ubi_tools: extend supported volume size to 64 bits; shrink code

function                                             old     new   delta
get_num_from_file                                      -      90     +90
packed_usage                                       29395   29377     -18
ubi_tools_main                                      1349    1141    -208
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 0/2 up/down: 90/-226)          Total: -136 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ubimkvol: add -m option to create volume of maximum size

Signed-off-by: Paul B. Henson <henson@acm.org>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

pstree: fix theoretically unsafe code

In practice, p->comm is never long enough to trigger the bug, but still.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ping: code shrink

function                                             old     new   delta
unpack_tail                                          262     243     -19

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Move create_icmp[6]_socket to its only user, and simplify it

function                                             old     new   delta
run_applet_and_exit                                  711     714      +3
sendping_tail                                        239     236      -3
common_ping_main                                    1798    1770     -28
create_icmp_socket                                    65       -     -65
create_icmp6_socket                                   65       -     -65
------------------------------------------------------------------------------
(add/remove: 0/4 grow/shrink: 1/2 up/down: 3/-161)           Total: -158 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

losetup: assorted fixes. Closes 6314

"losetup -d" was not complaining that LOOPDEV is missing.
"losetup -a" was listing only up to /dev/loop9.
"losetup -f" looped forever if llop0 was taken, and never tried
anything after /dev/loop9.
"-o" with other options (say, -r) had no effect.

function                                             old     new   delta
losetup_main                                         376     419     +43

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

lsof: correct check for symbolic link

Busybox lsof used the d_type field of a 'struct dirent' to verify whether the
entry is a symbolic link. This field, however, is not portable. On at least
one board [1] I have seen, that field is 0, and the check fails even though
the entry is a link.

The explicit check for a symbolic link is really only needed to skip the
default directory entries '.' and '..'. The directory /proc/<pid>/fd/
should not contain anything else but these two and symbolic links.
With these assumptions, this patch replaces the explicit link check with a
basic check for '.' and '..' (and any hidden file). In the unlikely case that
there are other file types, xmalloc_readlink() will return NULL, and we can
skip the entry.

[1] A MIPS-based board with glibc 2.9, Linux 2.6.32.27.

Signed-off-by: Thomas De Schampheleire <thomas.de.schampheleire@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

nslookup: set default DNS server again. Hopefully helps with 675

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

md5/sha512: a better fix for strict aliasing warnings

The locations *are* well-aligned for direct stores
on any architecture.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

md5/sha512: fix strict aliasing warnings

If the target can tolerate these issues, then gcc is smart enough
to generate the same code (x86_64 produces the same code).  If the
target can't, then it needs the memcpy anyways.

libbb/hash_md5_sha.c: In function 'common64_end':
libbb/hash_md5_sha.c:87:4: warning:
  dereferencing type-punned pointer will break strict-aliasing rules
    *(uint64_t *) (&ctx->wbuffer[64 - 8]) = t;
libbb/hash_md5_sha.c: In function 'sha512_end':
libbb/hash_md5_sha.c:886:4: warning:
  dereferencing type-punned pointer will break strict-aliasing rules
    *(uint64_t *) (&ctx->wbuffer[128 - 8]) = t;
libbb/hash_md5_sha.c:889:4: warning:
  dereferencing type-punned pointer will break strict-aliasing rules
    *(uint64_t *) (&ctx->wbuffer[128 - 16]) = t;

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

pstree: stop truncating thread names

This also fixes a minor buffer overflow when displaying threads as
add_proc() only expects COMM_LEN bytes, but we give it one more than
that.

Reported-by: Dag Wieers <dag@wieers.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ntpd: set offset to 0.0 in "usync" event

function                                             old     new   delta
ntpd_main                                           1030    1026      -4

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ntpd: set G.last_script_run even if script isn't configured

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ntpd: make "unsync" code actually work.

While at it: don't run "periodic" if interrupted by a signal.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ntpd: drop delay outliers; run "unsync" handler if no replies for some time

function                                             old     new   delta
ntpd_main                                            960    1029     +69
recv_and_process_peer_pkt                            834     892     +58
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 127/0)             Total: 127 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sulogin: allow system maintenance login if root password is empty

The current password checking is unable to distinguish between the user
entering an empty password or pressing Control-D. As a result, an empty
password always results in normal startup.

We modify bb_ask to return NULL if Control-D is pressed without entering
a password. The sulogin applet is then modified to only proceed to
normal startup if bb_ask returns NULL. This covers EOF with no password,
interrupt by timeout and ^C.

Signed-off-by: Jonathan Liu <net147@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sulogin: correct lying comment

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

init: remove special-case code for machines with less tham 1 MB of RAM.

function                                             old     new   delta
init_main                                            920     781    -139

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

stty: disable CIBAUD leak check - it is not necessary on Linux

function                                             old     new   delta
stty_main                                           1233    1196     -37

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

stty: code shrink

function                                             old     new   delta
set_mode                                             759     725     -34

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

udhcp: code shrink

function                                             old     new   delta
attach_option                                        416     413      -3
allocate_tempopt_if_needed                            74       -     -74

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

udhcpc: fix a problem with binary-encoded options #2

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

find:: get rid of nested function (it's a gcc-ism)

function                                             old     new   delta
alloc_action                                           -      80     +80
parse_params                                        1459    1445     -14
static.alloc_action                                   98       -     -98
------------------------------------------------------------------------------
(add/remove: 1/1 grow/shrink: 0/1 up/down: 80/-112)           Total: -32 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

grep: fix grep -x to not set REG_NOSUB

When -F isn't specified (and !ENABLE_EXTRA_COMPAT), grep -x uses
regexec's regmatch_t output to determine if the match was the entire
line. However it also set the REG_NOSUB flag which makes it ignore the
regmatch_t argument.

Add an exception to the setting of REG_NOSUB for OPT_x and add some test
cases to test the behaviour of -x.

Signed-off-by: James Hogan <james.hogan@imgtec.com>
Cc: Natanael Copa <ncopa@alpinelinux.org>
Cc: Denys Vlasenko <vda.linux@googlemail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

volume_id_internal.h: drop unoptimized byteswap routines

function                                             old     new   delta
volume_id_probe_hfs_hfsplus                          802     627    -175

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

grep: don't bail out on first mismatch if '-w' option is set

Signed-off-by: Bartosz Golaszewski <bartekgola@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Fix build failures caused by not compiling open_transformer.c (#2)

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

find: support -HLP

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

mdev: use $DEVNAME in uevent sysfs file for device node name

516530c932bd17d87c9eb4347a490be051e495f4 uses $DEVNAME variable
for device node name. This is fine, but only works for hotplugging,
"mdev -s" will behave differently when DEVNAME and basename(path)
differ.

This patch extracts the DEVNAME from the uevent sysfs file in
make_device(), thus works for hot- and coldplugging; so using
the environment DEVNAME on hotplug events is no longer necessary.

Signed-off-by: Nikolaus Voss <n.voss@weinmann.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: fix build failure if FEATURE_EDITING=y && !HUSH_INTERACTIVE

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Fix build failures caused by not compiling open_transformer.c

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

telnet: code shrink

function                                             old     new   delta
telnet_main                                         1519    1515      -4
con_escape                                           296     285     -11

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

volume_id: fix ntfs attribute search

len is declared as uint32_t but le16_to_cpu macro is used

Signed-off-by: Oliver Metz <oliver.metz@gmx.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

vi: make regexp search case-insensitive if ":set ignorecase" is active

Reported by Dan Moinescu <dan@moinescu.net>.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash,hush: history builtin

function                                             old     new   delta
show_history                                           -      39     +39
builtin_history                                        -      16     +16
historycmd                                             -      13     +13
bltins1                                              312     324     +12
builtintab                                           336     344      +8
popstring                                            134     140      +6
hush_main                                           1048    1046      -2
ash_main                                            1398    1396      -2
size_from_HISTFILESIZE                                44      40      -4
------------------------------------------------------------------------------
(add/remove: 3/0 grow/shrink: 3/3 up/down: 94/-8)              Total: 86 bytes

Signed-off-by: Flemming Madsen <busybox@themadsens.dk>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: fix for "while false && true; do echo BUG; break; done". closes 6170

function                                             old     new   delta
run_list                                             959     941     -18

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

mdev: call index_in_strings on $ACTION only after we checked it for NULL

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

examples/mdev.conf.change_blockdev.sh: update

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

examples: add mdev example for Android phone

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

examples/inittab: fix a few incorrect statements about init

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

lineedit: \W on "/bin" should show "bin", not "/bin"

function                                             old     new   delta
parse_and_put_prompt                                 793     785      -8

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

lineedit: implement \T \t \A \@ prompts escapes, fix \W escape, drop \!

function                                             old     new   delta
parse_and_put_prompt                                 742     793     +51
read_line_input                                     3836    3826     -10

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

libbb: introduce and use strftime_[YYYYMMDD]HHMMSS()

function                                             old     new   delta
strftime_fmt                                           -      53     +53
strftime_YYYYMMDDHHMMSS                                -      12     +12
strftime_HHMMSS                                        -      12     +12
human_time                                            44      43      -1
fmtstr_t                                               9       -      -9
step_time                                            361     345     -16
watch_main                                           261     232     -29
------------------------------------------------------------------------------
(add/remove: 3/1 grow/shrink: 0/3 up/down: 77/-55)             Total: 22 bytes
   text    data     bss     dec     hex filename
 919203     932   17692  937827   e4f63 busybox_old
 919209     932   17692  937833   e4f69 busybox_unstripped

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

lineedit: add handling of \H in prompt

Based on the patch by Arnaud Rébillout <rebillout@syscom.ch>

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hexdump: need to use xstrtoull for off_t

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hexdump: don't unconditionally limit the usable address range

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

man: install to /usr/bin, not /sbin

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

httpd: don't allow tabs and multiple spaces in request string

HTTP standard doesn't allow it and no sane clients should ever use it.

function                                             old     new   delta
handle_incoming_and_exit                            2795    2785     -10

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

setserial: Fix typo: s/baund_rate/baud_rate/g

as noted by Christoph Handel opensource at macht-blau.org

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

sendmail: code shrink on top of previous patches

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sendmail: don't add multiple To: headers

When adding To: header, add only a single header. If there are multiple
addresses, make it multiline.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sendmail: don't add To: header if it already exists

If the message we are sending already has To: header, don't add a new one.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sendmail: support long header fields for recipients

Support long header fields in To:, Cc: and Bcc: headers.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sendmail: support address lists

Headers To:, Cc: and Bcc: may have a list of comma-separated
addresses. Add support for that. Commas inside double quotes are ignored.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sendmail: support addresses inside angle brackets

When we extract addresses from the e-mail, try to first check for an
address inside angle brackets.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sendmail: don't mangle e-mail headers

Leave the original To: and Cc: headers untouched, when we try to extract
addresses from them.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sendmail: avoid sending mail to wrong addresses

If we get an address we cannot parse properly, we currently just strip
the unknown characters and still try to send it. This is considered
harmful as the resulting address may still be valid but different from
what the user originally intended.

Instead, skip sending to an address we cannot fully understand and
print the characters what we have scanned so far. Leading and trailing
whitespace is allowed and silently stripped.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

mkfs_minix: use get_volume_size_in_bytes instead of local version

Hopefully this also closes 4730

function                                             old     new   delta
valid_offset                                          55       -     -55
mkfs_minix_main                                     2925    2674    -251
------------------------------------------------------------------------------
(add/remove: 0/1 grow/shrink: 0/1 up/down: 0/-306)           Total: -306 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

mdev: fix build failure

The error was "error: 'struct globals' has no member named 'parser'"

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

nc: fix build failure ("subscripted value is neither array nor pointer")

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: add comment about failures in source builtin. No code changes.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: source builtin should override $N only if it has args

function                                             old     new   delta
builtin_source                                       174     184     +10

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: read $HOME/.profile instead of $(pwd)/.profile

ash --login should read ~/.profile instead of .profile in the current
directory. I noticed it while trying to figure out why /root/.profile
is only read sometimes.

function                                             old     new   delta
ash_main                                            1374    1398     +24

Signed-off-by: Stefan Hellermann <stefan@the2masters.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: move code to allow setting $HOME in /etc/profile

move HISTFILE=$HOME/.ash_history below reading /etc/profile,
so that /etc/profile can set $HOME. HOME can be unset when
directly invoking ash --login from init without going through
getty.

Signed-off-by: Stefan Hellermann <stefan@the2masters.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

vi: code shrink

function                                             old     new   delta
status_line_bold_errno                                 -      32     +32
colon                                               2891    2873     -18
file_insert                                          354     313     -41
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 0/2 up/down: 32/-59)            Total: -27 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

vi: fix adjustment of buffer on partial file read

The second argument to text_hole_delete was incorrect: it should
be a pointer to the end of the hole.

Signed-off-by: Ron Yorston <rmy@tigress.co.uk>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

sysklogd: trim help text

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

traceroute: free some memory allocated by xzalloc()

Signed-off-by: Guilherme Maciel Ferreira <guilherme.maciel.ferreira@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

flash_eraseall: implement -N

Signed-off-by: Alexander Shiyan <shc_work@mail.ru>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

nanddump: fix build if nandwrite isn't enabled

Signed-off-by: Alexander Shiyan <shc_work@mail.ru>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ping[6]: accept and ignore -n. Code shrink

function                                             old     new   delta
print_stats_and_exit                                 270     232     -38

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

udhcpd: also add -I ADDR to help text

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

udhcpd: add -I LOCAL_ADDR option

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

awk: Fix handling of functions with empty body

Signed-off-by: Bernhard Reutner-Fischer <rep.dot.nop@gmail.com>

losetup: fix util-linux compatibility

Added -a support. Also made sure -f works as follows:

losetup [-r] [-o offset] {-f|loopdev} file

Removed support for 'losetup -r' with no arguments.

Signed-off-by: Mandeep Singh Baines <msb@chromium.org>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

readlink: note that our -f is really -e

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

readlink: uClibc supports automatic allocation too now

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

platform: use KERNEL_VERSION to simplify uClibc version checking

This makes reading the logic (as well as adding new code) a lot simpler,
and fixes one or two cases that were broken due to incorrect sub-version
tests.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

udhcpc: use readlink rather than realpath

The realpath utility requires all paths exist when canonicalizing
symlinks.  If /etc/resolv.conf points to a tmpfs, then it might
not exist initially.  Use `readlink -f` so that we follow all
symlinks that are available.

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

hostid: do not output sign-extended host id. Closes 6056

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

archival: note implicit dependencies between lzop & bbunzip

Signed-off-by: Mike Frysinger <vapier@gentoo.org>

decompress_unlzma: move function, no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

decompress_unlzma: 10% speedup in "small" code

   text    data     bss     dec     hex filename
   1796       0       0    1796     704 decompress_unlzma.o
   1801       0       0    1801     709 decompress_unlzma.o

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

decompress_unlzma: make "fast" version a bit smaller

It is not slower. In fact it seems a tiny bit faster too.

   text    data     bss     dec     hex filename
   2827       0       0    2827     b0b decompress_unlzma.o
   2797       0       0    2797     aed decompress_unlzma.o

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

lzop: fiq -q and OPTION_DECOMPRESS mismatch

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

bbunzip: fix order of flags vs bit defines

Too much code shuffling.

Reported-by: Mandeep Singh Baines <msb@chromium.org>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

bbunzip: ignore the -q flag with the decompressors

The -q flag is used in shell scripts for suppressing output.
Have our applets swallow the flag for compatibility.

Reported-by: Mandeep Singh Baines <msb@chromium.org>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

ifupdown: support "source" stanza in /etc/network/interfaces

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

zcat: fix "zcat FILE" trying to do detection twice

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

zcat: if seamless uncompressors are defined, autodetect file's format

function                                             old     new   delta
bbunpack                                             526     622     +96
packed_usage                                       29335   29341      +6
gunzip_main                                           64      67      +3

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

fix error message on failure to oen /dev/null; fix zcat's help text

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

code shrink in check_errors_in_children()

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

pgrep: fix a "missing closing paren" build error

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ifplugd: make -k send SIGINT, not SIGQUIT

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

pgrep: fix -x option

Because when -x is used (exact match), then we cannot compile the
regular expression with REG_NOSUB. The manual page regcomp(3) states
in section "Byte offsets":

    Unless REG_NOSUB was set for the compilation of the pattern
    buffer, it is possible to obtain substring match addressing
    information.

The problem was detected on an ARM system with glibc 2.16.
Signed-off-by: Bernhard Walle <bernhard@bwalle.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Fix config help text

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

make --help return exitcode 0. Closes 5612

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

run-parts: fix unicode creep in --help

Signed-off-by: Peter Korsgaard <jacmet@sunsite.dk>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

nc_bloaty: support -ll and -lk. Closes 2245

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

nc: don't redirect stderr to network in -e PROG mode

This in incompatible with nc-1.10, but makes a lot of sense.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>