARG ('O', "o4"); /* obsolesent and undocumented */
1687 case 'B': /* obsolescent and undocumented alias */
1688 CASE_OLD_ARG ('o', "o2");
1689 CASE_OLD_ARG ('s', "d2");
1690 case 'h': /* obsolescent and undocumented alias */
1691 CASE_OLD_ARG ('x', "x2");
1697 width_specified
= true;
1705 s_err
= xstrtoumax (optarg
, NULL
, 10, &w_tmp
, "");
1706 if (s_err
!= LONGINT_OK
)
1707 xstrtol_fatal (s_err
, oi
, c
, long_options
, optarg
);
1708 if (SIZE_MAX
< w_tmp
)
1709 error (EXIT_FAILURE
, 0, _("%s is too large"), optarg
);
1710 desired_width
= w_tmp
;
1714 case_GETOPT_HELP_CHAR
;
1716 case_GETOPT_VERSION_CHAR (PROGRAM_NAME
, AUTHORS
);
1719 usage (EXIT_FAILURE
);
1725 exit (EXIT_FAILURE
);
1727 if (flag_dump_strings
&& n_specs
> 0)
1728 error (EXIT_FAILURE
, 0,
1729 _("no type may be specified when dumping strings"));
1731 n_files
= argc
- optind
;
1733 /* If the --traditional option is used, there may be from
1734 0 to 3 remaining command line arguments; handle each case
1736 od [file] [[+]offset[.][b] [[+]label[.][b]]]
1737 The offset and label have the same syntax.
1739 If --traditional is not given, and if no modern options are
1740 given, and if the offset begins with + or (if there are two
1741 operands) a digit, accept only this form, as per POSIX:
1742 od [file] [[+]offset[.][b]]
1745 if (!modern
|| traditional
)
1753 if ((traditional
|| argv
[optind
][0] == '+')
1754 && parse_old_offset (argv
[optind
], &o1
))
1756 n_bytes_to_skip
= o1
;
1763 if ((traditional
|| argv
[optind
+ 1][0] == '+'
1764 || ISDIGIT (argv
[optind
+ 1][0]))
1765 && parse_old_offset (argv
[optind
+ 1], &o2
))
1767 if (traditional
&& parse_old_offset (argv
[optind
], &o1
))
1769 n_bytes_to_skip
= o1
;
1770 flag_pseudo_start
= true;
1777 n_bytes_to_skip
= o2
;
1779 argv
[optind
+ 1] = argv
[optind
];
1787 && parse_old_offset (argv
[optind
+ 1], &o1
)
1788 && parse_old_offset (argv
[optind
+ 2], &o2
))
1790 n_bytes_to_skip
= o1
;
1791 flag_pseudo_start
= true;
1793 argv
[optind
+ 2] = argv
[optind
];
1800 if (traditional
&& 1 < n_files
)
1802 error (0, 0, _("extra operand %s"), quote (argv
[optind
+ 1]));
1804 _("compatibility mode supports at most one file"));
1805 usage (EXIT_FAILURE
);
1809 if (flag_pseudo_start
)
1811 if (format_address
== format_address_none
)
1814 address_pad_len
= 7;
1815 format_address
= format_address_paren
;
1818 format_address
= format_address_label
;
1821 if (limit_bytes_to_format
)
1823 end_offset
= n_bytes_to_skip
+ max_bytes_to_format
;
1824 if (end_offset
< n_bytes_to_skip
)
1825 error (EXIT_FAILURE
, 0, _("skip-bytes + read-bytes is too large"));
1829 decode_format_string ("oS");
1833 /* Set the global pointer FILE_LIST so that it
1834 references the first file-argument on the command-line. */
1836 file_list
= (char const *const *) &argv
[optind
];
1840 /* No files were listed on the command line.
1841 Set the global pointer FILE_LIST so that it
1842 references the null-terminated list of one name: "-". */
1844 file_list
= default_file_list
;
1847 /* open the first input file */
1848 ok
= open_next_file ();
1849 if (in_stream
== NULL
)
1852 /* skip over any unwanted header bytes */
1853 ok
&= skip (n_bytes_to_skip
);
1854 if (in_stream
== NULL
)
1857 pseudo_offset
= (flag_pseudo_start
? pseudo_start
- n_bytes_to_skip
: 0);
1859 /* Compute output block length. */
1862 if (width_specified
)
1864 if (desired_width
!= 0 && desired_width
% l_c_m
== 0)
1865 bytes_per_block
= desired_width
;
1868 error (0, 0, _("warning: invalid width %lu; using %d instead"),
1869 (unsigned long int) desired_width
, l_c_m
);
1870 bytes_per_block
= l_c_m
;
1875 if (l_c_m
< DEFAULT_BYTES_PER_BLOCK
)
1876 bytes_per_block
= l_c_m
* (DEFAULT_BYTES_PER_BLOCK
/ l_c_m
);
1878 bytes_per_block
= l_c_m
;
1881 /* Compute padding necessary to align output block. */
1882 for (i
= 0; i
< n_specs
; i
++)
1884 int fields_per_block
= bytes_per_block
/ width_bytes
[spec
[i
].size
];
1885 int block_width
= (spec
[i
].field_width
+ 1) * fields_per_block
;
1886 if (width_per_block
< block_width
)
1887 width_per_block
= block_width
;
1889 for (i
= 0; i
< n_specs
; i
++)
1891 int fields_per_block
= bytes_per_block
/ width_bytes
[spec
[i
].size
];
1892 int block_width
= spec
[i
].field_width
* fields_per_block
;
1893 spec
[i
].pad_width
= width_per_block
- block_width
;
1897 printf ("lcm=%d, width_per_block=%zu\n", l_c_m
, width_per_block
);
1898 for (i
= 0; i
< n_specs
; i
++)
1900 int fields_per_block
= bytes_per_block
/ width_bytes
[spec
[i
].size
];
1901 assert (bytes_per_block
% width_bytes
[spec
[i
].size
] == 0);
1902 assert (1 <= spec
[i
].pad_width
/ fields_per_block
);
1903 printf ("%d: fmt=\"%s\" in_width=%d out_width=%d pad=%d\n",
1904 i
, spec
[i
].fmt_string
, width_bytes
[spec
[i
].size
],
1905 spec
[i
].field_width
, spec
[i
].pad_width
);
1909 ok
&= (flag_dump_strings
? dump_strings () : dump ());
1913 if (have_read_stdin
&& fclose (stdin
) == EOF
)
1914 error (EXIT_FAILURE
, errno
, _("standard input"));
1916 exit (ok
? EXIT_SUCCESS
: EXIT_FAILURE
);