The man pages, annotated
Level: reference · for anyone who wants the primary documentation of text on their own machine, with every word on it explained
One line: 169 manual pages about text and files, 140 from this Mac and 29 from Ubuntu, dumped verbatim with the date each one carries and the machine it came from, then read in 28 family pages that quote each page, explain its terms, check its claims against both machines, and link the lesson here that owns each idea.
What is in this folder
The encoding man pages nobody opens made the case that a stock Mac ships about forty pages of primary material on this subject and that nobody opens them. This folder opens them, and the hundred and thirty others that turn out to be about text once you ask which pages on the machine decide what a character is. It has three layers.
| Layer |
Where |
What it is |
| The dumps |
raw/macos/, raw/linux/ |
169 text files, each exactly what man SECTION NAME prints, one per page. Never edited; regenerated by two scripts |
| The provenance |
raw/PROVENANCE-macos.txt, raw/PROVENANCE-linux.txt |
Which machine, which library version, which command, which day. A man page is documentation of a machine, so the machine is recorded beside it |
| The family pages |
the 28 .md files listed below |
One page per group of related man pages. Each quotes the passages that matter, explains the terms in a table, measures the page's claims on both machines in dated fences, says where the page is stale or silent, and links the lessons |
A family page is laid out the same way every time, so you can skip to the part you want: What the pages are for (who wrote it, when, which section, why it matters here), The page, with notes (verbatim excerpts, each followed by commentary), Terms on the page (a table: the word, what it means, where this library explains it), Try it on your machine (commands and what they printed, on macOS 26.6.2 and on ubuntu:24.04), Where the page is dated, and what it does not say, and See also.
To look a page up, find its name in the tables below. The first column links the verbatim dump, the heading links the family page that reads it.
How the dumps were made, and how to check them
dump.sh renders each macOS page with man SECTION NAME | col -bx and writes it to raw/macos/NAME.SECTION.txt. Into a pipe, man on this Mac runs mandoc at 78 columns and marks bold and underline the way a 1970s teletype did, by printing the character, a backspace, and the character again; col -bx removes those backspaces and turns tabs into spaces, and does nothing else. The man(1) family page shows the raw bytes. A page that man finds under another name is dumped under the name you would type: man 1 base64 on this Mac prints the page called bintrans(1), so it is here as base64.1.txt.
dump_linux.sh does the same for the 29 Linux pages in an ubuntu:24.04 container. A stock Ubuntu image ships no man pages at all: its dpkg is configured to drop /usr/share/man on every install, and its /usr/bin/man is a stub that says so. The script builds an image that lifts that exclusion, installs man-db and the manpages packages (release 6.7 of the Linux man-pages project), and renders with the real binary, which the stub leaves at /usr/bin/man.REAL. The experiments in the family pages ran in a second image built on that one with gcc, xxd and python3 added; both Dockerfiles are in the scripts and the provenance files.
The locale decides the bytes, so both scripts pin it. dump.sh runs under LC_ALL=C, and the 140 macOS dumps contain no byte above 0x7f: mandoc writes its special-character escapes as ASCII, EUR for the euro and 'e for an e-acute, and the dash between a page's name and its description as -. Read the same page in a UTF-8 terminal and that dash is –, and man 7 mandoc_char gains 1,028 bytes of real glyphs. dump_linux.sh runs under C.UTF-8, the locale a stock Ubuntu ships, so the 29 Linux dumps are the UTF-8 rendering and hold 9,420 bytes above 0x7f: man-db's hyphens, dashes and bullets. Two renderings of one page are a finding in their own right, measured on the man(1) page, which also shows why the pinning matters on the Mac: under C, this Mac's col deletes any byte above 0x7f it is given.
The dumps are verbatim so that they can be checked. On the same machine, bash 13_Documentation/manual_pages/dump.sh followed by git diff --stat 13_Documentation/manual_pages/raw/ should print nothing. On a different machine it will print the pages that differ, which is the point: a man page has a date, and so does the system it describes.
raw/PROVENANCE-macos.txt and raw/PROVENANCE-linux.txt, as written by the two scripts on 2026-09-13macOS dump: 2026-09-13
macOS 26.6.2 (x86_64), Darwin 25.6.0
man: /Library/Developer/CommandLineTools/SDKs/MacOSX.sdk/usr/share/man/man5/utf8.5
rendered with: LC_ALL=C man SECTION NAME | col -bx (mandoc, 78 columns, ASCII rendering)
pages: 140 dumped, 0 missing
Linux dump: 2026-09-13
Ubuntu 24.04.4 LTS in Docker, image enc-man (built from ubuntu:24.04)
manpages 6.7-2, man-db 2.12.0-4build2, glibc 2.39-0ubuntu8.9
rendered with: LANG=C.UTF-8 MANWIDTH=78 man.REAL -P cat SECTION NAME | col -bx (man-db, UTF-8 rendering)
pages: 29 dumped, 0 missing
Note the first path: man on this Mac serves most section-3 pages from the Command Line Tools SDK, not from /usr/share/man. The SDK is documentation of the headers you compile against, which is the right thing to read, and it is also why a machine without Xcode's tools has fewer of these pages.
Which pages, and why these
The rule was: a page is in if it is about text, files as sequences of bytes, or the tables and settings that turn one into the other. That gave four groups.
- The encodings themselves. BSD documents each encoding as a file format in section 5, one page each; Linux documents them as essays in section 7 under different names. Both sets are here, and the
utf8(5) page puts the two side by side.
- The C library. Every command-line tool that has an opinion about characters got it from libc, so the pages for the functions that decode (
multibyte(3)), classify (ctype(3)), measure (wcwidth(3)), order (strcoll(3)), convert (iconv(3)), escape (vis(3)) and read (stdio(3)) text are here, with the locale pages that give them their meaning.
- The tools. The commands the 11_Tools chapter measures, read here as pages: what each flag promises, in bytes or in characters, and where the page is silent.
- The formats.
magic(5), tar(5), the uuencode format on bintrans(1), and the format of the manual itself on mandoc_char(7).
Some pages on the machine were left out on purpose. textutil(1), plutil(1), plist(5), core(5) and sh(1) are Apple's, all rights reserved, with no licence that allows copying them into a public repository; read them with man on your own Mac. The Perl documentation (perlunitut, perluniintro, perlunicode, Encode::Supported) is a graded Unicode course in its own right, already recommended on the lesson page, and too large to annotate here. Pages for tools that are not on a stock machine (rg, hexyl, uchardet, recode, dos2unix) belong with The five worth installing.
What the licences allow
Every dump in raw/macos/ except three is a rendering of a BSD-licensed manual page: from the Regents of the University of California (the 1980s and 1990s pages), Tim J. Robbins (the multibyte and CJK pages, 2002 to 2004), the Citrus Project and Gabor Kovesdan (iconv), Henry Spencer (regex, re_format), Kristaps Dzonsons and Ingo Schwarze (mandoc, ISC licence), Tim Kientzle (tar(5), cpio(5)), Ian F. Darwin and Christos Zoulas (file, magic), and the vim project (xxd, under the vim licence). The BSD licence permits redistribution in any form provided the copyright notice, the conditions and the disclaimer are reproduced with the distribution. The rendered pages do not carry those notices; they are comments in the roff sources, at the paths man -w SECTION NAME prints on a Mac and which the provenance file records, and the notice common to them is this:
The BSD notice, as it appears in the roff source of utf8.5 and most of the pages hereCopyright (c) 1993
The Regents of the University of California. All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
3. Neither the name of the University nor the names of its contributors
may be used to endorse or promote products derived from this software
without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
SUCH DAMAGE.
The three exceptions are Apple's own: locale(1) (Darwin, 2004), the xlocale(3) family (newlocale, uselocale, duplocale, querylocale, 2005) and strings(1) from cctools. Those ship in Apple's open-source releases of Libc and cctools under the Apple Public Source License 2.0 ↗, which permits redistribution of unmodified copies with that notice. The 29 Linux pages are from release 6.7 of the Linux man-pages project ↗, whose pages are licensed page by page under the GPL version 2 or later or under the project's own verbatim-copying licence; each page's roff source carries its notice, and the rendered text does not. None of this is legal advice; it is where the notices are.
The families
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
utf8(5) |
|
macOS |
April 7, 2004 |
UTF-8, a transformation format of ISO 10646 |
utf2(5) |
|
macOS |
October 11, 2002 |
Universal character set Transformation Format encoding of runes |
utf-8(7) |
UTF-8 |
Linux |
2024-03-14 |
an ASCII compatible multibyte Unicode encoding |
unicode(7) |
|
Linux |
2024-01-28 |
universal character set |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
euc(5) |
|
macOS |
November 8, 2003 |
EUC encoding of wide characters |
big5(5) |
|
macOS |
August 7, 2003 |
"Big Five" encoding for Traditional Chinese text |
gb2312(5) |
|
macOS |
November 7, 2003 |
GB2312 encoding method for Chinese text |
gbk(5) |
|
macOS |
August 10, 2003 |
Guojia biaozhun kuozhan (GBK) encoding method for Chinese text |
gb18030(5) |
|
macOS |
August 10, 2003 |
GB 18030 encoding method for Chinese text |
mskanji(5) |
|
macOS |
August 7, 2003 |
Shift-JIS (MS Kanji) encoding for Japanese text |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
charsets(7) |
|
Linux |
2024-01-28 |
character set standards and internationalization |
iso_8859-1(7) |
|
Linux |
2024-01-28 |
ISO/IEC 8859-1 character set encoded in octal, decimal, and hexadecimal |
iso_8859-2(7) |
|
Linux |
2024-01-28 |
ISO/IEC 8859-2 character set encoded in octal, decimal, and hexadecimal |
iso_8859-15(7) |
|
Linux |
2024-01-28 |
ISO/IEC 8859-15 character set encoded in octal, deci‐ mal, and hexadecimal |
iso_8859-16(7) |
|
Linux |
2024-01-28 |
ISO/IEC 8859-16 character set encoded in octal, deci‐ mal, and hexadecimal |
cp1251(7) |
|
Linux |
2024-01-28 |
CP 1251 character set encoded in octal, decimal, and hexa‐ decimal |
cp1252(7) |
|
Linux |
2024-01-28 |
CP 1252 character set encoded in octal, decimal, and hexa‐ decimal |
koi8-r(7) |
|
Linux |
2022-12-15 |
Russian character set encoded in octal, decimal, and hexa‐ decimal |
koi8-u(7) |
|
Linux |
2022-12-15 |
Ukrainian character set encoded in octal, decimal, and hexa‐ decimal |
armscii-8(7) |
|
Linux |
2022-12-15 |
Armenian character set encoded in octal, decimal, and hexadecimal |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
ascii(7) |
|
macOS |
June 5, 1993 |
octal, hexadecimal and decimal ASCII character sets |
ascii(7) |
|
Linux |
2024-01-28 |
ASCII character set encoded in octal, decimal, and hexadeci‐ mal |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
multibyte(3) |
|
macOS |
April 8, 2004 |
multibyte and wide character manipulation functions |
mbrtowc(3) |
mbrtowc_l |
macOS |
April 8, 2004 |
convert a character to a wide-character code (restartable) |
wcrtomb(3) |
wcrtomb_l |
macOS |
April 8, 2004 |
convert a wide-character code to a character (restartable) |
mbsrtowcs(3) |
mbsnrtowcs, mbsnrtowcs_l, mbsrtowcs_l |
macOS |
July 21, 2004 |
convert a character string to a wide-character string (restartable) |
wcsrtombs(3) |
wcsnrtombs, wcsnrtombs_l, wcsrtombs_l |
macOS |
July 21, 2004 |
convert a wide- character string to a character string (restartable) |
mbstowcs(3) |
mbstowcs_l |
macOS |
April 8, 2004 |
convert a character string to a wide-character string |
wcstombs(3) |
wcstombs_l |
macOS |
April 8, 2004 |
convert a wide-character string to a character string |
mblen(3) |
mblen_l |
macOS |
April 11, 2004 |
get number of bytes in a character |
mbrlen(3) |
mbrlen_l |
macOS |
April 7, 2004 |
get number of bytes in a character (restartable) |
mbtowc(3) |
mbtowc_l |
macOS |
April 11, 2004 |
convert a character to a wide-character code |
wctomb(3) |
wctomb_l |
macOS |
April 8, 2004 |
convert a wide-character code to a character |
btowc(3) |
btowc_l, wctob, wctob_l |
macOS |
August 3, 2002 |
convert between wide and single-byte characters |
wctob(3) |
btowc, btowc_l, wctob_l |
macOS |
August 3, 2002 |
convert between wide and single-byte characters |
mbsinit(3) |
mbsinit_l |
macOS |
April 8, 2004 |
determine conversion object status |
mbrtowc(3) |
|
Linux |
2023-10-31 |
convert a multibyte sequence to a wide character |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
mbrune(3) |
mbrrune, mbmb |
macOS |
April 19, 1994 |
multibyte rune support for C |
rune(3) |
setrunelocale, setinvalidrune, sgetrune, sputrune, fgetrune, fungetrune, fputrune |
macOS |
October 6, 2002 |
rune support for C |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
ctype(3) |
|
macOS |
March 30, 2004 |
digittoint, isalnum, isalpha, isascii, isblank, iscntrl, isdigit, isgraph, ishexnumber, isideogram, islower, isnumber, isphonogram, isprint, ispunct, isrune, isspace, isspecial, isupper, isxdigit, toascii, |
isalpha(3) |
|
macOS |
July 17, 2005 |
alphabetic character test |
isprint(3) |
|
macOS |
July 17, 2005 |
printing character test (space character inclusive) |
isascii(3) |
|
macOS |
October 6, 2002 |
test for ASCII character |
toascii(3) |
|
macOS |
June 4, 1993 |
convert a byte to 7-bit ASCII |
tolower(3) |
tolower_l |
macOS |
July 17, 2005 |
upper case to lower case letter conversion |
toupper(3) |
toupper_l |
macOS |
July 17, 2005 |
lower case to upper case letter conversion |
wctype(3) |
iswctype, iswctype_l, wctype_l |
macOS |
March 27, 2004 |
wide character class functions |
iswalpha(3) |
iswalnum, iswascii, iswblank, iswcntrl, iswdigit, iswgraph, iswhexnumber, iswideogram, iswlower, iswnumber, iswphonogram, iswprint, iswpunct, iswrune, iswspace, iswspecial, iswupper, iswxdigit |
macOS |
October 3, 2002 |
wide |
iswctype(3) |
iswctype_l, wctype, wctype_l |
macOS |
March 27, 2004 |
wide character class functions |
towlower(3) |
towlower_l |
macOS |
October 3, 2002 |
upper case to lower case letter conversion (wide character version) |
towupper(3) |
towupper_l |
macOS |
October 3, 2002 |
lower case to upper case letter conversion (wide character version) |
wctrans(3) |
towctrans, towctrans_l, wctrans_l |
macOS |
October 3, 2002 |
wide character mapping functions |
towctrans(3) |
towctrans_l, wctrans, wctrans_l |
macOS |
October 3, 2002 |
wide character mapping functions |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
wcwidth(3) |
wcwidth_l |
macOS |
August 17, 2004 |
number of column positions of a wide-character code |
wcswidth(3) |
wcswidth_l |
macOS |
August 20, 2002 |
number of column positions in wide-character string |
wcwidth(3) |
|
Linux |
2023-10-31 |
determine columns needed for a wide character |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
strcoll(3) |
strcoll_l |
macOS |
June 4, 1993 |
compare strings, according to current collation |
strxfrm(3) |
strxfrm_l |
macOS |
June 4, 1993 |
transform a string under locale |
wcscoll(3) |
wcscoll_l |
macOS |
October 4, 2002 |
compare wide strings according to current collation |
wcsxfrm(3) |
wcsxfrm_l |
macOS |
October 4, 2002 |
transform a wide string under locale |
strcasecmp(3) |
strcasecmp_l, strncasecmp, strncasecmp_l |
macOS |
June 9, 1993 |
compare strings, ignoring case |
wcscasecmp(3) |
|
macOS |
March 4, 2009 |
wcpcpy, wcpncpy, wcscasecmp, wcscat, wcschr, wcscmp, wcscpy, wcscspn, wcsdup, wcslcat, wcslcpy, wcslen, wcsncasecmp, wcsncat, wcsncmp, wcsncpy, wcsnlen, wcspbrk, wcsrchr, wcsspn, wcsstr, wmemchr, wmemcmp, wmemcpy, |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
locale(1) |
|
macOS |
August 27, 2004 |
display locale settings |
setlocale(3) |
|
macOS |
November 21, 2003 |
natural language formatting for C |
localeconv(3) |
localeconv_l |
macOS |
November 21, 2003 |
natural language formatting for C |
nl_langinfo(3) |
nl_langinfo_l |
macOS |
May 3, 2001 |
language information |
xlocale(3) |
|
macOS |
March 11, 2005 |
Extended locale support. |
newlocale(3) |
|
macOS |
March 11, 2005 |
Create a new locale |
uselocale(3) |
|
macOS |
March 11, 2005 |
Set the per-thread locale |
duplocale(3) |
|
macOS |
March 11, 2005 |
Duplicate existing locale |
querylocale(3) |
|
macOS |
March 11, 2005 |
Get locale name for a specified category |
environ(7) |
|
macOS |
April 12, 2003 |
user environment |
environ(7) |
|
Linux |
2023-10-31 |
user environment |
setlocale(3) |
|
Linux |
2024-02-25 |
set the current locale |
nl_langinfo(3) |
nl_langinfo_l |
Linux |
2024-01-28 |
query language and locale information |
locale(1) |
|
Linux |
2023-10-31 |
get locale-specific information |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
locale(7) |
|
Linux |
2024-02-25 |
description of multilanguage support |
locale(5) |
|
Linux |
2024-01-28 |
describes a locale definition file |
charmap(5) |
|
Linux |
2023-10-31 |
character set description file |
repertoiremap(5) |
|
Linux |
2023-10-31 |
map symbolic character names to Unicode code points |
localedef(1) |
|
Linux |
2023-10-31 |
compile locale definition files |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
iconv(1) |
|
macOS |
October 22, 2009 |
codeset conversion utility |
iconv(3) |
iconv_open, iconv_open_into, iconv_close |
macOS |
August 4, 2014 |
codeset conversion functions |
iconvctl(3) |
|
macOS |
November 25, 2009 |
controlling and diagnostical facility for iconv(3) |
iconvlist(3) |
|
macOS |
February 23, 2023 |
retrieving a list of character encodings supported by iconv(3) |
iconv(1) |
|
Linux |
2024-01-28 |
convert text from one character encoding to another |
iconv(3) |
|
Linux |
2023-10-31 |
perform character set conversion |
iconv_open(3) |
|
Linux |
2023-10-31 |
allocate descriptor for character set conversion |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
vis(1) |
|
macOS |
February 18, 2021 |
display non-printable characters in a visual format |
unvis(1) |
|
macOS |
November 27, 2010 |
revert a visual representation of data back to original form |
vis(3) |
nvis, strvis, stravis, strnvis, strvisx, strnvisx, strenvisx, svis, snvis, strsvis, strsnvis, strsvisx, strsnvisx, strsenvisx |
macOS |
April 22, 2017 |
visually encode characters |
unvis(3) |
strunvis, strnunvis, strunvisx, strnunvisx |
macOS |
March 12, 2011 |
decode a visual representation of characters |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
base64(1) |
bintrans, uuencode, uudecode, b64encode, b64decode |
macOS |
January 23, 2024 |
encode / decode a binary file |
uuencode(1) |
bintrans, uudecode, b64encode, b64decode, base64 |
macOS |
January 23, 2024 |
encode / decode a binary file |
uudecode(1) |
bintrans, uuencode, b64encode, b64decode, base64 |
macOS |
January 23, 2024 |
encode / decode a binary file |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
byteorder(3) |
htonl, htons, htonll, ntohl, ntohs, ntohll |
macOS |
June 4, 1993 |
convert values between host and network byte order |
swab(3) |
|
macOS |
February 24, 2010 |
swap adjacent bytes |
bitstring(3) |
bit_alloc, bit_clear, bit_decl, bit_ffs, bit_nclear, bit_nset, bit_set, bitstr_size, bit_test |
macOS |
July 19, 1993 |
bit-string manipulation macros |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
stdio(3) |
|
macOS |
March 3, 2009 |
standard input/output library functions |
fopen(3) |
fdopen, freopen, fmemopen |
macOS |
September 1, 2023 |
stream open functions |
fgets(3) |
gets |
macOS |
June 4, 1993 |
get a line from a stream |
fgetln(3) |
|
macOS |
April 19, 1994 |
get a line from a stream |
getline(3) |
getdelim |
macOS |
November 30, 2010 |
get a line from a stream |
fgetws(3) |
fgetws_l |
macOS |
August 6, 2002 |
get a line of wide characters from a stream |
getwc(3) |
fgetwc, getwchar |
macOS |
March 3, 2004 |
get next wide character from input stream |
putwc(3) |
fputwc, putwchar |
macOS |
March 3, 2004 |
output a wide character to a stream |
fwide(3) |
|
macOS |
October 24, 2001 |
get/set orientation of a stream |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
printf(3) |
fprintf, sprintf, snprintf, asprintf, dprintf, vprintf, vfprintf, vsprintf, vsnprintf, vasprintf, vdprintf |
macOS |
December 2, 2009 |
formatted output conversion |
wprintf(3) |
fwprintf, swprintf, vwprintf, vfwprintf, vswprintf |
macOS |
July 5, 2003 |
formatted wide character output conversion |
printf(1) |
|
macOS |
July 1, 2020 |
formatted output |
echo(1) |
|
macOS |
April 12, 2003 |
write arguments to the standard output |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
strtol(3) |
strtoimax, strtoll, strtoq |
macOS |
November 28, 2001 |
convert a string value to a long, long long, intmax_t or quad_t integer |
strtoul(3) |
strtoull, strtoumax, strtouq |
macOS |
November 28, 2001 |
convert a string to an unsigned long, unsigned long long, uintmax_t, or u_quad_t integer |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
re_format(7) |
|
macOS |
Sept 29, 2011 |
POSIX 1003.2 regular expressions |
regex(3) |
regcomp, regcomp_l, regerror, regexec, regfree, regncomp, regncomp_l, regnexec, regnwcomp, regnwcomp_l, regnwexec, regwcomp, regwcomp_l, regwexec |
macOS |
July 11, 2025 |
regular-expression library |
fnmatch(3) |
|
macOS |
April 7, 2025 |
test whether a filename or pathname matches a shell-style pattern |
glob(3) |
glob_b, globfree |
macOS |
June 23, 2025 |
generate pathnames matching a pattern |
grep(1) |
egrep, fgrep, rgrep, bzgrep, bzegrep, bzfgrep, zgrep, zegrep, zfgrep |
macOS |
November 10, 2021 |
file pattern searcher |
sed(1) |
|
macOS |
December 17, 2024 |
stream editor |
regex(7) |
|
Linux |
2023-11-01 |
POSIX.2 regular expressions |
glob(7) |
|
Linux |
2023-10-31 |
globbing pathnames |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
hexdump(1) |
|
macOS |
June 29, 2020 |
ASCII, decimal, hexadecimal, octal dump |
od(1) |
|
macOS |
December 22, 2011 |
octal, decimal, hex, ASCII dump |
xxd(1) |
|
macOS |
May 2024 |
make a hex dump or do the reverse. |
strings(1) |
|
macOS |
June 7, 2016 |
find the printable strings in a object, or other binary, file |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
file(1) |
|
macOS |
February 5, 2021 |
determine file type |
magic(5) |
|
macOS |
May 9, 2021 |
file command's magic pattern file |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
tr(1) |
|
macOS |
October 13, 2006 |
translate characters |
cut(1) |
|
macOS |
August 3, 2017 |
cut out selected portions of each line of a file |
fold(1) |
|
macOS |
October 29, 2020 |
fold long lines for finite width output device |
fmt(1) |
|
macOS |
October 29, 2020 |
simple text formatter |
expand(1) |
unexpand |
macOS |
June 6, 2015 |
expand tabs to spaces, and vice versa |
unexpand(1) |
expand |
macOS |
June 6, 2015 |
expand tabs to spaces, and vice versa |
col(1) |
|
macOS |
October 21, 2020 |
filter reverse line feeds from input |
colrm(1) |
|
macOS |
June 23, 2020 |
remove columns from a file |
pr(1) |
|
macOS |
July 3, 2004 |
print files |
ul(1) |
|
macOS |
October 7, 2020 |
do underlining |
rev(1) |
|
macOS |
June 27, 2020 |
reverse lines of a file |
wc(1) |
|
macOS |
April 11, 2020 |
word, line, character, and byte count |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
cat(1) |
|
macOS |
January 29, 2013 |
concatenate and print files |
head(1) |
|
macOS |
April 10, 2018 |
display first lines of a file |
tail(1) |
|
macOS |
November 28, 2023 |
display the last part of a file |
nl(1) |
|
macOS |
June 18, 2020 |
line numbering filter |
split(1) |
|
macOS |
May 26, 2023 |
split a file into pieces |
csplit(1) |
|
macOS |
February 6, 2014 |
split files based on context |
tee(1) |
|
macOS |
June 23, 2020 |
duplicate standard input |
look(1) |
|
macOS |
December 29, 2020 |
display lines beginning with a given string |
uniq(1) |
|
macOS |
December 9, 2024 |
report or filter out repeated lines in a file |
sort(1) |
|
macOS |
September 4, 2019 |
sort or merge records (lines) of text and binary files |
comm(1) |
|
macOS |
July 27, 2020 |
select or reject lines common to two files |
join(1) |
|
macOS |
June 20, 2020 |
relational database operator |
paste(1) |
|
macOS |
June 25, 2004 |
merge corresponding or subsequent lines of files |
lam(1) |
|
macOS |
April 7, 2015 |
laminate files |
rs(1) |
|
macOS |
April 7, 2015 |
reshape a data array |
jot(1) |
|
macOS |
September 21, 2019 |
print sequential or random data |
seq(1) |
|
macOS |
June 20, 2020 |
print sequences of numbers |
xargs(1) |
|
macOS |
September 21, 2020 |
construct argument list(s) and execute utility |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
diff(1) |
|
macOS |
January 7, 2025 |
differential file and directory comparator |
cmp(1) |
|
macOS |
September 23, 2021 |
compare two files |
cksum(1) |
sum |
macOS |
April 28, 1995 |
display file checksums and block counts |
md5(1) |
sha1, sha224, sha256, sha384, sha512, md5sum, sha1sum, sha224sum, sha256sum, sha384sum, sha512sum |
macOS |
February 13, 2024 |
calculate a message-digest fingerprint (checksum) for a file |
Moving blocks, with six code pages inside: dd(1)
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
dd(1) |
|
macOS |
May 19, 2021 |
convert and copy a file |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
tty(4) |
|
macOS |
August 14, 1992 |
general terminal interface |
stty(1) |
|
macOS |
October 20, 2018 |
set the options for a terminal device interface |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
man(1) |
apropos, whatis |
macOS |
January 9, 2021 |
display online manual documentation pages |
mandoc(1) |
|
macOS |
August 14, 2021 |
format manual pages |
mandoc_char(7) |
|
macOS |
October 31, 2020 |
mandoc special characters |
| Page |
Also on this page |
Machine |
Date on the page |
What its NAME line says |
tar(5) |
|
macOS |
December 27, 2016 |
format of tape archive files |
cpio(5) |
|
macOS |
December 23, 2011 |
format of cpio archive files |
What the measurements found
The family pages hold about a hundred dated fences. These are the findings worth knowing before you open any of them, one line each, with the page that measured it.
- The 2004
utf8(5) table is still what this Mac's iconv implements: it decodes the five- and six-byte spellings the page prints, to numbers no character has; GNU's refuses them; both accept the noncharacter U+FFFE and both refuse an overlong NUL and a surrogate. Python refuses the long forms on both. utf8(5)
utf-8(7), dated 2024, still prints six rows and opens with a sentence about a 16-bit code space that has been false since 2001. utf8(5)- Neither UTF-8 page mentions the byte order mark, the replacement character, or what to do with bad input.
utf8(5)
MB_CUR_MAX under a UTF-8 locale is 4 on macOS and 6 on glibc, and MB_LEN_MAX is 6 against 16, so a buffer sized by one platform's header is not sized for the other's. multibyte(3)- Under the
C locale, macOS decodes every byte and glibc refuses every byte above 0x7f; both are conforming, and no page says which its library does. multibyte(3)
- On macOS a
wchar_t is not a code point outside UTF-8 locales: under ja_JP.eucJP, mbrtowc on c3 a9 returns 0xC3A9, and __STDC_ISO_10646__ is not defined. glibc defines it and promises the opposite. multibyte(3)
- This Mac's
iconv transliterates without being asked: Ł into Latin-1, CP1252 or GBK is L, ź is ´z, − is -, all with exit 0 and nothing on stderr, which only the BUGS paragraph of iconvctl(3) admits. The ? its iconv(3) page promises appears for some targets and never for ASCII, where it stops as GNU does. iconv(1)
- The GNU
iconv(1) and iconv_open(3) pages document //TRANSLIT and //IGNORE, the flags What the page does not say found undocumented on the Mac. The GNU iconv(3) page also states the BSD ? fallback that this library measured by hand. iconv(1)
-
iconv -c drops and exits 1 on the Mac, drops and exits 0 on Ubuntu; //TRANSLIT writes 'e for é on the Mac and e on Ubuntu, and the Mac counts its own transliterations as invalid characters. iconv(1)
-
Apple's GB18030 stops at the BMP: its iconv decodes 84 31 A4 37 but refuses 90 30 81 30 (U+10000) and everything above, while GNU decodes the whole four-byte range, and 😀 is 94 39 FC 36. Python's codec on the same Mac agrees with GNU. The CJK pages
euc(5)'s recipe for a wchar_t is stale: under ja_JP.eucJP the single-shift byte the page says is first removed is packed into the value (0x8eb1 for SS2), and the page says codesets 2 and 3 where its own definitions mean 3 and 4. The CJK pages- Shift-JIS by two names:
-f SHIFT_JIS and -f CP932 decode 5c as ¥ and \, 7e as ‾ and ~, identically on both machines; MSKANJI, the page's own ENCODING string, is refused by both iconvs and MS_KANJI accepted. The CJK pages
- The code-page tables check out row for row: Latin-9 differs from Latin-1 in exactly eight rows,
cp1252(7) has 27 rows in 0x80–0x9F and is missing exactly 81 8D 8F 90 9D, which both iconvs and Python refuse, and clearing bit 7 of KOI8-R Привет, мир gives pRIWET, MIR. charsets(7)
- Neither
tar ever wrote a hdrcharset record, and bsdtar lists café.txt from a pax archive in NFD and from a GNU-format archive in NFC, with the same bytes in both headers. A 1960 timestamp gets six different treatments from the two tars' four formats. tar(5) and cpio(5)
bitstring(3)'s bit 0 is the least significant bit of byte 0, bit_test returns 32 rather than 1, and Ubuntu has no bitstring.h at all; htonl(0x0A0B0C0D) is 0x0D0C0B0A on both of these x86_64 machines. byteorder(3)- glibc's
fnmatch gives two answers in a UTF-8 locale: café matches caf? and caf??, a character-wise and a byte-wise reading accepted side by side; macOS gives one answer per locale. Patterns
- A range is the locale's:
[a-z] matches é only under glibc's en_US.UTF-8, [[=e=]] matches it under en_US.UTF-8 on both machines but not under C.utf8, and [[:lower:]] is the only portable spelling. REG_ILLSEQ exists on macOS only. Patterns
- GNU
cmp says byte 4 under C.UTF-8 and char 4 under C; BSD says char always. So the library's finding 18, that both print char, holds only in the C locale. Comparing and summing
bintrans(1) misdescribes its own base64: BSD base64 accepts -w as well as -b (finding 14 lists only -b), and rejects --ignore-garbage with exit 64 where the page says it is silently ignored. Binary to textfile's encoding guess is code, not magic: -e ascii and -e encoding each switch off half of it, -e soft (no magic at all) leaves it intact, and the page's own TODO says encoding.c is hard-coded. file(1) and magic(5)- BSD
hexdump's %_u and %_c consult the locale and util-linux's do not; od's page says ** and hexdump's says *, and the newline is nl on one page and lf on the other. Seeing the bytes
- The 4.4BSD rune API is still there and half-broken:
sgetrune decodes c3 a9 to U+00E9 but libc prints a three-line depreciated notice to stderr, and setrunelocale compiles with a deprecated in macOS 10.4 warning and then segfaults. rune_t is wchar_t by typedef. mbrune(3) and rune(3)
isalpha(0xE9) is 1 in every non-C locale on the Mac and 0 in every locale on glibc, including C.UTF-8, and toupper(0xE9) is 0xC9 on the Mac and the identity on glibc; the obsolete 4.4BSD extension that lets isalpha take a code point still answers 1 for U+4E2D in 2026. ctype(3)- The two
wcwidth tables are a version apart: the Mac's is Unicode 15.0, glibc 2.39's is 15.1, both return -1 for every Unicode 16.0 character, and SOFT HYPHEN is -1 on one and 1 on the other. Nothing on either machine pads printf '%-10s' by columns: the shells and both /usr/bin/printf pad by bytes, Python by code points. wcwidth(3)
wcscoll(3)'s BUGS paragraph is stale: it says the function falls back to wcscmp, and wcscoll(L"é", L"f") is -1 under en_US.UTF-8 on this Mac. strcasecmp folds 0xC9 and 0xE9 together in every non-C locale on the Mac, including UTF-8 ones, and never on glibc. strcoll(3)strtol under an unsupported base: glibc returns 0 with EINVAL and never writes *endptr, where the Mac sets it to the input; "0b101" is 0 on both under C11 and 5 on glibc under C23, because the header redirects to __isoc23_strtol. Finding 16 reproduced on both. strtol(3)- util-linux
rev never returns under LC_ALL=C on café (killed by timeout, exit 124) and is fine under C.UTF-8; BSD rev in C reverses the bytes. BSD fmt and BSD col in the C locale delete every byte above 0x7f, so col -b must run under the locale man rendered with. Text tools that count
- BSD
dd conv=ucase in a UTF-8 locale corrupts UTF-8 lead bytes (CAFÉ comes out 63 61 66 e3 89) because it is per-byte toupper; GNU folds ASCII only. conv=ibm is CP1047 on every ASCII byte tested and conv=ebcdic differs from it only at ^ and ~; none of the six tables is CP037, and above 0x7f they are a permutation rather than a code page. dd(1)
- Bold reaches a pipe as
N backspace N on the Mac and not at all on Ubuntu, where man-db strips formatting off a pipe; LC_CTYPE changes man 7 mandoc_char from 0 to 1,040 non-ASCII bytes, and mandoc -T ascii writes \('e as 'e where the page's table says e. The manual itself
- Canonical-mode erase is per byte until
iutf8 is set: Backspace after café hands the program 63 61 66 c3 on both platforms, and both fresh ptys start -iutf8; tty(4), dated 1992, never mentions it. The terminal
- Smaller splits from the text tools: GNU
fold counts bytes even in C.UTF-8 while BSD counts columns; wc -L is bytes on BSD and columns on GNU; uniq -i and paste -d '€' honour the locale on BSD only; look é misses on both machines in every locale on an alphabetically sorted file, exactly as its BUGS section says. Text tools that cut on lines
- The two
C locales are different locales, three pages along: under C, fgetws reads é as U+00C3 U+00A9 on the Mac and fails on the first line on glibc; printf("%ls", L"café") writes 63 61 66 e9 on the Mac and returns -1 with EILSEQ on glibc; putwc(U+00E9) writes e9 on the Mac and a ? with a success return on glibc. Streams, Formatting
- Stream orientation is enforced only by glibc: after
getwc, fgetc on the same stream returns the next byte on the Mac and EOF on glibc. After an EILSEQ, fgetws is stuck for good on both, and the Mac leaves ferror at 0. Streams
/usr/bin/vis on this Mac imports no locale function at all, so the MULTIBYTE CHARACTER SUPPORT section of its page describes nothing it does; and strnvisx in a UTF-8 locale writes é as the single byte e9 when the buffer holds a NUL, breaking the page's own invariant. Escaping into ASCII- glibc's
//TRANSLIT is two tables: ż é € ß become ? ? EUR ss under LC_ALL=C and z e EUR ss under C.UTF-8; the € and ß rules are built in, the é and ż rules are lines of translit_combining in /usr/share/i18n/locales. pl_PL's thousands separator is U+00A0 on the Mac and U+202F on glibc. The locale's source files, The locale
printf '%d' "'Ł" is 321 under UTF-8 and 197 under C from the BSD binary, GNU, zsh and bash 5.2, and -59, a signed byte, from bash 3.2 in either locale. Formatting
The remaining family pages add their own; each one's Try it section is the place to look.
See also