share/mk/: Consistency fixes in copyright lines

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: build-html: Skip known errors (in groff-1.23.0)

Signed-off-by: Alejandro Colomar <alx@kernel.org>

LICENSES/: Add LGPL-3.0-or-later and LGPL-3.0-linking-exception

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Update license

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Use grohtml(1) instead of man2html(1)

man2html(1) crashes on tzfile(5), and the upstream project is defunct.

Link: <https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1067022>
Cc: Paul Eggert <eggert@cs.ucla.edu>
Cc: "G. Branden Robinson" <branden@debian.org>
Cc: Robert Luberda <robert@debian.org>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

proc_pid_io.5: Move note to CAVEATS

Cc: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

bpf.2: wfix

Signed-off-by: Jing Peng <pj.hades@gmail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

ip.7: Some socket options are not supported by SOCK_STREAM

It was not made clear in several socket options that they were not
supported by SOCK_STREAM; this patch fixes that.

Socket options not supported by SOCK_STREAM are handled in the
ip_cmsg_recv_offset() function in <net/ipv4/ip_sockglue.c>.  The
function is called for udp sockets, and indirectly by ping and raw
sockets, but not for STREAM sockets, as they don't support these
options.

Link: <https://lore.kernel.org/linux-man/ejhphmjh74ebtk4br3id66f27a4yoh4aukrcz7m6dp7acsu6zr@crtueyadqzmp/T/#mb298ac7f71a348d1e6b423cfa32bfad9c28efa40>
Link: <https://lore.kernel.org/linux-man/ejhphmjh74ebtk4br3id66f27a4yoh4aukrcz7m6dp7acsu6zr@crtueyadqzmp/T/#u>
Signed-off-by: Oliver Crumrine <ozlinuxc@gmail.com>
Cc: Peter Seiderer <ps.report@gmx.net>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

proc_pid_io.5: Dewafflify

This page copies verbatim the contents of
Documentation/filesystems/proc.rst, added wholesale in
commit f9c99463b0cd05603d125c915e2886d55a686b82 ("[PATCH] Documentation
for io-accounting / reporting via procfs") in 2007.

As such, it mirrors the sensibilities of the time ‒
writing "successful read returns" as "data pulled from storage. actually
just the data the process gave to read(). this also means from non-regular
files! whether the data was pulled from storage doesn't matter actually
(obligatory cache mention)"
for the modern reader this is just a lot of waffling
(note also that processes give no data to read()!)
‒ and sensibilities of the sheepish implementer in kernel documentation ‒
"an attempt" for a well-defined kernel behaviour, mentioning the
"current implementation", consistent mentions of specific kernel-internal
caching mechanisms, "the big inaccuracy here".

Re-write to be more useful and less misleading as documentation;
the syscall enumeration is accurate for kernel v6.8, but the sysc? stats
are also bumped by kernel_{read,write}(), which is sometimes used by too
many syscalls in too many scenarios to usefully enumerate.

Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

proc_pid_io.5: Stats include children

This file is like getrusage(2) RUSAGE_SELF + RUSAGE_CHILDREN,
the current wording implies it's like just RUSAGE_SELF.

Compare:
  #include <fcntl.h>
  #include <signal.h>
  #include <string.h>
  #include <sys/wait.h>
  #include <unistd.h>

  void copy() {
    char buf[1024];
    int fd = open("/proc/self/io", 0);
    write(1, buf, read(fd, buf, sizeof(buf)));
    close(fd);
  }

  int main() {
    copy();
    copy();
    if (!fork()) {
    zero:;
      int fd = open("/dev/zero", 0);
      char buf[64 * 1024] = {};
      write(1, buf, 10000);
      for (int i = 0; i < 1000; ++i)
        read(fd, buf, sizeof(buf));
      _exit(0);
    }
    sleep(1);
    copy();
    wait(NULL);
    copy();
    signal(SIGCHLD, SIG_IGN);
    if (!fork())
      goto zero;
    sleep(1);
    copy();
    wait(NULL);
    copy();
  }
yielding
  rchar: 3980
  wchar: 0
  syscr: 9
  syscw: 0
  read_bytes: 2968
  write_bytes: 0
  cancelled_write_bytes: 0

  rchar: 4076
  wchar: 96
  syscr: 10
  syscw: 1
  read_bytes: 2968
  write_bytes: 96
  cancelled_write_bytes: 0

  rchar: 4175
  wchar: 195
  syscr: 11
  syscw: 2
  read_bytes: 2968
  write_bytes: 195
  cancelled_write_bytes: 0

  rchar: 65540276
  wchar: 10296
  syscr: 1012
  syscw: 4
  read_bytes: 2968
  write_bytes: 10296
  cancelled_write_bytes: 0

  rchar: 65540387
  wchar: 10407
  syscr: 1013
  syscw: 5
  read_bytes: 2968
  write_bytes: 10407
  cancelled_write_bytes: 0

  rchar: 65540498
  wchar: 10518
  syscr: 1014
  syscw: 6
  read_bytes: 2968
  write_bytes: 10518
  cancelled_write_bytes: 0

Just s/process/& and its waited-for children/ but re-broken per review.

Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

tzfile.5: ffix

Cc: "G. Branden Robinson" <branden@debian.org>
Cc: Paul Eggert <eggert@cs.ucla.edu>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

tzfile.5: Avoid crashing man2html(1)

Link: <https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=1067022>
Cc: Paul Eggert <eggert@cs.ucla.edu>
Cc: "G. Branden Robinson" <branden@debian.org>
Cc: Robert Luberda <robert@debian.org>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

tzfile.5, tzselect.8, zdump.8, zic.8: Refresh pages

Refreshed from tzdb-2024a.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

bpf-helpers.7: Refresh page

Scripted change:

$ ~/src/linux/linux/6.8/scripts/bpf_doc.py | rst2man > man7/bpf-helpers.7

Signed-off-by: Alejandro Colomar <alx@kernel.org>

getgrouplist.3: Use same type in both sides of comparison

This removes a warning.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

unix.7: EXAMPLES: server.c: End connection with END

Client wants to send END while server already closed the connection on
DOWN, so connection is broken instead of the Result = 0 in the sample on
sending a single DOWN.
Now, the server disconnects only on first END. After DOWN, all further
processing of number stops.
Patch does not handle cases of double END sending, multiple clients etc.

Fixes: 15545eb6d7ae ("unix.7: Add example")
Cc: Heinrich Schuchardt <xypron.glpk@gmx.de>
Signed-off-by: Lili Püspök <poordirtylili@gmail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

printf.3: Return the number of bytes printed, not characters

From <https://pubs.opengroup.org/onlinepubs/9699919799/functions/printf.html>:

    Upon successful completion, the dprintf(), fprintf(), and printf()
    functions shall return the number of bytes transmitted.

Closes: <https://bugzilla.kernel.org/show_bug.cgi?id=218600>
Reported-by: <quirin.blaeser@freenet.de>
Signed-off-by: Jeremy Baxter <jtbx@disroot.org>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

utf-8.7: ffix

Signed-off-by: Alejandro Colomar <alx@kernel.org>

utf-8.7: ffix

Signed-off-by: Alejandro Colomar <alx@kernel.org>

printf.h.3head: tfix

Signed-off-by: Alejandro Colomar <alx@kernel.org>

getrusage.2, proc_pid_io.5: Crosslink

These serve the same purpose from different perspectives.

Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

getrusage.2: proc(5) /proc/pid/stat -> proc_pid_stat(5)

Fixes: bbc21bc4dbef ("proc.5, proc_pid_stat.5: Split /proc/PID/stat from proc(5)")
Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

proc_pid_io.5: Mention atomicity in atomicity note

And drop "current implementation"
(wording still literal from Documentation/filesystems/proc.rst).
Of course this describes the current implementation.

Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

proc_pid_io.5: wfix

Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

scripts/LinuxManBook/build.sh, share/mk/: Use command variables

Signed-off-by: Alejandro Colomar <alx@kernel.org>

logb.3: floor(log2(fabs(x))) has rounding errors

Link: <https://inbox.sourceware.org/libc-alpha/20240305150131.GD3653@qaa.vinc17.org/T/#m3ceecda630012995339bcc5448fee451cf277a8b>
Reported-by: Vincent Lefevre <vincent@vinc17.net>
Suggested-by: Vincent Lefevre <vincent@vinc17.net>
Cc: Morten Welinder <mwelinder@gmail.com>
Cc: Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

scripts/LinuxManBook/prepare.pl: Support some autogenerated pages from other projects

And remove a warning about an experimental perl feature, by using a
while instead of a for loop.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

TIMESPEC_TO_TIMEVAL.3, TIMEVAL_TO_TIMESPEC.3: Document these macros

Reported-by: Guillem Jover <guillem@hadrons.org>
Reviewed-by: Oskari Pirhonen <xxc3ncoredxx@gmail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/configure/build-depends/cc.mk: Use -Wno-unsafe-buffer-usage with clang(1)

That warning has false positives, such as in unix(7).

Signed-off-by: Alejandro Colomar <alx@kernel.org>

unix.7: Use a 'ssize_t' for the return value of read(2) and write(2)

Signed-off-by: Alejandro Colomar <alx@kernel.org>

unix.7: Use 'int' to iterate over 'argv'

'argc' is an 'int', so we should compare it with a variable of the same
type.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

unix.7: Use int main(void)

The arguments were unused.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

unix.7: Remove spurious #include

Signed-off-by: Alejandro Colomar <alx@kernel.org>

unix.7: Build example programs

Signed-off-by: Alejandro Colomar <alx@kernel.org>

read.2: tfix

Signed-off-by: Hugo Gabriel Eyherabide <hugogabiel.eyherabide@gmail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

log2.3, log10.3: wfix

Reported-by: Morten Welinder <mwelinder@gmail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

cbrt.3: wfix: Cube roots are rarely representable

On Thu, Feb 29, 2024 at 07:28:10PM -0500, Morten Welinder wrote:
> The phrase "every representable real value has a representable real
> cube root" is wrong.  In fact, a representable cube root is quite
> rare.

Link: <https://lore.kernel.org/linux-man/ZeYKUOKYS7G90SaV@debian/T/#mff0ab388000c6afdb5e5162804d4a0073de481de>
Reported-by: Morten Welinder <mwelinder@gmail.com>
Cowritten-by: Morten Welinder <mwelinder@gmail.com>
Cc: Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>
Cc: Vincent Lefevre <vincent@vinc17.net>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

significand.3: significand() uses FLT_RADIX, not 2

It's implemented using scalb(), which uses FLT_RADIX.

Reported-by: Morten Welinder <mwelinder@gmail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

logb.3: logb(x) is floor(log2(fabs(x)))

log2(3) doesn't accept negative input, but it seems logb(3) does accept
it.

Link: <https://lore.kernel.org/linux-man/ZeYKUOKYS7G90SaV@debian/T/#u>
Reported-by: Morten Welinder <mwelinder@gmail.com>
Cc: Adhemerval Zanella Netto <adhemerval.zanella@linaro.org>
Cc: Vincent Lefevre <vincent@vinc17.net>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: dist: Don't package .git/

Signed-off-by: Alejandro Colomar <alx@kernel.org>

clock_getres.2: HISTORY: Clarify that POSIX.1-2008 only makes mandatory the functions

The different clocks are still optional.

Closes: <https://bugzilla.kernel.org/show_bug.cgi?id=218452>
Fixes: 4131356cdab8 ("man*/, man-pages.7: VERSIONS, STANDARDS, HISTORY: Reorganize sections")
Reported-by: Enrique Garcia <cquike@arcor.de>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

cgroups.7: tfix

Closes: <https://bugzilla.kernel.org/show_bug.cgi?id=218105>
Reported-by: Göran Uddeborg <goeran@uddeborg.se>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Add missing include

Signed-off-by: Alejandro Colomar <alx@kernel.org>

clock_nanosleep.2: Use 't' rather than 'request'

It seems much more clear.

Cc: Elliott Hughes <enh@google.com>
Cc: Stefan Puiu <stefan.puiu@gmail.com>
Cc: Bruno Haible <bruno@clisp.org>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

nanosleep.2: Use 'duration' rather than 'request'

It seems much more clear.

Suggested-by: Elliott Hughes <enh@google.com>
Cc: Stefan Puiu <stefan.puiu@gmail.com>
Cc: Bruno Haible <bruno@clisp.org>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

man*/: epoll_*(), fcntl(), flock(), ioctl(), msgctl(), *prctl(), ptrace(), quotactl(), reboot(), semctl(), shmctl(), lockf(): Consistently use 'op' and 'operation'

Reported-by: Bruno Haible <bruno@clisp.org>
Cc: Elliott Hughes <enh@google.com>
Cc: Stefan Puiu <stefan.puiu@gmail.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

Changes.old: tfix

Signed-off-by: Jakub Wilk <jwilk@jwilk.net>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck: Rewrite most of it

This implementation is simpler.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: build-ex-src: Support long $DISTNAME

When $DISTNAME is too long, mandoc(1) breaks the last line.  The last
two lines can always be removed safely, which makes it work also in this
case.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Fix path

Signed-off-by: Alejandro Colomar <alx@kernel.org>

process_madvise.2: Fix comments about commit ids

Move the comment about the Linux commit id, specify that it's a Linux
commit, and add the glibc commit id too.

Link: <https://sourceware.org/bugzilla/show_bug.cgi?id=27380>
Link: <https://lore.kernel.org/linux-man/Zd5fMGvIlmhQyONs@thunder.hadrons.org/T/#m9129640e1293a94ff1606a2f973522f40c968306>
Fixes: 28628d826661 ("process_madvise.2: Document the glibc wrapper")
Reported-by: Guillem Jover <guillem@hadrons.org>
Cc: Florian Weimer <fweimer@redhat.com>
Cc: Carlos O'Donell <carlos@redhat.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

CONTRIBUTING.d/lint: Rewrite now that we have $SKIP_XFAIL

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/, man*/: build-ex: Skip expected failures

Signed-off-by: Alejandro Colomar <alx@kernel.org>

etc/cppcheck/cppcheck.suppress: Silence report

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/, etc/: Fix dependencies

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: lint-c: Skip expected failures

Signed-off-by: Alejandro Colomar <alx@kernel.org>

man3/: Say cancelation instead of cancellation

It's the most common spelling in this project.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: $SKIP_XFAIL: Add variable to allow skipping expected failures

Default to 'yes'.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck: string_copying.7 fails 'check-catman'

Except with groff(1) from git HEAD, which will be 1.24.0 eventually.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

mbsrtowcs.3: ffix

With groff-1.23.0, which has a base paragraph indentation of 7, this
line was longer than 80 columns.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/configure/version.mk: $DISTDATECMD: Use committer date, not author date

Comitter date is always increasing, while author date may jump
backwards, which is problematic with make(1).

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Simplify code

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck: Ignore fanotify_init.2 build errors

In groff from git HEAD, it doesn't fail, but in 1.23.0, which still
has the default base paragraph indentation set to 7, it reports

troff:man2/fanotify_init.2:322: warning [p 4, 0.7i]: cannot adjust line

Signed-off-by: Alejandro Colomar <alx@kernel.org>

man*/: srcfix

Found with:

$ grep -rn '"""' man* \
| grep -v '"""""""""""' \
| sed 's/:.*//' \
| sort \
| uniq;
man2/add_key.2
man2/getrlimit.2
man2/keyctl.2
man2/pivot_root.2
man2/request_key.2
man3/isalpha.3
man3/setlocale.3
man3/toupper.3
man7/capabilities.7
man7/cgroups.7
man7/keyrings.7
man7/locale.7
man7/user_namespaces.7

Signed-off-by: Alejandro Colomar <alx@kernel.org>

man*/: ffix

Found with:

$ grep -rn '^\.[BI] .* [.,:;)]*$' man*
man2/prctl.2:382:.B FR=1 .
man2/openat2.2:377:.B EAGAIN .
man2/openat2.2:424:.I how.resolve .
man5/elf.5:788:.B PF_R .
man5/networks.5:18:.I name number aliases ...
man5/protocols.5:31:.I protocol number aliases ...
man7/cgroups.7:980:.I """max""" .

Cc: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck-*: Add targets to run 'distcheck' partially

This is mainly for debugging purposes.  I won't document it in
'make help' for now, as it will clutter the output, and isn't so useful
for normal users.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

vsock.7: ffix

Signed-off-by: Ahelenia Ziemiańska <nabijaczleweli@nabijaczleweli.xyz>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

process_madvise.2: Rename parameter, and use array syntax

Signed-off-by: Alejandro Colomar <alx@kernel.org>

process_madvise.2: Document the glibc wrapper

$ grepc -n process_madvise /usr/include/
/usr/include/x86_64-linux-gnu/bits/mman_ext.h:25:
extern __ssize_t process_madvise (int __pid_fd, const struct iovec *__iov,
  size_t __count, int __advice,
  unsigned __flags)
  __THROW;

Reported-by: Florian Weimer <fweimer@redhat.com>
Cc: Carlos O'Donell <carlos@redhat.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

sigaction.2: HISTORY: Some constants are or were XSI extensions

Link: <https://lore.kernel.org/linux-man/20240214095707.1824c25c@plasteblaster/T/>
Reported-by: Thomas Orgis <thomas@orgis.org>
Cc: Eric Blake <eblake@redhat.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

sigaction.2: HISTORY: SA_NOCLDSTOP wasn't added in POSIX.1-2001

It already existed in POSIX.1-1996, according to just a few lines above.

Link: <https://lore.kernel.org/linux-man/20240214095707.1824c25c@plasteblaster/T/>
Cc: Thomas Orgis <thomas@orgis.org>
Cc: Eric Blake <eblake@redhat.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

getdents.2: Add note to misleading field "d_off" in struct linux_dirent64

The getdents.2 man page details a pair syscalls: getdents() and
getdents64(), both of which are used to get the entries of a directory.
The results are populated into a structure, with the difference between
both syscalls being mostly bitwidth related.

However, the behaviour or the 'd_off' field in both struct linux_dirent
and linux_dirent64 is wrongly documented in this man page.

According to the current manual page, 'd_off' is used to store the
"Offset to the next linux_dirent [...] the distance from the start of
the directory to the start of the next linux_dirent."

This value, though, is filesystem dependent, and much of the time it
stores no such offset.

According to readdir.3 [1] manpage:

 > The value returned in d_off is the same as would be returned by
 > calling telldir(3) at the current position in the directory stream.
 > Be aware that despite its type and name, the d_off field is seldom
 > any kind of directory offset on modern filesystems.  Applications
 > should treat this field as an opaque value, making no assumptions
 > about its contents; see also telldir(3).

Of course, readdir(3) is a glibc function with no ties to
getdents(2), but it was implemented with such syscall and considering
that readdir(3) doesn't process the data from getdents(2) my belief is
that it inherited said behaviour from it [2].  telldir(3) tells a
similar story.

On the example provided at the end of getdents.2, notable is the d_off
value of the very last entry:

--------------- nread=120 ---------------
inode#    file type  d_reclen  d_off   d_name
       2  directory    16         12  .
       2  directory    16         24  ..
      11  directory    24         44  lost+found
      12  regular      16         56  a
  228929  directory    16         68  sub
   16353  directory    16         80  sub2
  130817  directory    16       4096  sub3

which makes a very sudden jump that is obviously not where the entry is
located.

Rerunning this same example but on a ext4 partition gives you garbage
values:

--------------- nread=176 ---------------
inode#    file type  d_reclen  d_off   d_name
    2050  directory    24 4842312636391754590  sub2
       2  directory    24 4844777444668968292  ..
    2051  directory    24 7251781863886579875  sub3
      12  regular      24 7470722685224223838  a
    2049  directory    24 7653193867028490235  sub
      11  directory    32 7925945214358802294  lost+found
       2  directory    24 9223372036854775807  .

In fact, I've had a hard time reproducing nice d_off values on ext2 too,
so what the filesystem does with d_off must have change since then.

On tmpfs it's a count:

--------------- nread=144 ---------------
inode#    file type  d_reclen  d_off   d_name
       1  directory    24          1  .
       1  directory    24          2  ..
       5  directory    24          3  sub3
       4  directory    24          4  sub2
       3  directory    24          5  sub
       2  regular      24          6  a

I've also not been the first to notice this, as you can see from this
stackoverflow issue opened last year:

https://stackoverflow.com/q/75119224

Safe to say, it's a very unreliable field.

Below is a patch that adds a warning besides the d_off field in both
structures, plus a brief explanation on why this field can be mislea-
ding (while also directing the user towards the readdir.3 man page).

Link: [1] <https://git.kernel.org/pub/scm/docs/man-pages/man-pages.git/tree/man3/readdir.3>
Link: [2] <https://elixir.bootlin.com/glibc/glibc-2.39/source/sysdeps/unix/sysv/linux/readdir.c>
Signed-off-by: Vinícius Schütz Piva <vinicius.vsczpv@outlook.com>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: srcfix

Use a variable for the options passed to recursive make(1), to avoid
repetition.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: dist: Show CP instead of INSTALL

We're using install(1), but it's just an implementation detail.
Since we're not installing into the system, CP is less confusing.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: dist: FORCE regeneration of version file if necessary

If any of $DISTVERSION, $DISTNAME, or $DISTDATE have changed since the
last 'make dist', force regeneration of the version file, even if it
wouldn't change due to normal dependencies.  This makes sure that the
tarball has correct values.

It doesn't need to depend on all $DISTFILES.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: $DISTDATE: If the repository is "dirty", use the newest timestamp

Otherwise, the files within the 'dist' tarball will have a timestamp
older than their last actual modification, which is problematic with
'distcheck'.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: dist: Preserve links with cp(1)

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Consistently use 'cp -T'

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/configure/version.mk: $DISTNAME: Pass --dirty to git-describe(1)

This makes sure we don't accidentally produce release tarballs from a
dirty repository.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/configure/version.mk: $DISTNAME: Don't hide git(1) errors

We don't call git(1) inside tarballs anymore to get the $DISTNAME, so we
can safely assume that git(1) should never fail, and if it fails, we
better get an error message.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck: Remove old directory before untarring the new tarball

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck: Fix dependencies

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Fix path

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck: Support an absolute path in $builddir

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck: Use temporary dirs within $builddir

Signed-off-by: Alejandro Colomar <alx@kernel.org>

close_range.2: srcfix

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Use a variable to prefix recursive make(1) output

Piping the output to sed(1) didn't behave well with -Orecurse.  Using a
variable behaves well, keeping output synchronized.  The length of the
variable name is specific, so that `$(INFO_)` uses exactly 8 characters,
a tab.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

GNUmakefile: help: Document 'distcheck'

Signed-off-by: Alejandro Colomar <alx@kernel.org>

close_range.2: Add _GNU_SOURCE and <unistd.h> to SYNOPSIS

close_range() is defined in <unistd.h> when _GNU_SOURCE is defined.
The <linux/close_range.h> header file only defines the (linux-specific)
flags constants.  The flags argument is an int, not an unsigned int, in
the glibc wrapper.  Use the close_range() library call in the example
code instead of syscall().

Fixes: 71a62d6c3c56 ("close_range.2: Glibc added a wrapper recently")
Fixes: c2356ba085ed ("close_range.2: Glibc 2.34 has added a close_range() wrapper")
Reported-by: Alexandra Hájková <ahajkova@redhat.com>
Signed-off-by: Mark Wielaard <mark@klomp.org>
[alx: ffix]
Signed-off-by: Alejandro Colomar <alx@kernel.org>

getaddrinfo.3: tfix

(Incidentally, the glibc function source does name the parameter
"name" rather than "node", unlike this man page.  (The POSIX man
page uses "nodename".))

Signed-off-by: Štěpán Němec <stepnem@smrk.net>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

man*/: tfix

Signed-off-by: Štěpán Němec <stepnem@smrk.net>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

path_resolution.7: Work around groff(1) bug

Link: <https://savannah.gnu.org/bugs/?65320>
Link: <https://savannah.gnu.org/bugs/?65322>
Link: <https://lists.gnu.org/archive/html/groff/2024-02/msg00039.html>
Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: clean: Also remove 'distcheck' tmp dirs

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: $(info ...): Make output more consistent

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Use $(TRUE) instead of ':'

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/dist/: dist: Support recursive 'dist'

Hardcode the version and date, and don't use git(1) for selecting the
distribution files.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: distcheck: Add target

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Reorganize build system

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Move configuration variables to share/mk/configure/

Some variables are only part of the implementation of our build system,
and users should not modify them; others are designed to be set by users
when they invoke make(1).  Define the latter in share/mk/configure/, so
that they are more visible.

Signed-off-by: Alejandro Colomar <alx@kernel.org>

share/mk/: Move '.SILENT:' to GNUmakefile

Signed-off-by: Alejandro Colomar <alx@kernel.org>