1FILE(1) BSD General Commands Manual FILE(1)
2
4 file — determine file type
5
7 file [-bcdEhiklLNnprsSvzZ0] [--apple] [--extension] [--mime-encoding]
8 [--mime-type] [-e testname] [-F separator] [-f namefile]
9 [-m magicfiles] [-P name=value] file ...
10 file -C [-m magicfiles]
11 file [--help]
12
14 This manual page documents version 5.36 of the file command.
15 file tests each argument in an attempt to classify it. There are three
17 sets of tests, performed in this order: filesystem tests, magic tests,
18 and language tests. The first test that succeeds causes the file type to
19 be printed.
20 The type printed will usually contain one of the words text (the file
22 contains only printing characters and a few common control characters and
23 is probably safe to read on an ASCII terminal), executable (the file con‐
24 tains the result of compiling a program in a form understandable to some
25 UNIX kernel or another), or data meaning anything else (data is usually
26 “binary” or non-printable). Exceptions are well-known file formats (core
27 files, tar archives) that are known to contain binary data. When modify‐
28 ing magic files or the program itself, make sure to preserve these
29 keywords. Users depend on knowing that all the readable files in a
30 directory have the word “text” printed. Don't do as Berkeley did and
31 change “shell commands text” to “shell script”.
32 The filesystem tests are based on examining the return from a stat(2)
34 system call. The program checks to see if the file is empty, or if it's
35 some sort of special file. Any known file types appropriate to the sys‐
36 tem you are running on (sockets, symbolic links, or named pipes (FIFOs)
37 on those systems that implement them) are intuited if they are defined in
38 the system header file <sys/stat.h>.
39 The magic tests are used to check for files with data in particular fixed
41 formats. The canonical example of this is a binary executable (compiled
42 program) a.out file, whose format is defined in <elf.h>, <a.out.h> and
43 possibly <exec.h> in the standard include directory. These files have a
44 “magic number” stored in a particular place near the beginning of the
45 file that tells the UNIX operating system that the file is a binary exe‐
46 cutable, and which of several types thereof. The concept of a “magic”
47 has been applied by extension to data files. Any file with some invari‐
48 ant identifier at a small fixed offset into the file can usually be
49 described in this way. The information identifying these files is read
50 from the compiled magic file /usr/share/misc/magic.mgc, or the files in
51 the directory /usr/share/misc/magic if the compiled file does not exist.
52 In addition, if $HOME/.magic.mgc or $HOME/.magic exists, it will be used
53 in preference to the system magic files.
54 If a file does not match any of the entries in the magic file, it is
56 examined to see if it seems to be a text file. ASCII, ISO-8859-x, non-
57 ISO 8-bit extended-ASCII character sets (such as those used on Macintosh
58 and IBM PC systems), UTF-8-encoded Unicode, UTF-16-encoded Unicode, and
59 EBCDIC character sets can be distinguished by the different ranges and
60 sequences of bytes that constitute printable text in each set. If a file
61 passes any of these tests, its character set is reported. ASCII,
62 ISO-8859-x, UTF-8, and extended-ASCII files are identified as “text”
63 because they will be mostly readable on nearly any terminal; UTF-16 and
64 EBCDIC are only “character data” because, while they contain text, it is
65 text that will require translation before it can be read. In addition,
66 file will attempt to determine other characteristics of text-type files.
67 If the lines of a file are terminated by CR, CRLF, or NEL, instead of the
68 Unix-standard LF, this will be reported. Files that contain embedded
69 escape sequences or overstriking will also be identified.
70 Once file has determined the character set used in a text-type file, it
72 will attempt to determine in what language the file is written. The lan‐
73 guage tests look for particular strings (cf. <names.h>) that can appear
74 anywhere in the first few blocks of a file. For example, the keyword .br
75 indicates that the file is most likely a troff(1) input file, just as the
76 keyword struct indicates a C program. These tests are less reliable than
77 the previous two groups, so they are performed last. The language test
78 routines also test for some miscellany (such as tar(1) archives, JSON
79 files).
80 Any file that cannot be identified as having been written in any of the
82 character sets listed above is simply said to be “data”.
83
85 --apple
86 Causes the file command to output the file type and creator code
87 as used by older MacOS versions. The code consists of eight let‐
88 ters, the first describing the file type, the latter the creator.
89 the first describing the file type, the latter the creator. This
90 option works properly only for file formats that have the apple-
91 style output defined.
92 -b, --brief
94 Do not prepend filenames to output lines (brief mode).
95 -C, --compile
97 Write a magic.mgc output file that contains a pre-parsed version
98 of the magic file or directory.
99 -c, --checking-printout
101 Cause a checking printout of the parsed form of the magic file.
102 This is usually used in conjunction with the -m flag to debug a
103 new magic file before installing it.
104 -d Prints internal debugging information to stderr.
106 -E On filesystem errors (file not found etc), instead of handling
108 the error as regular output as POSIX mandates and keep going,
109 issue an error message and exit.
110 -e, --exclude testname
112 Exclude the test named in testname from the list of tests made to
113 determine the file type. Valid test names are:
114 apptype EMX application type (only on EMX).
116 ascii Various types of text files (this test will try to
118 guess the text encoding, irrespective of the setting of
119 the ‘encoding’ option).
120 encoding Different text encodings for soft magic tests.
122 tokens Ignored for backwards compatibility.
124 cdf Prints details of Compound Document Files.
126 compress Checks for, and looks inside, compressed files.
128 elf Prints ELF file details, provided soft magic tests are
130 enabled and the elf magic is found.
131 json Examines JSON (RFC-7159) files by parsing them for com‐
133 pliance.
134 soft Consults magic files.
136 tar Examines tar files by verifying the checksum of the 512
138 byte tar header. Excluding this test can provide more
139 detailed content description by using the soft magic
140 method.
141 text A synonym for ‘ascii’.
143 --extension
145 Print a slash-separated list of valid extensions for the file
146 type found.
147 -F, --separator separator
149 Use the specified string as the separator between the filename
150 and the file result returned. Defaults to ‘:’.
151 -f, --files-from namefile
153 Read the names of the files to be examined from namefile (one per
154 line) before the argument list. Either namefile or at least one
155 filename argument must be present; to test the standard input,
156 use ‘-’ as a filename argument. Please note that namefile is
157 unwrapped and the enclosed filenames are processed when this
158 option is encountered and before any further options processing
159 is done. This allows one to process multiple lists of files with
160 different command line arguments on the same file invocation.
161 Thus if you want to set the delimiter, you need to do it before
162 you specify the list of files, like: “-F @ -f namefile”, instead
163 of: “-f namefile -F @”.
164 -h, --no-dereference
166 option causes symlinks not to be followed (on systems that sup‐
167 port symbolic links). This is the default if the environment
168 variable POSIXLY_CORRECT is not defined.
169 -i, --mime
171 Causes the file command to output mime type strings rather than
172 the more traditional human readable ones. Thus it may say
173 ‘text/plain; charset=us-ascii’ rather than “ASCII text”.
174 --mime-type, --mime-encoding
176 Like -i, but print only the specified element(s).
177 -k, --keep-going
179 Don't stop at the first match, keep going. Subsequent matches
180 will be have the string ‘\012- ’ prepended. (If you want a new‐
181 line, see the -r option.) The magic pattern with the highest
182 strength (see the -l option) comes first.
183 -l, --list
185 Shows a list of patterns and their strength sorted descending by
186 magic(4) strength which is used for the matching (see also the -k
187 option).
188 -L, --dereference
190 option causes symlinks to be followed, as the like-named option
191 in ls(1) (on systems that support symbolic links). This is the
192 default if the environment variable POSIXLY_CORRECT is defined.
193 -m, --magic-file magicfiles
195 Specify an alternate list of files and directories containing
196 magic. This can be a single item, or a colon-separated list. If
197 a compiled magic file is found alongside a file or directory, it
198 will be used instead.
199 -N, --no-pad
201 Don't pad filenames so that they align in the output.
202 -n, --no-buffer
204 Force stdout to be flushed after checking each file. This is
205 only useful if checking a list of files. It is intended to be
206 used by programs that want filetype output from a pipe.
207 -p, --preserve-date
209 On systems that support utime(3) or utimes(2), attempt to pre‐
210 serve the access time of files analyzed, to pretend that file
211 never read them.
212 -P, --parameter name=value
214 Set various parameter limits.
215 Name Default Explanation
217 indir 15 recursion limit for indirect magic
218 name 30 use count limit for name/use magic
219 elf_notes 256 max ELF notes processed
220 elf_phnum 128 max ELF program sections processed
221 elf_shnum 32768 max ELF sections processed
222 regex 8192 length limit for regex searches
223 bytes 1048576 max number of bytes to read from
224 file
225 -r, --raw
227 Don't translate unprintable characters to \ooo. Normally file
228 translates unprintable characters to their octal representation.
229 -s, --special-files
231 Normally, file only attempts to read and determine the type of
232 argument files which stat(2) reports are ordinary files. This
233 prevents problems, because reading special files may have pecu‐
234 liar consequences. Specifying the -s option causes file to also
235 read argument files which are block or character special files.
236 This is useful for determining the filesystem types of the data
237 in raw disk partitions, which are block special files. This
238 option also causes file to disregard the file size as reported by
239 stat(2) since on some systems it reports a zero size for raw disk
240 partitions.
241 -S, --no-sandbox
243 On systems where libseccomp
244 (https://github.com/seccomp/libseccomp) is available, the -S flag
245 disables sandboxing which is enabled by default. This option is
246 needed for file to execute external descompressing programs, i.e.
247 when the -z flag is specified and the built-in decompressors are
248 not available.
249 -v, --version
251 Print the version of the program and exit.
252 -z, --uncompress
254 Try to look inside compressed files.
255 -Z, --uncompress-noreport
257 Try to look inside compressed files, but report information about
258 the contents only not the compression.
259 -0, --print0
261 Output a null character ‘\0’ after the end of the filename. Nice
262 to cut(1) the output. This does not affect the separator, which
263 is still printed.
264 If this option is repeated more than once, then file prints just
266 the filename followed by a NUL followed by the description (or
267 ERROR: text) followed by a second NUL for each entry.
268 --help Print a help message and exit.
270
272 The environment variable MAGIC can be used to set the default magic file
273 name. If that variable is set, then file will not attempt to open
274 $HOME/.magic. file adds “.mgc” to the value of this variable as appro‐
275 priate. The environment variable POSIXLY_CORRECT controls (on systems
276 that support symbolic links), whether file will attempt to follow sym‐
277 links or not. If set, then file follows symlink, otherwise it does not.
278 This is also controlled by the -L and -h options.
279
281 /usr/share/misc/magic.mgc Default compiled list of magic.
282 /usr/share/misc/magic Directory containing default magic files.
283
285 file will exit with 0 if the operation was successful or >0 if an error
286 was encountered. The following errors cause diagnostic messages, but
287 don't affect the program exit code (as POSIX requires), unless -E is
288 specified:
289 · A file cannot be found
290 · There is no permission to read a file
291 · The file type cannot be determined
292
294 $ file file.c file /dev/{wd0a,hda}
295 file.c: C program text
296 file: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV),
297 dynamically linked (uses shared libs), stripped
298 /dev/wd0a: block special (0/0)
299 /dev/hda: block special (3/0)
300 $ file -s /dev/wd0{b,d}
302 /dev/wd0b: data
303 /dev/wd0d: x86 boot sector
304 $ file -s /dev/hda{,1,2,3,4,5,6,7,8,9,10}
306 /dev/hda: x86 boot sector
307 /dev/hda1: Linux/i386 ext2 filesystem
308 /dev/hda2: x86 boot sector
309 /dev/hda3: x86 boot sector, extended partition table
310 /dev/hda4: Linux/i386 ext2 filesystem
311 /dev/hda5: Linux/i386 swap file
312 /dev/hda6: Linux/i386 swap file
313 /dev/hda7: Linux/i386 swap file
314 /dev/hda8: Linux/i386 swap file
315 /dev/hda9: empty
316 /dev/hda10: empty
317 $ file -i file.c file /dev/{wd0a,hda}
319 file.c: text/x-c
320 file: application/x-executable
321 /dev/hda: application/x-not-regular-file
322 /dev/wd0a: application/x-not-regular-file
323
326 hexdump(1), od(1), strings(1), magic(5)
327
329 This program is believed to exceed the System V Interface Definition of
330 FILE(CMD), as near as one can determine from the vague language contained
331 therein. Its behavior is mostly compatible with the System V program of
332 the same name. This version knows more magic, however, so it will pro‐
333 duce different (albeit more accurate) output in many cases.
334 The one significant difference between this version and System V is that
336 this version treats any white space as a delimiter, so that spaces in
337 pattern strings must be escaped. For example,
338 >10 string language impress (imPRESS data)
340 in an existing magic file would have to be changed to
342 >10 string language\ impress (imPRESS data)
344 In addition, in this version, if a pattern string contains a backslash,
346 it must be escaped. For example
347 0 string \begindata Andrew Toolkit document
349 in an existing magic file would have to be changed to
351 0 string \\begindata Andrew Toolkit document
353 SunOS releases 3.2 and later from Sun Microsystems include a file command
355 derived from the System V one, but with some extensions. This version
356 differs from Sun's only in minor ways. It includes the extension of the
357 ‘&’ operator, used as, for example,
358 >16 long&0x7fffffff >0 not stripped
360
362 On systems where libseccomp (https://github.com/seccomp/libseccomp) is
363 available, file is enforces limiting system calls to only the ones neces‐
364 sary for the operation of the program. This enforcement does not provide
365 any security benefit when file is asked to decompress input files running
366 external programs with the -z option. To enable execution of external
367 decompressors, one needs to disable sandboxing using the -S flag.
368
370 The magic file entries have been collected from various sources, mainly
371 USENET, and contributed by various authors. Christos Zoulas (address
372 below) will collect additional or corrected magic file entries. A con‐
373 solidation of magic file entries will be distributed periodically.
374 The order of entries in the magic file is significant. Depending on what
376 system you are using, the order that they are put together may be incor‐
377 rect. If your old file command uses a magic file, keep the old magic
378 file around for comparison purposes (rename it to
379 /usr/share/misc/magic.orig).
380
382 There has been a file command in every UNIX since at least Research
383 Version 4 (man page dated November, 1973). The System V version intro‐
384 duced one significant major change: the external list of magic types.
385 This slowed the program down slightly but made it a lot more flexible.
386 This program, based on the System V version, was written by Ian Darwin
388 ⟨ian@darwinsys.com⟩ without looking at anybody else's source code.
389 John Gilmore revised the code extensively, making it better than the
391 first version. Geoff Collyer found several inadequacies and provided
392 some magic file entries. Contributions of the ‘&’ operator by Rob McMa‐
393 hon, ⟨cudcv@warwick.ac.uk⟩, 1989.
394 Guy Harris, ⟨guy@netapp.com⟩, made many changes from 1993 to the present.
396 Primary development and maintenance from 1990 to the present by Christos
398 Zoulas ⟨christos@astron.com⟩.
399 Altered by Chris Lowth ⟨chris@lowth.com⟩, 2000: handle the -i option to
401 output mime type strings, using an alternative magic file and internal
402 logic.
403 Altered by Eric Fischer ⟨enf@pobox.com⟩, July, 2000, to identify charac‐
405 ter codes and attempt to identify the languages of non-ASCII files.
406 Altered by Reuben Thomas ⟨rrt@sc3d.org⟩, 2007-2011, to improve MIME sup‐
408 port, merge MIME and non-MIME magic, support directories as well as files
409 of magic, apply many bug fixes, update and fix a lot of magic, improve
410 the build system, improve the documentation, and rewrite the Python bind‐
411 ings in pure Python.
412 The list of contributors to the ‘magic’ directory (magic files) is too
414 long to include here. You know who you are; thank you. Many contribu‐
415 tors are listed in the source files.
416
418 Copyright (c) Ian F. Darwin, Toronto, Canada, 1986-1999. Covered by the
419 standard Berkeley Software Distribution copyright; see the file COPYING
420 in the source distribution.
421 The files tar.h and is_tar.c were written by John Gilmore from his pub‐
423 lic-domain tar(1) program, and are not covered by the above license.
424
426 Please report bugs and send patches to the bug tracker at
427 https://bugs.astron.com/ or the mailing list at ⟨file@astron.com⟩ (visit
428 https://mailman.astron.com/mailman/listinfo/file first to subscribe).
429
431 Fix output so that tests for MIME and APPLE flags are not needed all over
432 the place, and actual output is only done in one place. This needs a
433 design. Suggestion: push possible outputs on to a list, then pick the
434 last-pushed (most specific, one hopes) value at the end, or use a default
435 if the list is empty. This should not slow down evaluation.
436 The handling of MAGIC_CONTINUE and printing \012- between entries is
438 clumsy and complicated; refactor and centralize.
439 Some of the encoding logic is hard-coded in encoding.c and can be moved
441 to the magic files if we had a !:charset annotation
442 Continue to squash all magic bugs. See Debian BTS for a good source.
444 Store arbitrarily long strings, for example for %s patterns, so that they
446 can be printed out. Fixes Debian bug #271672. This can be done by allo‐
447 cating strings in a string pool, storing the string pool at the end of
448 the magic file and converting all the string pointers to relative offsets
449 from the string pool.
450 Add syntax for relative offsets after current level (Debian bug #466037).
452 Make file -ki work, i.e. give multiple MIME types.
454 Add a zip library so we can peek inside Office2007 documents to print
456 more details about their contents.
457 Add an option to print URLs for the sources of the file descriptions.
459 Combine script searches and add a way to map executable names to MIME
461 types (e.g. have a magic value for !:mime which causes the resulting
462 string to be looked up in a table). This would avoid adding the same
463 magic repeatedly for each new hash-bang interpreter.
464 When a file descriptor is available, we can skip and adjust the buffer
466 instead of the hacky buffer management we do now.
467 Fix “name” and “use” to check for consistency at compile time (duplicate
469 “name”, “use” pointing to undefined “name” ). Make “name” / “use” more
470 efficient by keeping a sorted list of names. Special-case ^ to flip
471 endianness in the parser so that it does not have to be escaped, and doc‐
472 ument it.
473 If the offsets specified internally in the file exceed the buffer size (
475 HOWMANY variable in file.h), then we don't seek to that offset, but we
476 give up. It would be better if buffer managements was done when the file
477 descriptor is available so move around the file. One must be careful
478 though because this has performance (and thus security considerations).
479
481 You can obtain the original author's latest version by anonymous FTP on
482 ftp.astron.com in the directory /pub/file/file-X.YZ.tar.gz.
483BSD February 18, 2019 BSD