1 /* od -- dump files in octal and other formats
2 Copyright (C) 1992-2013 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 If first and second call formats both apply, the second format is assumed\n\
309 if the last operand begins with + or (if there are 2 operands) a digit.\n\
310 An OFFSET operand means -j OFFSET. LABEL is the pseudo-address\n\
311 at first byte printed, incremented when dump is progressing.\n\
312 For OFFSET and LABEL, a 0x or 0X prefix indicates hexadecimal;\n\
313 suffixes may be . for octal and b for multiply by 512.\n\
316 emit_mandatory_arg_note ();
319 -A, --address-radix=RADIX output format for file offsets; RADIX is one\n\
320 of [doxn], for Decimal, Octal, Hex or None\n\
321 -j, --skip-bytes=BYTES skip BYTES input bytes first\n\
324 -N, --read-bytes=BYTES limit dump to BYTES input bytes\n\
325 -S BYTES, --strings[=BYTES] output strings of at least BYTES graphic chars;\
327 3 is implied when BYTES is not specified\n\
328 -t, --format=TYPE select output format or formats\n\
329 -v, --output-duplicates do not use * to mark line suppression\n\
330 -w[BYTES], --width[=BYTES] output BYTES bytes per output line;\n\
331 32 is implied when BYTES is not specified\n\
332 --traditional accept arguments in third form above\n\
334 fputs (HELP_OPTION_DESCRIPTION
, stdout
);
335 fputs (VERSION_OPTION_DESCRIPTION
, stdout
);
339 Traditional format specifications may be intermixed; they accumulate:\n\
340 -a same as -t a, select named characters, ignoring high-order bit\n\
341 -b same as -t o1, select octal bytes\n\
342 -c same as -t c, select printable characters or backslash escapes\n\
343 -d same as -t u2, select unsigned decimal 2-byte units\n\
346 -f same as -t fF, select floats\n\
347 -i same as -t dI, select decimal ints\n\
348 -l same as -t dL, select decimal longs\n\
349 -o same as -t o2, select octal 2-byte units\n\
350 -s same as -t d2, select decimal 2-byte units\n\
351 -x same as -t x2, select hexadecimal 2-byte units\n\
356 TYPE is made up of one or more of these specifications:\n\
357 a named character, ignoring high-order bit\n\
358 c printable character or backslash escape\n\
361 d[SIZE] signed decimal, SIZE bytes per integer\n\
362 f[SIZE] floating point, SIZE bytes per integer\n\
363 o[SIZE] octal, SIZE bytes per integer\n\
364 u[SIZE] unsigned decimal, SIZE bytes per integer\n\
365 x[SIZE] hexadecimal, SIZE bytes per integer\n\
369 SIZE is a number. For TYPE in [doux], SIZE may also be C for\n\
370 sizeof(char), S for sizeof(short), I for sizeof(int) or L for\n\
371 sizeof(long). If TYPE is f, SIZE may also be F for sizeof(float), D\n\
372 for sizeof(double) or L for sizeof(long double).\n\
376 Adding a z suffix to any type displays printable characters at the end of\n\
382 BYTES is hex with 0x or 0X prefix, and may have a multiplier suffix:\n\
388 and so on for G, T, P, E, Z, Y.\n\
390 emit_ancillary_info ();
395 /* Define the print functions. */
397 #define PRINT_FIELDS(N, T, FMT_STRING, ACTION) \
399 N (size_t fields, size_t blank, void const *block, \
400 char const *FMT_STRING, int width, int pad) \
402 T const *p = block; \
404 int pad_remaining = pad; \
405 for (i = fields; blank < i; i--) \
407 int next_pad = pad * (i - 1) / fields; \
408 int adjusted_width = pad_remaining - next_pad + width; \
411 pad_remaining = next_pad; \
415 #define PRINT_TYPE(N, T) \
416 PRINT_FIELDS (N, T, fmt_string, xprintf (fmt_string, adjusted_width, x))
418 #define PRINT_FLOATTYPE(N, T, FTOASTR, BUFSIZE) \
419 PRINT_FIELDS (N, T, fmt_string _GL_UNUSED, \
421 FTOASTR (buf, sizeof buf, 0, 0, x); \
422 xprintf ("%*s", adjusted_width, buf))
424 PRINT_TYPE (print_s_char
, signed char)
425 PRINT_TYPE (print_char
, unsigned char)
426 PRINT_TYPE (print_s_short
, short int)
427 PRINT_TYPE (print_short
, unsigned short int)
428 PRINT_TYPE (print_int
, unsigned int)
429 PRINT_TYPE (print_long
, unsigned long int)
430 PRINT_TYPE (print_long_long
, unsigned_long_long_int
)
432 PRINT_FLOATTYPE (print_float
, float, ftoastr
, FLT_BUFSIZE_BOUND
)
433 PRINT_FLOATTYPE (print_double
, double, dtoastr
, DBL_BUFSIZE_BOUND
)
434 PRINT_FLOATTYPE (print_long_double
, long double, ldtoastr
, LDBL_BUFSIZE_BOUND
)
437 #undef PRINT_FLOATTYPE
440 dump_hexl_mode_trailer (size_t n_bytes
, const char *block
)
443 fputs (" >", stdout
);
444 for (i
= n_bytes
; i
> 0; i
--)
446 unsigned char c
= *block
++;
447 unsigned char c2
= (isprint (c
) ? c
: '.');
454 print_named_ascii (size_t fields
, size_t blank
, void const *block
,
455 const char *unused_fmt_string _GL_UNUSED
,
458 unsigned char const *p
= block
;
460 int pad_remaining
= pad
;
461 for (i
= fields
; blank
< i
; i
--)
463 int next_pad
= pad
* (i
- 1) / fields
;
464 int masked_c
= *p
++ & 0x7f;
470 else if (masked_c
<= 040)
471 s
= charname
[masked_c
];
479 xprintf ("%*s", pad_remaining
- next_pad
+ width
, s
);
480 pad_remaining
= next_pad
;
485 print_ascii (size_t fields
, size_t blank
, void const *block
,
486 const char *unused_fmt_string _GL_UNUSED
, int width
,
489 unsigned char const *p
= block
;
491 int pad_remaining
= pad
;
492 for (i
= fields
; blank
< i
; i
--)
494 int next_pad
= pad
* (i
- 1) / fields
;
495 unsigned char c
= *p
++;
534 sprintf (buf
, (isprint (c
) ? "%c" : "%03o"), c
);
538 xprintf ("%*s", pad_remaining
- next_pad
+ width
, s
);
539 pad_remaining
= next_pad
;
543 /* Convert a null-terminated (possibly zero-length) string S to an
544 unsigned long integer value. If S points to a non-digit set *P to S,
545 *VAL to 0, and return true. Otherwise, accumulate the integer value of
546 the string of digits. If the string of digits represents a value
547 larger than ULONG_MAX, don't modify *VAL or *P and return false.
548 Otherwise, advance *P to the first non-digit after S, set *VAL to
549 the result of the conversion and return true. */
552 simple_strtoul (const char *s
, const char **p
, unsigned long int *val
)
554 unsigned long int sum
;
560 if (sum
> (ULONG_MAX
- c
) / 10)
569 /* If S points to a single valid modern od format string, put
570 a description of that format in *TSPEC, make *NEXT point at the
571 character following the just-decoded format (if *NEXT is non-NULL),
572 and return true. If S is not valid, don't modify *NEXT or *TSPEC,
573 give a diagnostic, and return false. For example, if S were
574 "d4afL" *NEXT would be set to "afL" and *TSPEC would be
576 fmt = SIGNED_DECIMAL;
577 size = INT or LONG; (whichever integral_type_size[4] resolves to)
578 print_function = print_int; (assuming size == INT)
582 pad_width is determined later, but is at least as large as the
583 number of fields printed per row.
584 S_ORIG is solely for reporting errors. It should be the full format
589 decode_one_format (const char *s_orig
, const char *s
, const char **next
,
592 enum size_spec size_spec
;
593 unsigned long int size
;
594 enum output_format fmt
;
595 void (*print_function
) (size_t, size_t, void const *, char const *,
601 assert (tspec
!= NULL
);
615 size
= sizeof (char);
620 size
= sizeof (short int);
630 size
= sizeof (long int);
634 if (! simple_strtoul (s
, &p
, &size
))
636 /* The integer at P in S would overflow an unsigned long int.
637 A digit string that long is sufficiently odd looking
638 that the following diagnostic is sufficient. */
639 error (0, 0, _("invalid type string %s"), quote (s_orig
));
646 if (MAX_INTEGRAL_TYPE_SIZE
< size
647 || integral_type_size
[size
] == NO_SIZE
)
649 error (0, 0, _("invalid type string %s;\nthis system"
650 " doesn't provide a %lu-byte integral type"),
651 quote (s_orig
), size
);
659 #define ISPEC_TO_FORMAT(Spec, Min_format, Long_format, Max_format) \
660 ((Spec) == LONG_LONG ? (Max_format) \
661 : ((Spec) == LONG ? (Long_format) \
664 size_spec = integral_type_size[size];
669 fmt
= SIGNED_DECIMAL
;
670 field_width
= bytes_to_signed_dec_digits
[size
];
671 sprintf (tspec
->fmt_string
, "%%*%s",
672 ISPEC_TO_FORMAT (size_spec
, "d", "ld", PRIdMAX
));
677 sprintf (tspec
->fmt_string
, "%%*.%d%s",
678 (field_width
= bytes_to_oct_digits
[size
]),
679 ISPEC_TO_FORMAT (size_spec
, "o", "lo", PRIoMAX
));
683 fmt
= UNSIGNED_DECIMAL
;
684 field_width
= bytes_to_unsigned_dec_digits
[size
];
685 sprintf (tspec
->fmt_string
, "%%*%s",
686 ISPEC_TO_FORMAT (size_spec
, "u", "lu", PRIuMAX
));
691 sprintf (tspec
->fmt_string
, "%%*.%d%s",
692 (field_width
= bytes_to_hex_digits
[size
]),
693 ISPEC_TO_FORMAT (size_spec
, "x", "lx", PRIxMAX
));
700 assert (strlen (tspec
->fmt_string
) < FMT_BYTES_ALLOCATED
);
705 print_function
= (fmt
== SIGNED_DECIMAL
711 print_function
= (fmt
== SIGNED_DECIMAL
717 print_function
= print_int
;
721 print_function
= print_long
;
725 print_function
= print_long_long
;
734 fmt
= FLOATING_POINT
;
740 size
= sizeof (float);
745 size
= sizeof (double);
750 size
= sizeof (long double);
754 if (! simple_strtoul (s
, &p
, &size
))
756 /* The integer at P in S would overflow an unsigned long int.
757 A digit string that long is sufficiently odd looking
758 that the following diagnostic is sufficient. */
759 error (0, 0, _("invalid type string %s"), quote (s_orig
));
763 size
= sizeof (double);
766 if (size
> MAX_FP_TYPE_SIZE
767 || fp_type_size
[size
] == NO_SIZE
)
770 _("invalid type string %s;\n"
771 "this system doesn't provide a %lu-byte"
772 " floating point type"),
773 quote (s_orig
), size
);
780 size_spec
= fp_type_size
[size
];
783 struct lconv
const *locale
= localeconv ();
784 size_t decimal_point_len
=
785 (locale
->decimal_point
[0] ? strlen (locale
->decimal_point
) : 1);
790 print_function
= print_float
;
791 field_width
= FLT_STRLEN_BOUND_L (decimal_point_len
);
795 print_function
= print_double
;
796 field_width
= DBL_STRLEN_BOUND_L (decimal_point_len
);
799 case FLOAT_LONG_DOUBLE
:
800 print_function
= print_long_double
;
801 field_width
= LDBL_STRLEN_BOUND_L (decimal_point_len
);
813 fmt
= NAMED_CHARACTER
;
815 print_function
= print_named_ascii
;
823 print_function
= print_ascii
;
828 error (0, 0, _("invalid character '%c' in type string %s"),
833 tspec
->size
= size_spec
;
835 tspec
->print_function
= print_function
;
837 tspec
->field_width
= field_width
;
838 tspec
->hexl_mode_trailer
= (*s
== 'z');
839 if (tspec
->hexl_mode_trailer
)
848 /* Given a list of one or more input filenames FILE_LIST, set the global
849 file pointer IN_STREAM and the global string INPUT_FILENAME to the
850 first one that can be successfully opened. Modify FILE_LIST to
851 reference the next filename in the list. A file name of "-" is
852 interpreted as standard input. If any file open fails, give an error
853 message and return false. */
856 open_next_file (void)
862 input_filename
= *file_list
;
863 if (input_filename
== NULL
)
867 if (STREQ (input_filename
, "-"))
869 input_filename
= _("standard input");
871 have_read_stdin
= true;
872 if (O_BINARY
&& ! isatty (STDIN_FILENO
))
873 xfreopen (NULL
, "rb", stdin
);
877 in_stream
= fopen (input_filename
, (O_BINARY
? "rb" : "r"));
878 if (in_stream
== NULL
)
880 error (0, errno
, "%s", input_filename
);
885 while (in_stream
== NULL
);
887 if (limit_bytes_to_format
&& !flag_dump_strings
)
888 setvbuf (in_stream
, NULL
, _IONBF
, 0);
893 /* Test whether there have been errors on in_stream, and close it if
894 it is not standard input. Return false if there has been an error
895 on in_stream or stdout; return true otherwise. This function will
896 report more than one error only if both a read and a write error
897 have occurred. IN_ERRNO, if nonzero, is the error number
898 corresponding to the most recent action for IN_STREAM. */
901 check_and_close (int in_errno
)
905 if (in_stream
!= NULL
)
907 if (ferror (in_stream
))
909 error (0, in_errno
, _("%s: read error"), input_filename
);
910 if (! STREQ (file_list
[-1], "-"))
914 else if (! STREQ (file_list
[-1], "-") && fclose (in_stream
) != 0)
916 error (0, errno
, "%s", input_filename
);
925 error (0, 0, _("write error"));
932 /* Decode the modern od format string S. Append the decoded
933 representation to the global array SPEC, reallocating SPEC if
934 necessary. Return true if S is valid. */
937 decode_format_string (const char *s
)
939 const char *s_orig
= s
;
946 if (n_specs_allocated
<= n_specs
)
947 spec
= X2NREALLOC (spec
, &n_specs_allocated
);
949 if (! decode_one_format (s_orig
, s
, &next
, &spec
[n_specs
]))
960 /* Given a list of one or more input filenames FILE_LIST, set the global
961 file pointer IN_STREAM to position N_SKIP in the concatenation of
962 those files. If any file operation fails or if there are fewer than
963 N_SKIP bytes in the combined input, give an error message and return
964 false. When possible, use seek rather than read operations to
965 advance IN_STREAM. */
968 skip (uintmax_t n_skip
)
976 while (in_stream
!= NULL
) /* EOF. */
978 struct stat file_stats
;
980 /* First try seeking. For large offsets, this extra work is
981 worthwhile. If the offset is below some threshold it may be
982 more efficient to move the pointer by reading. There are two
983 issues when trying to seek:
984 - the file must be seekable.
985 - before seeking to the specified position, make sure
986 that the new position is in the current file.
987 Try to do that by getting file's size using fstat.
988 But that will work only for regular files. */
990 if (fstat (fileno (in_stream
), &file_stats
) == 0)
992 /* The st_size field is valid for regular files.
993 If the number of bytes left to skip is larger than
994 the size of the current file, we can decrement n_skip
995 and go on to the next file. Skip this optimization also
996 when st_size is 0, because some kernels report that
997 nonempty files in /proc have st_size == 0. */
998 if (S_ISREG (file_stats
.st_mode
) && 0 < file_stats
.st_size
)
1000 if ((uintmax_t) file_stats
.st_size
< n_skip
)
1001 n_skip
-= file_stats
.st_size
;
1004 if (fseeko (in_stream
, n_skip
, SEEK_CUR
) != 0)
1013 /* If it's not a regular file with nonnegative size,
1014 position the file pointer by reading. */
1019 size_t n_bytes_read
, n_bytes_to_read
= BUFSIZ
;
1023 if (n_skip
< n_bytes_to_read
)
1024 n_bytes_to_read
= n_skip
;
1025 n_bytes_read
= fread (buf
, 1, n_bytes_to_read
, in_stream
);
1026 n_skip
-= n_bytes_read
;
1027 if (n_bytes_read
!= n_bytes_to_read
)
1041 else /* cannot fstat() file */
1043 error (0, errno
, "%s", input_filename
);
1047 ok
&= check_and_close (in_errno
);
1049 ok
&= open_next_file ();
1053 error (EXIT_FAILURE
, 0, _("cannot skip past end of combined input"));
1059 format_address_none (uintmax_t address _GL_UNUSED
,
1065 format_address_std (uintmax_t address
, char c
)
1067 char buf
[MAX_ADDRESS_LENGTH
+ 2];
1068 char *p
= buf
+ sizeof buf
;
1073 pbound
= p
- address_pad_len
;
1075 /* Use a special case of the code for each base. This is measurably
1076 faster than generic code. */
1077 switch (address_base
)
1081 *--p
= '0' + (address
& 7);
1082 while ((address
>>= 3) != 0);
1087 *--p
= '0' + (address
% 10);
1088 while ((address
/= 10) != 0);
1093 *--p
= "0123456789abcdef"[address
& 15];
1094 while ((address
>>= 4) != 0);
1105 format_address_paren (uintmax_t address
, char c
)
1108 format_address_std (address
, ')');
1114 format_address_label (uintmax_t address
, char c
)
1116 format_address_std (address
, ' ');
1117 format_address_paren (address
+ pseudo_offset
, c
);
1120 /* Write N_BYTES bytes from CURR_BLOCK to standard output once for each
1121 of the N_SPEC format specs. CURRENT_OFFSET is the byte address of
1122 CURR_BLOCK in the concatenation of input files, and it is printed
1123 (optionally) only before the output line associated with the first
1124 format spec. When duplicate blocks are being abbreviated, the output
1125 for a sequence of identical input blocks is the output for the first
1126 block followed by an asterisk alone on a line. It is valid to compare
1127 the blocks PREV_BLOCK and CURR_BLOCK only when N_BYTES == BYTES_PER_BLOCK.
1128 That condition may be false only for the last input block. */
1131 write_block (uintmax_t current_offset
, size_t n_bytes
,
1132 const char *prev_block
, const char *curr_block
)
1134 static bool first
= true;
1135 static bool prev_pair_equal
= false;
1137 #define EQUAL_BLOCKS(b1, b2) (memcmp (b1, b2, bytes_per_block) == 0)
1139 if (abbreviate_duplicate_blocks
1140 && !first
&& n_bytes
== bytes_per_block
1141 && EQUAL_BLOCKS (prev_block
, curr_block
))
1143 if (prev_pair_equal
)
1145 /* The two preceding blocks were equal, and the current
1146 block is the same as the last one, so print nothing. */
1151 prev_pair_equal
= true;
1158 prev_pair_equal
= false;
1159 for (i
= 0; i
< n_specs
; i
++)
1161 int datum_width
= width_bytes
[spec
[i
].size
];
1162 int fields_per_block
= bytes_per_block
/ datum_width
;
1163 int blank_fields
= (bytes_per_block
- n_bytes
) / datum_width
;
1165 format_address (current_offset
, '\0');
1167 printf ("%*s", address_pad_len
, "");
1168 (*spec
[i
].print_function
) (fields_per_block
, blank_fields
,
1169 curr_block
, spec
[i
].fmt_string
,
1170 spec
[i
].field_width
, spec
[i
].pad_width
);
1171 if (spec
[i
].hexl_mode_trailer
)
1173 /* space-pad out to full line width, then dump the trailer */
1174 int field_width
= spec
[i
].field_width
;
1175 int pad_width
= (spec
[i
].pad_width
* blank_fields
1176 / fields_per_block
);
1177 printf ("%*s", blank_fields
* field_width
+ pad_width
, "");
1178 dump_hexl_mode_trailer (n_bytes
, curr_block
);
1186 /* Read a single byte into *C from the concatenation of the input files
1187 named in the global array FILE_LIST. On the first call to this
1188 function, the global variable IN_STREAM is expected to be an open
1189 stream associated with the input file INPUT_FILENAME. If IN_STREAM
1190 is at end-of-file, close it and update the global variables IN_STREAM
1191 and INPUT_FILENAME so they correspond to the next file in the list.
1192 Then try to read a byte from the newly opened file. Repeat if
1193 necessary until EOF is reached for the last file in FILE_LIST, then
1194 set *C to EOF and return. Subsequent calls do likewise. Return
1195 true if successful. */
1204 while (in_stream
!= NULL
) /* EOF. */
1206 *c
= fgetc (in_stream
);
1211 ok
&= check_and_close (errno
);
1213 ok
&= open_next_file ();
1219 /* Read N bytes into BLOCK from the concatenation of the input files
1220 named in the global array FILE_LIST. On the first call to this
1221 function, the global variable IN_STREAM is expected to be an open
1222 stream associated with the input file INPUT_FILENAME. If all N
1223 bytes cannot be read from IN_STREAM, close IN_STREAM and update
1224 the global variables IN_STREAM and INPUT_FILENAME. Then try to
1225 read the remaining bytes from the newly opened file. Repeat if
1226 necessary until EOF is reached for the last file in FILE_LIST.
1227 On subsequent calls, don't modify BLOCK and return true. Set
1228 *N_BYTES_IN_BUFFER to the number of bytes read. If an error occurs,
1229 it will be detected through ferror when the stream is about to be
1230 closed. If there is an error, give a message but continue reading
1231 as usual and return false. Otherwise return true. */
1234 read_block (size_t n
, char *block
, size_t *n_bytes_in_buffer
)
1238 assert (0 < n
&& n
<= bytes_per_block
);
1240 *n_bytes_in_buffer
= 0;
1245 while (in_stream
!= NULL
) /* EOF. */
1250 n_needed
= n
- *n_bytes_in_buffer
;
1251 n_read
= fread (block
+ *n_bytes_in_buffer
, 1, n_needed
, in_stream
);
1253 *n_bytes_in_buffer
+= n_read
;
1255 if (n_read
== n_needed
)
1258 ok
&= check_and_close (errno
);
1260 ok
&= open_next_file ();
1266 /* Return the least common multiple of the sizes associated
1267 with the format specs. */
1269 static int _GL_ATTRIBUTE_PURE
1275 for (i
= 0; i
< n_specs
; i
++)
1276 l_c_m
= lcm (l_c_m
, width_bytes
[spec
[i
].size
]);
1280 /* If S is a valid traditional offset specification with an optional
1281 leading '+' return true and set *OFFSET to the offset it denotes. */
1284 parse_old_offset (const char *s
, uintmax_t *offset
)
1291 /* Skip over any leading '+'. */
1295 /* Determine the radix we'll use to interpret S. If there is a '.',
1296 it's decimal, otherwise, if the string begins with '0X'or '0x',
1297 it's hexadecimal, else octal. */
1298 if (strchr (s
, '.') != NULL
)
1302 if (s
[0] == '0' && (s
[1] == 'x' || s
[1] == 'X'))
1308 return xstrtoumax (s
, NULL
, radix
, offset
, "Bb") == LONGINT_OK
;
1311 /* Read a chunk of size BYTES_PER_BLOCK from the input files, write the
1312 formatted block to standard output, and repeat until the specified
1313 maximum number of bytes has been read or until all input has been
1314 processed. If the last block read is smaller than BYTES_PER_BLOCK
1315 and its size is not a multiple of the size associated with a format
1316 spec, extend the input block with zero bytes until its length is a
1317 multiple of all format spec sizes. Write the final block. Finally,
1318 write on a line by itself the offset of the byte after the last byte
1319 read. Accumulate return values from calls to read_block and
1320 check_and_close, and if any was false, return false.
1321 Otherwise, return true. */
1327 uintmax_t current_offset
;
1330 size_t n_bytes_read
;
1332 block
[0] = xnmalloc (2, bytes_per_block
);
1333 block
[1] = block
[0] + bytes_per_block
;
1335 current_offset
= n_bytes_to_skip
;
1337 if (limit_bytes_to_format
)
1342 if (current_offset
>= end_offset
)
1347 n_needed
= MIN (end_offset
- current_offset
,
1348 (uintmax_t) bytes_per_block
);
1349 ok
&= read_block (n_needed
, block
[idx
], &n_bytes_read
);
1350 if (n_bytes_read
< bytes_per_block
)
1352 assert (n_bytes_read
== bytes_per_block
);
1353 write_block (current_offset
, n_bytes_read
,
1354 block
[!idx
], block
[idx
]);
1355 current_offset
+= n_bytes_read
;
1363 ok
&= read_block (bytes_per_block
, block
[idx
], &n_bytes_read
);
1364 if (n_bytes_read
< bytes_per_block
)
1366 assert (n_bytes_read
== bytes_per_block
);
1367 write_block (current_offset
, n_bytes_read
,
1368 block
[!idx
], block
[idx
]);
1369 current_offset
+= n_bytes_read
;
1374 if (n_bytes_read
> 0)
1377 size_t bytes_to_write
;
1381 /* Ensure zero-byte padding up to the smallest multiple of l_c_m that
1382 is at least as large as n_bytes_read. */
1383 bytes_to_write
= l_c_m
* ((n_bytes_read
+ l_c_m
- 1) / l_c_m
);
1385 memset (block
[idx
] + n_bytes_read
, 0, bytes_to_write
- n_bytes_read
);
1386 write_block (current_offset
, n_bytes_read
, block
[!idx
], block
[idx
]);
1387 current_offset
+= n_bytes_read
;
1390 format_address (current_offset
, '\n');
1392 if (limit_bytes_to_format
&& current_offset
>= end_offset
)
1393 ok
&= check_and_close (0);
1400 /* STRINGS mode. Find each "string constant" in the input.
1401 A string constant is a run of at least 'string_min' ASCII
1402 graphic (or formatting) characters terminated by a null.
1403 Based on a function written by Richard Stallman for a
1404 traditional version of od. Return true if successful. */
1409 size_t bufsize
= MAX (100, string_min
);
1410 char *buf
= xmalloc (bufsize
);
1411 uintmax_t address
= n_bytes_to_skip
;
1419 /* See if the next 'string_min' chars are all printing chars. */
1422 if (limit_bytes_to_format
1423 && (end_offset
< string_min
|| end_offset
- string_min
<= address
))
1426 for (i
= 0; i
< string_min
; i
++)
1428 ok
&= read_char (&c
);
1436 /* Found a non-printing. Try again starting with next char. */
1441 /* We found a run of 'string_min' printable characters.
1442 Now see if it is terminated with a null byte. */
1443 while (!limit_bytes_to_format
|| address
< end_offset
)
1447 buf
= X2REALLOC (buf
, &bufsize
);
1449 ok
&= read_char (&c
);
1457 break; /* It is; print this string. */
1459 goto tryline
; /* It isn't; give up on this string. */
1460 buf
[i
++] = c
; /* String continues; store it all. */
1463 /* If we get here, the string is all printable and null-terminated,
1464 so print it. It is all in 'buf' and 'i' is its length. */
1466 format_address (address
- i
- 1, ' ');
1468 for (i
= 0; (c
= buf
[i
]); i
++)
1473 fputs ("\\a", stdout
);
1477 fputs ("\\b", stdout
);
1481 fputs ("\\f", stdout
);
1485 fputs ("\\n", stdout
);
1489 fputs ("\\r", stdout
);
1493 fputs ("\\t", stdout
);
1497 fputs ("\\v", stdout
);
1507 /* We reach this point only if we search through
1508 (max_bytes_to_format - string_min) bytes before reaching EOF. */
1512 ok
&= check_and_close (0);
1517 main (int argc
, char **argv
)
1522 size_t desired_width
IF_LINT ( = 0);
1523 bool modern
= false;
1524 bool width_specified
= false;
1526 size_t width_per_block
= 0;
1527 static char const multipliers
[] = "bEGKkMmPTYZ0";
1529 /* The old-style 'pseudo starting address' to be printed in parentheses
1530 after any true address. */
1531 uintmax_t pseudo_start
IF_LINT ( = 0);
1533 initialize_main (&argc
, &argv
);
1534 set_program_name (argv
[0]);
1535 setlocale (LC_ALL
, "");
1536 bindtextdomain (PACKAGE
, LOCALEDIR
);
1537 textdomain (PACKAGE
);
1539 atexit (close_stdout
);
1541 for (i
= 0; i
<= MAX_INTEGRAL_TYPE_SIZE
; i
++)
1542 integral_type_size
[i
] = NO_SIZE
;
1544 integral_type_size
[sizeof (char)] = CHAR
;
1545 integral_type_size
[sizeof (short int)] = SHORT
;
1546 integral_type_size
[sizeof (int)] = INT
;
1547 integral_type_size
[sizeof (long int)] = LONG
;
1548 #if HAVE_UNSIGNED_LONG_LONG_INT
1549 /* If 'long int' and 'long long int' have the same size, it's fine
1550 to overwrite the entry for 'long' with this one. */
1551 integral_type_size
[sizeof (unsigned_long_long_int
)] = LONG_LONG
;
1554 for (i
= 0; i
<= MAX_FP_TYPE_SIZE
; i
++)
1555 fp_type_size
[i
] = NO_SIZE
;
1557 fp_type_size
[sizeof (float)] = FLOAT_SINGLE
;
1558 /* The array entry for 'double' is filled in after that for 'long double'
1559 so that if they are the same size, we avoid any overhead of
1560 long double computation in libc. */
1561 fp_type_size
[sizeof (long double)] = FLOAT_LONG_DOUBLE
;
1562 fp_type_size
[sizeof (double)] = FLOAT_DOUBLE
;
1565 n_specs_allocated
= 0;
1568 format_address
= format_address_std
;
1570 address_pad_len
= 7;
1571 flag_dump_strings
= false;
1576 enum strtol_error s_err
;
1578 int c
= getopt_long (argc
, argv
, short_options
, long_options
, &oi
);
1589 format_address
= format_address_std
;
1591 address_pad_len
= 7;
1594 format_address
= format_address_std
;
1596 address_pad_len
= 7;
1599 format_address
= format_address_std
;
1601 address_pad_len
= 6;
1604 format_address
= format_address_none
;
1605 address_pad_len
= 0;
1608 error (EXIT_FAILURE
, 0,
1609 _("invalid output address radix '%c';\
1610 it must be one character from [doxn]"),
1618 s_err
= xstrtoumax (optarg
, NULL
, 0, &n_bytes_to_skip
, multipliers
);
1619 if (s_err
!= LONGINT_OK
)
1620 xstrtol_fatal (s_err
, oi
, c
, long_options
, optarg
);
1625 limit_bytes_to_format
= true;
1627 s_err
= xstrtoumax (optarg
, NULL
, 0, &max_bytes_to_format
,
1629 if (s_err
!= LONGINT_OK
)
1630 xstrtol_fatal (s_err
, oi
, c
, long_options
, optarg
);
1639 s_err
= xstrtoumax (optarg
, NULL
, 0, &tmp
, multipliers
);
1640 if (s_err
!= LONGINT_OK
)
1641 xstrtol_fatal (s_err
, oi
, c
, long_options
, optarg
);
1643 /* The minimum string length may be no larger than SIZE_MAX,
1644 since we may allocate a buffer of this size. */
1646 error (EXIT_FAILURE
, 0, _("%s is too large"), optarg
);
1650 flag_dump_strings
= true;
1655 ok
&= decode_format_string (optarg
);
1660 abbreviate_duplicate_blocks
= false;
1663 case TRADITIONAL_OPTION
:
1667 /* The next several cases map the traditional format
1668 specification options to the corresponding modern format
1669 specs. GNU od accepts any combination of old- and
1670 new-style options. Format specification options accumulate.
1671 The obsolescent and undocumented formats are compatible
1672 with FreeBSD 4.10 od. */
1674 #define CASE_OLD_ARG(old_char,new_string) \
1676 ok &= decode_format_string (new_string); \
1679 CASE_OLD_ARG ('a', "a");
1680 CASE_OLD_ARG ('b', "o1");
1681 CASE_OLD_ARG ('c', "c");
1682 CASE_OLD_ARG ('D', "u4"); /* obsolescent and undocumented */
1683 CASE_OLD_ARG ('d', "u2");
1684 case 'F': /* obsolescent and undocumented alias */
1685 CASE_OLD_ARG ('e', "fD"); /* obsolescent and undocumented */
1686 CASE_OLD_ARG ('f', "fF");
1687 case 'X': /* obsolescent and undocumented alias */
1688 CASE_OLD_ARG ('H', "x4"); /* obsolescent and undocumented */
1689 CASE_OLD_ARG ('i', "dI");
1690 case 'I': case 'L': /* obsolescent and undocumented aliases */
1691 CASE_OLD_ARG ('l', "dL");
1692 CASE_OLD_ARG ('O', "o4"); /* obsolesent and undocumented */
1693 case 'B': /* obsolescent and undocumented alias */
1694 CASE_OLD_ARG ('o', "o2");
1695 CASE_OLD_ARG ('s', "d2");
1696 case 'h': /* obsolescent and undocumented alias */
1697 CASE_OLD_ARG ('x', "x2");
1703 width_specified
= true;
1711 s_err
= xstrtoumax (optarg
, NULL
, 10, &w_tmp
, "");
1712 if (s_err
!= LONGINT_OK
)
1713 xstrtol_fatal (s_err
, oi
, c
, long_options
, optarg
);
1714 if (SIZE_MAX
< w_tmp
)
1715 error (EXIT_FAILURE
, 0, _("%s is too large"), optarg
);
1716 desired_width
= w_tmp
;
1720 case_GETOPT_HELP_CHAR
;
1722 case_GETOPT_VERSION_CHAR (PROGRAM_NAME
, AUTHORS
);
1725 usage (EXIT_FAILURE
);
1731 exit (EXIT_FAILURE
);
1733 if (flag_dump_strings
&& n_specs
> 0)
1734 error (EXIT_FAILURE
, 0,
1735 _("no type may be specified when dumping strings"));
1737 n_files
= argc
- optind
;
1739 /* If the --traditional option is used, there may be from
1740 0 to 3 remaining command line arguments; handle each case
1742 od [file] [[+]offset[.][b] [[+]label[.][b]]]
1743 The offset and label have the same syntax.
1745 If --traditional is not given, and if no modern options are
1746 given, and if the offset begins with + or (if there are two
1747 operands) a digit, accept only this form, as per POSIX:
1748 od [file] [[+]offset[.][b]]
1751 if (!modern
|| traditional
)
1759 if ((traditional
|| argv
[optind
][0] == '+')
1760 && parse_old_offset (argv
[optind
], &o1
))
1762 n_bytes_to_skip
= o1
;
1769 if ((traditional
|| argv
[optind
+ 1][0] == '+'
1770 || ISDIGIT (argv
[optind
+ 1][0]))
1771 && parse_old_offset (argv
[optind
+ 1], &o2
))
1773 if (traditional
&& parse_old_offset (argv
[optind
], &o1
))
1775 n_bytes_to_skip
= o1
;
1776 flag_pseudo_start
= true;
1783 n_bytes_to_skip
= o2
;
1785 argv
[optind
+ 1] = argv
[optind
];
1793 && parse_old_offset (argv
[optind
+ 1], &o1
)
1794 && parse_old_offset (argv
[optind
+ 2], &o2
))
1796 n_bytes_to_skip
= o1
;
1797 flag_pseudo_start
= true;
1799 argv
[optind
+ 2] = argv
[optind
];
1806 if (traditional
&& 1 < n_files
)
1808 error (0, 0, _("extra operand %s"), quote (argv
[optind
+ 1]));
1810 _("compatibility mode supports at most one file"));
1811 usage (EXIT_FAILURE
);
1815 if (flag_pseudo_start
)
1817 if (format_address
== format_address_none
)
1820 address_pad_len
= 7;
1821 format_address
= format_address_paren
;
1824 format_address
= format_address_label
;
1827 if (limit_bytes_to_format
)
1829 end_offset
= n_bytes_to_skip
+ max_bytes_to_format
;
1830 if (end_offset
< n_bytes_to_skip
)
1831 error (EXIT_FAILURE
, 0, _("skip-bytes + read-bytes is too large"));
1835 decode_format_string ("oS");
1839 /* Set the global pointer FILE_LIST so that it
1840 references the first file-argument on the command-line. */
1842 file_list
= (char const *const *) &argv
[optind
];
1846 /* No files were listed on the command line.
1847 Set the global pointer FILE_LIST so that it
1848 references the null-terminated list of one name: "-". */
1850 file_list
= default_file_list
;
1853 /* open the first input file */
1854 ok
= open_next_file ();
1855 if (in_stream
== NULL
)
1858 /* skip over any unwanted header bytes */
1859 ok
&= skip (n_bytes_to_skip
);
1860 if (in_stream
== NULL
)
1863 pseudo_offset
= (flag_pseudo_start
? pseudo_start
- n_bytes_to_skip
: 0);
1865 /* Compute output block length. */
1868 if (width_specified
)
1870 if (desired_width
!= 0 && desired_width
% l_c_m
== 0)
1871 bytes_per_block
= desired_width
;
1874 error (0, 0, _("warning: invalid width %lu; using %d instead"),
1875 (unsigned long int) desired_width
, l_c_m
);
1876 bytes_per_block
= l_c_m
;
1881 if (l_c_m
< DEFAULT_BYTES_PER_BLOCK
)
1882 bytes_per_block
= l_c_m
* (DEFAULT_BYTES_PER_BLOCK
/ l_c_m
);
1884 bytes_per_block
= l_c_m
;
1887 /* Compute padding necessary to align output block. */
1888 for (i
= 0; i
< n_specs
; i
++)
1890 int fields_per_block
= bytes_per_block
/ width_bytes
[spec
[i
].size
];
1891 int block_width
= (spec
[i
].field_width
+ 1) * fields_per_block
;
1892 if (width_per_block
< block_width
)
1893 width_per_block
= block_width
;
1895 for (i
= 0; i
< n_specs
; i
++)
1897 int fields_per_block
= bytes_per_block
/ width_bytes
[spec
[i
].size
];
1898 int block_width
= spec
[i
].field_width
* fields_per_block
;
1899 spec
[i
].pad_width
= width_per_block
- block_width
;
1903 printf ("lcm=%d, width_per_block=%zu\n", l_c_m
, width_per_block
);
1904 for (i
= 0; i
< n_specs
; i
++)
1906 int fields_per_block
= bytes_per_block
/ width_bytes
[spec
[i
].size
];
1907 assert (bytes_per_block
% width_bytes
[spec
[i
].size
] == 0);
1908 assert (1 <= spec
[i
].pad_width
/ fields_per_block
);
1909 printf ("%d: fmt=\"%s\" in_width=%d out_width=%d pad=%d\n",
1910 i
, spec
[i
].fmt_string
, width_bytes
[spec
[i
].size
],
1911 spec
[i
].field_width
, spec
[i
].pad_width
);
1915 ok
&= (flag_dump_strings
? dump_strings () : dump ());
1919 if (have_read_stdin
&& fclose (stdin
) == EOF
)
1920 error (EXIT_FAILURE
, errno
, _("standard input"));
1922 exit (ok
? EXIT_SUCCESS
: EXIT_FAILURE
);