1.\" $File: file.man,v 1.138 2019/10/15 18:00:40 christos Exp $ 2.Dd July 13, 2019 3.Dt FILE __CSECTION__ 4.Os 5.Sh NAME 6.Nm file 7.Nd determine file type 8.Sh SYNOPSIS 9.Nm 10.Bk -words 11.Op Fl bcdEhiklLNnprsSvzZ0 12.Op Fl Fl apple 13.Op Fl Fl extension 14.Op Fl Fl mime-encoding 15.Op Fl Fl mime-type 16.Op Fl e Ar testname 17.Op Fl F Ar separator 18.Op Fl f Ar namefile 19.Op Fl m Ar magicfiles 20.Op Fl P Ar name=value 21.Ar 22.Ek 23.Nm 24.Fl C 25.Op Fl m Ar magicfiles 26.Nm 27.Op Fl Fl help 28.Sh DESCRIPTION 29This manual page documents version __VERSION__ of the 30.Nm 31command. 32.Pp 33.Nm 34tests each argument in an attempt to classify it. 35There are three sets of tests, performed in this order: 36filesystem tests, magic tests, and language tests. 37The 38.Em first 39test that succeeds causes the file type to be printed. 40.Pp 41The type printed will usually contain one of the words 42.Em text 43(the file contains only 44printing characters and a few common control 45characters and is probably safe to read on an 46.Dv ASCII 47terminal), 48.Em executable 49(the file contains the result of compiling a program 50in a form understandable to some 51.Tn UNIX 52kernel or another), 53or 54.Em data 55meaning anything else (data is usually 56.Dq binary 57or non-printable). 58Exceptions are well-known file formats (core files, tar archives) 59that are known to contain binary data. 60When modifying magic files or the program itself, make sure to 61.Em "preserve these keywords" . 62Users depend on knowing that all the readable files in a directory 63have the word 64.Dq text 65printed. 66Don't do as Berkeley did and change 67.Dq shell commands text 68to 69.Dq shell script . 70.Pp 71The filesystem tests are based on examining the return from a 72.Xr stat 2 73system call. 74The program checks to see if the file is empty, 75or if it's some sort of special file. 76Any known file types appropriate to the system you are running on 77(sockets, symbolic links, or named pipes (FIFOs) on those systems that 78implement them) 79are intuited if they are defined in the system header file 80.In sys/stat.h . 81.Pp 82The magic tests are used to check for files with data in 83particular fixed formats. 84The canonical example of this is a binary executable (compiled program) 85.Dv a.out 86file, whose format is defined in 87.In elf.h , 88.In a.out.h 89and possibly 90.In exec.h 91in the standard include directory. 92These files have a 93.Dq "magic number" 94stored in a particular place 95near the beginning of the file that tells the 96.Tn UNIX 97operating system 98that the file is a binary executable, and which of several types thereof. 99The concept of a 100.Dq "magic" 101has been applied by extension to data files. 102Any file with some invariant identifier at a small fixed 103offset into the file can usually be described in this way. 104The information identifying these files is read from the compiled 105magic file 106.Pa __MAGIC__.mgc , 107or the files in the directory 108.Pa __MAGIC__ 109if the compiled file does not exist. 110In addition, if 111.Pa $HOME/.magic.mgc 112or 113.Pa $HOME/.magic 114exists, it will be used in preference to the system magic files. 115.Pp 116If a file does not match any of the entries in the magic file, 117it is examined to see if it seems to be a text file. 118ASCII, ISO-8859-x, non-ISO 8-bit extended-ASCII character sets 119(such as those used on Macintosh and IBM PC systems), 120UTF-8-encoded Unicode, UTF-16-encoded Unicode, and EBCDIC 121character sets can be distinguished by the different 122ranges and sequences of bytes that constitute printable text 123in each set. 124If a file passes any of these tests, its character set is reported. 125ASCII, ISO-8859-x, UTF-8, and extended-ASCII files are identified 126as 127.Dq text 128because they will be mostly readable on nearly any terminal; 129UTF-16 and EBCDIC are only 130.Dq character data 131because, while 132they contain text, it is text that will require translation 133before it can be read. 134In addition, 135.Nm 136will attempt to determine other characteristics of text-type files. 137If the lines of a file are terminated by CR, CRLF, or NEL, instead 138of the Unix-standard LF, this will be reported. 139Files that contain embedded escape sequences or overstriking 140will also be identified. 141.Pp 142Once 143.Nm 144has determined the character set used in a text-type file, 145it will 146attempt to determine in what language the file is written. 147The language tests look for particular strings (cf. 148.In names.h ) 149that can appear anywhere in the first few blocks of a file. 150For example, the keyword 151.Em .br 152indicates that the file is most likely a 153.Xr troff 1 154input file, just as the keyword 155.Em struct 156indicates a C program. 157These tests are less reliable than the previous 158two groups, so they are performed last. 159The language test routines also test for some miscellany 160(such as 161.Xr tar 1 162archives, JSON files). 163.Pp 164Any file that cannot be identified as having been written 165in any of the character sets listed above is simply said to be 166.Dq data . 167.Sh OPTIONS 168.Bl -tag -width indent 169.It Fl Fl apple 170Causes the file command to output the file type and creator code as 171used by older MacOS versions. 172The code consists of eight letters, 173the first describing the file type, the latter the creator. 174This option works properly only for file formats that have the 175apple-style output defined. 176.It Fl b , Fl Fl brief 177Do not prepend filenames to output lines (brief mode). 178.It Fl C , Fl Fl compile 179Write a 180.Pa magic.mgc 181output file that contains a pre-parsed version of the magic file or directory. 182.It Fl c , Fl Fl checking-printout 183Cause a checking printout of the parsed form of the magic file. 184This is usually used in conjunction with the 185.Fl m 186flag to debug a new magic file before installing it. 187.It Fl d 188Prints internal debugging information to stderr. 189.It Fl E 190On filesystem errors (file not found etc), instead of handling the error 191as regular output as POSIX mandates and keep going, issue an error message 192and exit. 193.It Fl e , Fl Fl exclude Ar testname 194Exclude the test named in 195.Ar testname 196from the list of tests made to determine the file type. 197Valid test names are: 198.Bl -tag -width compress 199.It apptype 200.Dv EMX 201application type (only on EMX). 202.It ascii 203Various types of text files (this test will try to guess the text 204encoding, irrespective of the setting of the 205.Sq encoding 206option). 207.It encoding 208Different text encodings for soft magic tests. 209.It tokens 210Ignored for backwards compatibility. 211.It cdf 212Prints details of Compound Document Files. 213.It compress 214Checks for, and looks inside, compressed files. 215.It csv 216Checks Comma Separated Value files. 217.It elf 218Prints ELF file details, provided soft magic tests are enabled and the 219elf magic is found. 220.It json 221Examines JSON (RFC-7159) files by parsing them for compliance. 222.It soft 223Consults magic files. 224.It tar 225Examines tar files by verifying the checksum of the 512 byte tar header. 226Excluding this test can provide more detailed content description by using 227the soft magic method. 228.It text 229A synonym for 230.Sq ascii . 231.El 232.It Fl Fl extension 233Print a slash-separated list of valid extensions for the file type found. 234.It Fl F , Fl Fl separator Ar separator 235Use the specified string as the separator between the filename and the 236file result returned. 237Defaults to 238.Sq \&: . 239.It Fl f , Fl Fl files-from Ar namefile 240Read the names of the files to be examined from 241.Ar namefile 242(one per line) 243before the argument list. 244Either 245.Ar namefile 246or at least one filename argument must be present; 247to test the standard input, use 248.Sq - 249as a filename argument. 250Please note that 251.Ar namefile 252is unwrapped and the enclosed filenames are processed when this option is 253encountered and before any further options processing is done. 254This allows one to process multiple lists of files with different command line 255arguments on the same 256.Nm 257invocation. 258Thus if you want to set the delimiter, you need to do it before you specify 259the list of files, like: 260.Dq Fl F Ar @ Fl f Ar namefile , 261instead of: 262.Dq Fl f Ar namefile Fl F Ar @ . 263.It Fl h , Fl Fl no-dereference 264option causes symlinks not to be followed 265(on systems that support symbolic links). 266This is the default if the environment variable 267.Dv POSIXLY_CORRECT 268is not defined. 269.It Fl i , Fl Fl mime 270Causes the file command to output mime type strings rather than the more 271traditional human readable ones. 272Thus it may say 273.Sq text/plain; charset=us-ascii 274rather than 275.Dq ASCII text . 276.It Fl Fl mime-type , Fl Fl mime-encoding 277Like 278.Fl i , 279but print only the specified element(s). 280.It Fl k , Fl Fl keep-going 281Don't stop at the first match, keep going. 282Subsequent matches will be 283have the string 284.Sq "\[rs]012\- " 285prepended. 286(If you want a newline, see the 287.Fl r 288option.) 289The magic pattern with the highest strength (see the 290.Fl l 291option) comes first. 292.It Fl l , Fl Fl list 293Shows a list of patterns and their strength sorted descending by 294.Xr magic __FSECTION__ 295strength 296which is used for the matching (see also the 297.Fl k 298option). 299.It Fl L , Fl Fl dereference 300option causes symlinks to be followed, as the like-named option in 301.Xr ls 1 302(on systems that support symbolic links). 303This is the default if the environment variable 304.Ev POSIXLY_CORRECT 305is defined. 306.It Fl m , Fl Fl magic-file Ar magicfiles 307Specify an alternate list of files and directories containing magic. 308This can be a single item, or a colon-separated list. 309If a compiled magic file is found alongside a file or directory, 310it will be used instead. 311.It Fl N , Fl Fl no-pad 312Don't pad filenames so that they align in the output. 313.It Fl n , Fl Fl no-buffer 314Force stdout to be flushed after checking each file. 315This is only useful if checking a list of files. 316It is intended to be used by programs that want filetype output from a pipe. 317.It Fl p , Fl Fl preserve-date 318On systems that support 319.Xr utime 3 320or 321.Xr utimes 2 , 322attempt to preserve the access time of files analyzed, to pretend that 323.Nm 324never read them. 325.It Fl P , Fl Fl parameter Ar name=value 326Set various parameter limits. 327.Bl -column "elf_phnum" "Default" "XXXXXXXXXXXXXXXXXXXXXXXXXXX" -offset indent 328.It Sy "Name" Ta Sy "Default" Ta Sy "Explanation" 329.It Li indir Ta 15 Ta recursion limit for indirect magic 330.It Li name Ta 30 Ta use count limit for name/use magic 331.It Li elf_notes Ta 256 Ta max ELF notes processed 332.It Li elf_phnum Ta 128 Ta max ELF program sections processed 333.It Li elf_shnum Ta 32768 Ta max ELF sections processed 334.It Li regex Ta 8192 Ta length limit for regex searches 335.It Li bytes Ta 1048576 Ta max number of bytes to read from file 336.El 337.It Fl r , Fl Fl raw 338Don't translate unprintable characters to \eooo. 339Normally 340.Nm 341translates unprintable characters to their octal representation. 342.It Fl s , Fl Fl special-files 343Normally, 344.Nm 345only attempts to read and determine the type of argument files which 346.Xr stat 2 347reports are ordinary files. 348This prevents problems, because reading special files may have peculiar 349consequences. 350Specifying the 351.Fl s 352option causes 353.Nm 354to also read argument files which are block or character special files. 355This is useful for determining the filesystem types of the data in raw 356disk partitions, which are block special files. 357This option also causes 358.Nm 359to disregard the file size as reported by 360.Xr stat 2 361since on some systems it reports a zero size for raw disk partitions. 362.It Fl S , Fl Fl no-sandbox 363On systems where libseccomp 364.Pa ( https://github.com/seccomp/libseccomp ) 365is available, the 366.Fl S 367flag disables sandboxing which is enabled by default. 368This option is needed for file to execute external decompressing programs, 369i.e. when the 370.Fl z 371flag is specified and the built-in decompressors are not available. 372On systems where sandboxing is not available, this option has no effect. 373.It Fl v , Fl Fl version 374Print the version of the program and exit. 375.It Fl z , Fl Fl uncompress 376Try to look inside compressed files. 377.It Fl Z , Fl Fl uncompress-noreport 378Try to look inside compressed files, but report information about the contents 379only not the compression. 380.It Fl 0 , Fl Fl print0 381Output a null character 382.Sq \e0 383after the end of the filename. 384Nice to 385.Xr cut 1 386the output. 387This does not affect the separator, which is still printed. 388.Pp 389If this option is repeated more than once, then 390.Nm 391prints just the filename followed by a NUL followed by the description 392(or ERROR: text) followed by a second NUL for each entry. 393.It Fl -help 394Print a help message and exit. 395.El 396.Sh ENVIRONMENT 397The environment variable 398.Ev MAGIC 399can be used to set the default magic file name. 400If that variable is set, then 401.Nm 402will not attempt to open 403.Pa $HOME/.magic . 404.Nm 405adds 406.Dq Pa .mgc 407to the value of this variable as appropriate. 408The environment variable 409.Ev POSIXLY_CORRECT 410controls (on systems that support symbolic links), whether 411.Nm 412will attempt to follow symlinks or not. 413If set, then 414.Nm 415follows symlink, otherwise it does not. 416This is also controlled by the 417.Fl L 418and 419.Fl h 420options. 421.Sh FILES 422.Bl -tag -width __MAGIC__.mgc -compact 423.It Pa __MAGIC__.mgc 424Default compiled list of magic. 425.It Pa __MAGIC__ 426Directory containing default magic files. 427.El 428.Sh EXIT STATUS 429.Nm 430will exit with 431.Dv 0 432if the operation was successful or 433.Dv >0 434if an error was encountered. 435The following errors cause diagnostic messages, but don't affect the program 436exit code (as POSIX requires), unless 437.Fl E 438is specified: 439.Bl -bullet -compact -offset indent 440.It 441A file cannot be found 442.It 443There is no permission to read a file 444.It 445The file type cannot be determined 446.El 447.Sh EXAMPLES 448.Bd -literal -offset indent 449$ file file.c file /dev/{wd0a,hda} 450file.c: C program text 451file: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), 452 dynamically linked (uses shared libs), stripped 453/dev/wd0a: block special (0/0) 454/dev/hda: block special (3/0) 455 456$ file -s /dev/wd0{b,d} 457/dev/wd0b: data 458/dev/wd0d: x86 boot sector 459 460$ file -s /dev/hda{,1,2,3,4,5,6,7,8,9,10} 461/dev/hda: x86 boot sector 462/dev/hda1: Linux/i386 ext2 filesystem 463/dev/hda2: x86 boot sector 464/dev/hda3: x86 boot sector, extended partition table 465/dev/hda4: Linux/i386 ext2 filesystem 466/dev/hda5: Linux/i386 swap file 467/dev/hda6: Linux/i386 swap file 468/dev/hda7: Linux/i386 swap file 469/dev/hda8: Linux/i386 swap file 470/dev/hda9: empty 471/dev/hda10: empty 472 473$ file -i file.c file /dev/{wd0a,hda} 474file.c: text/x-c 475file: application/x-executable 476/dev/hda: application/x-not-regular-file 477/dev/wd0a: application/x-not-regular-file 478 479.Ed 480.Sh SEE ALSO 481.Xr hexdump 1 , 482.Xr od 1 , 483.Xr strings 1 , 484.Xr magic __FSECTION__ , 485.Xr fstyp 8 486.Sh STANDARDS CONFORMANCE 487This program is believed to exceed the System V Interface Definition 488of FILE(CMD), as near as one can determine from the vague language 489contained therein. 490Its behavior is mostly compatible with the System V program of the same name. 491This version knows more magic, however, so it will produce 492different (albeit more accurate) output in many cases. 493.\" URL: http://www.opengroup.org/onlinepubs/009695399/utilities/file.html 494.Pp 495The one significant difference 496between this version and System V 497is that this version treats any white space 498as a delimiter, so that spaces in pattern strings must be escaped. 499For example, 500.Bd -literal -offset indent 501\*[Gt]10 string language impress\ (imPRESS data) 502.Ed 503.Pp 504in an existing magic file would have to be changed to 505.Bd -literal -offset indent 506\*[Gt]10 string language\e impress (imPRESS data) 507.Ed 508.Pp 509In addition, in this version, if a pattern string contains a backslash, 510it must be escaped. 511For example 512.Bd -literal -offset indent 5130 string \ebegindata Andrew Toolkit document 514.Ed 515.Pp 516in an existing magic file would have to be changed to 517.Bd -literal -offset indent 5180 string \e\ebegindata Andrew Toolkit document 519.Ed 520.Pp 521SunOS releases 3.2 and later from Sun Microsystems include a 522.Nm 523command derived from the System V one, but with some extensions. 524This version differs from Sun's only in minor ways. 525It includes the extension of the 526.Sq \*[Am] 527operator, used as, 528for example, 529.Bd -literal -offset indent 530\*[Gt]16 long\*[Am]0x7fffffff \*[Gt]0 not stripped 531.Ed 532.Sh SECURITY 533On systems where libseccomp 534.Pa ( https://github.com/seccomp/libseccomp ) 535is available, 536.Nm 537is enforces limiting system calls to only the ones necessary for the 538operation of the program. 539This enforcement does not provide any security benefit when 540.Nm 541is asked to decompress input files running external programs with 542the 543.Fl z 544option. 545To enable execution of external decompressors, one needs to disable 546sandboxing using the 547.Fl S 548flag. 549.Sh MAGIC DIRECTORY 550The magic file entries have been collected from various sources, 551mainly USENET, and contributed by various authors. 552Christos Zoulas (address below) will collect additional 553or corrected magic file entries. 554A consolidation of magic file entries 555will be distributed periodically. 556.Pp 557The order of entries in the magic file is significant. 558Depending on what system you are using, the order that 559they are put together may be incorrect. 560If your old 561.Nm 562command uses a magic file, 563keep the old magic file around for comparison purposes 564(rename it to 565.Pa __MAGIC__.orig ) . 566.Sh HISTORY 567There has been a 568.Nm 569command in every 570.Dv UNIX since at least Research Version 4 571(man page dated November, 1973). 572The System V version introduced one significant major change: 573the external list of magic types. 574This slowed the program down slightly but made it a lot more flexible. 575.Pp 576This program, based on the System V version, 577was written by Ian Darwin 578.Aq ian@darwinsys.com 579without looking at anybody else's source code. 580.Pp 581John Gilmore revised the code extensively, making it better than 582the first version. 583Geoff Collyer found several inadequacies 584and provided some magic file entries. 585Contributions of the 586.Sq \*[Am] 587operator by Rob McMahon, 588.Aq cudcv@warwick.ac.uk , 5891989. 590.Pp 591Guy Harris, 592.Aq guy@netapp.com , 593made many changes from 1993 to the present. 594.Pp 595Primary development and maintenance from 1990 to the present by 596Christos Zoulas 597.Aq christos@astron.com . 598.Pp 599Altered by Chris Lowth 600.Aq chris@lowth.com , 6012000: handle the 602.Fl i 603option to output mime type strings, using an alternative 604magic file and internal logic. 605.Pp 606Altered by Eric Fischer 607.Aq enf@pobox.com , 608July, 2000, 609to identify character codes and attempt to identify the languages 610of non-ASCII files. 611.Pp 612Altered by Reuben Thomas 613.Aq rrt@sc3d.org , 6142007-2011, to improve MIME support, merge MIME and non-MIME magic, 615support directories as well as files of magic, apply many bug fixes, 616update and fix a lot of magic, improve the build system, improve the 617documentation, and rewrite the Python bindings in pure Python. 618.Pp 619The list of contributors to the 620.Sq magic 621directory (magic files) 622is too long to include here. 623You know who you are; thank you. 624Many contributors are listed in the source files. 625.Sh LEGAL NOTICE 626Copyright (c) Ian F. Darwin, Toronto, Canada, 1986-1999. 627Covered by the standard Berkeley Software Distribution copyright; see the file 628COPYING in the source distribution. 629.Pp 630The files 631.Pa tar.h 632and 633.Pa is_tar.c 634were written by John Gilmore from his public-domain 635.Xr tar 1 636program, and are not covered by the above license. 637.Sh BUGS 638Please report bugs and send patches to the bug tracker at 639.Pa https://bugs.astron.com/ 640or the mailing list at 641.Aq file@astron.com 642(visit 643.Pa https://mailman.astron.com/mailman/listinfo/file 644first to subscribe). 645.Sh TODO 646Fix output so that tests for MIME and APPLE flags are not needed all 647over the place, and actual output is only done in one place. 648This needs a design. 649Suggestion: push possible outputs on to a list, then pick the 650last-pushed (most specific, one hopes) value at the end, or 651use a default if the list is empty. 652This should not slow down evaluation. 653.Pp 654The handling of 655.Dv MAGIC_CONTINUE 656and printing \e012- between entries is clumsy and complicated; refactor 657and centralize. 658.Pp 659Some of the encoding logic is hard-coded in encoding.c and can be moved 660to the magic files if we had a !:charset annotation 661.Pp 662Continue to squash all magic bugs. 663See Debian BTS for a good source. 664.Pp 665Store arbitrarily long strings, for example for %s patterns, so that 666they can be printed out. 667Fixes Debian bug #271672. 668This can be done by allocating strings in a string pool, storing the 669string pool at the end of the magic file and converting all the string 670pointers to relative offsets from the string pool. 671.Pp 672Add syntax for relative offsets after current level (Debian bug #466037). 673.Pp 674Make file -ki work, i.e. give multiple MIME types. 675.Pp 676Add a zip library so we can peek inside Office2007 documents to 677print more details about their contents. 678.Pp 679Add an option to print URLs for the sources of the file descriptions. 680.Pp 681Combine script searches and add a way to map executable names to MIME 682types (e.g. have a magic value for !:mime which causes the resulting 683string to be looked up in a table). 684This would avoid adding the same magic repeatedly for each new 685hash-bang interpreter. 686.Pp 687When a file descriptor is available, we can skip and adjust the buffer 688instead of the hacky buffer management we do now. 689.Pp 690Fix 691.Dq name 692and 693.Dq use 694to check for consistency at compile time (duplicate 695.Dq name , 696.Dq use 697pointing to undefined 698.Dq name 699). 700Make 701.Dq name 702/ 703.Dq use 704more efficient by keeping a sorted list of names. 705Special-case ^ to flip endianness in the parser so that it does not 706have to be escaped, and document it. 707.Pp 708If the offsets specified internally in the file exceed the buffer size 709( 710.Dv HOWMANY 711variable in file.h), then we don't seek to that offset, but we give up. 712It would be better if buffer managements was done when the file descriptor 713is available so move around the file. 714One must be careful though because this has performance (and thus security 715considerations). 716.Sh AVAILABILITY 717You can obtain the original author's latest version by anonymous FTP 718on 719.Pa ftp.astron.com 720in the directory 721.Pa /pub/file/file-X.YZ.tar.gz . 722