nmeter: fix a bug with unterminated varargs

function                                             old     new   delta
collect_mem                                          361     371     +10
collect_swp                                          116     120      +4
vrdval                                               168     170      +2
collect_thread_nr                                     63      65      +2
collect_int                                          121     123      +2
collect_if                                           205     207      +2
collect_fork                                         117     119      +2
collect_fd                                            79      81      +2
collect_ctx                                          117     119      +2
collect_cpu                                          621     623      +2
collect_blk                                          557     559      +2
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 11/0 up/down: 32/0)              Total: 32 bytes

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

New example config: android_502_defconfig

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

more bionic fixes

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

fixes for problems found by bionic build

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

udhcpc: fix a warning in debug code

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

udhcp: fix a SEGV on malformed RFC1035-encoded domain name

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

renice: tweak help text

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

ls: fix columnar output. Closes 8731

In coreutils/ls.c, 1.19 introduced commit
2f7d9e8903029b1b5e51a15f9cb0dcb6ca17c3ac, removing the variable tabstops and
hard coding the column separation to 2 characters, but was not done correctly.
The column_width assumes a gap of 1 character, so the computed number of
columns exceeds the terminal width when many small files are encountered.

A minor problem but surprisingly annoying.

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

inotifyd: swap meaning of 'y' and 'm' events in help text. Closes 8726

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

udhcpc: do not use -t NUM for counting "select" packets, use 3

Otherwise, "-t 0" usage may end up sending them forever
if server does not respond.

function                                             old     new   delta
udhcpc_main                                         2846    2836     -10

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

ntpd: add experimental patch

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

ntpd: if peer does not reply anymore, try re-resolving its hostname

function                                             old     new   delta
ntpd_main                                           1053    1130     +77
add_peers                                            166     195     +29

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

ntpd: do not use a peer more than once (say, if two peers resolve to the same IP)

function                                             old     new   delta
add_peers                                             98     166     +68

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

ntpd: more informative poll lowering message

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

ifupdowm: fix "warning: unused variable 'iface_list'"

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

dhcpc: Use client IP address as source address for DHCP renew/rebind messages

RFC2131 paragraph 4.1 states DHCP messages broadcast by a client prior to
that client obtaining its IP address must have the source IP address
field in the header set to 0.
Request messages transmitted in renewing and rebinding state need to use
the obtained IP address as source IP address in the header; this behavior
lines up with other implementations like ISC dhcp client.

Signed-off-by: Hans Dedecker <dedeckeh@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

udhcpd: keep expired leases at startup

Let udhcpd retain the information about expired leases when restarting
so that the leases are reserved until they possibly become the oldest
expired lease.

This reduces the frequency of IP address changes for example when the
DHCP server and a group of clients, who do not store and request their
previously offered IP address across restarts, are collectively restarted
and the startup order of the clients are not guaranteed.

Signed-off-by: Christian Lindeberg <christian.lindeberg@axis.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ifupdown: allow duplicate interface definitions

This patch allow to have multiple interface definitions, much like
Debian's ifupdown.  More specifically, it removes the check for a
duplicate definition, so the impact on binary size should be fairly
minimal.

This configuration:

iface eth0 inet static
        address 192.168.0.15
        netmask 255.255.0.0
        gateway 192.168.0.1

iface eth0 inet static
        address 10.0.0.1
        netmask 255.255.255.0

Will add two addresses to eth0 if ip is used.  If ifconfig is used,
the standards methods will likely not stack, but the administrator may
still use the manual method.  The DHCP method may work depending on the
DHCP client in use.

This is a fairly advanced feature for power users who knows what they
are doing.  There are not many other network configuration systems that
allows multiple addresses on an interface.

Signed-off-by: Nicolas Cavallari <nicolas.cavallari@green-communications.fr>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

udhcpc: fix OPTION_6RD parsing (could overflow its malloced buffer)

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

udhcp: trivial shrink

function                                             old     new   delta
dname_dec                                            337     332      -5

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

taskset: fix non-fancy cpuset printing on big-endian

The non-fancy version of the from_cpuset uses CPU_SETSIZE as if it
represents the number of bytes in the cpuset, while it is actually
the number of bits. This leads to out-of-bounds accesses on the
cpu_set_t in the big-endian case. Basically all uses of CPU_SETSIZE
have to be divided by 8. This is done correctly in the fancy version
of from_cpuset.

In addition, the big-endian case is completely wrong to begin with.
All standard C libraries that I know of implement cpu_set_t as an
unsigned long array, so both for big and little endian, the least
significant bits are in the beginning of the array. Therefore, the
approach taken for the little endian case is equally valid. We only
need special handling for big endian when CPU_SETSIZE is large and
we use an unsigned long long to get more bits out.

Signed-off-by: Arnout Vandecappelle (Essensium/Mind) <arnout@mind.be>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ntpd: print packet delay in clock update message

function                                             old     new   delta
update_local_clock                                   820     826      +6

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

modprobe: only parse files that end in .conf

This matches behavior with kmod which has been the standard for a long
time at this point.

URL: https://bugs.busybox.net/8021
Reported-by: Jö <jorrit@jorrit.de>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

build: add a sanitizer debug option

Building & running with ASAN is super helpful, so add a dedicated config
knob for it.  This way people don't have to guess at the right compiler
settings in order to get a good build.  We can just tell people to enable
this one option.

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

networking: allow dot at the end of the domain name in dhcp response

Patch based on Balaji Punnuru <balaji.punnuru@gmail.com> work.

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

init: make the command-line rewrite optional

When launched as PID 1 and after parsing its arguments, init wipes all
all of them except argv[0] and rewrites argv[0] to contain only "init",
so that its command-line appears solely as "init" in tools such as ps.

This patch adds the FEATURE_INIT_MODIFY_CMDLINE which, if set to n, will
make init preserve all its arguments including argv[0], be they parsed or
ignored.
The original command-line used to launch init can then be retrieved in
/proc/1/cmdline on Linux, for example.

Signed-off-by: Nicolas Carrier <nicolas.carrier@parrot.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ntpd: shorter message

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

ntpd: step when |offset| > 1 sec, not 0.125 sec

update_local_clock                                   769     820     +51
recv_and_process_peer_pkt                            838     862     +24
reset_peer_stats                                     137     133      -4
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/1 up/down: 75/-4)              Total: 71 bytes

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

busybox: fix uninitialized memory when displaying IPv6 addresses

After commit 8e74adab0107658e3dc04ed342206272a284f43e
("libbb: make INET[6]_rresolve use sockaddr2{host,dotted}_noport")
INET_sprint6 uses more than just sin6_addr, it also tries to display the
scope id, which is uninitialized when called from ife_print6.

Signed-off-by: Felix Fietkau <nbd@openwrt.org>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

dd: support iflag=skip_bytes

It allows specifying amount of bytes directly (not only amount of
blocks) is also supported by GNU's Coreutils.

function                                             old     new   delta
parse_comma_flags                                      -      93     +93
static.iflag_words                                     -      12     +12
dd_main                                             1569    1580     +11
packed_usage                                       30591   30600      +9
------------------------------------------------------------------------------
(add/remove: 2/0 grow/shrink: 2/0 up/down: 125/0)             Total: 125 bytes

Signed-off-by: Rafał Miłecki <zajec5@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Fix compiling with musl's utmp stubs

This patch fixes compiling busybox with FEATURE_UTMP and _WTMP enabled.
musl, while not really support utmp/wtmp, provides stub functions, as well
as variables such as _PATH_UTMP, so that programs using utmp or wtmp can
still compile fine.

My reasoning for this patch is that on Exherbo, I'm currently trying to get
us to be able to use the same busybox config file for both glibc and musl
systems, using utmp/wtmp on systems that support it, and using the stubs
on musl without needing two different configs.

As of latest musl git, it provides all utmp functions needed; 1.1.12 doesn't,
but I sent a patch to Rich to add the utmp{,x}name functions expected to
exist, which was merged into musl upstream.

Signed-off-by: Kylie McClain <somasissounds@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

printf: short-circuit output when argument to %b includes \c

printf wasn't correctly handling \c in an argument to the %b format
specifier.

   printf %bXX OK\\c

returned 'OK\cXX' rather than the expected 'OK'.

function                                             old     new   delta
printf_main                                          886     899     +13

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

sed: make 's///w FILE' actually write to FILE. Closes 8251

function                                             old     new   delta
add_cmd                                             1167    1210     +43

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

patch: correctly handle "patch FILE_TO_PATCH" form. Closes 7736

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

dos2unix: try to preserve ownership. closes 8311

function                                             old     new   delta
dos2unix_main                                        426     441     +15

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

setsid: implement -c

function                                             old     new   delta
setsid_main                                           53      96     +43
packed_usage                                       30846   30833     -13

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

truncate: use O_WRONLY|O_NONBLOCK

This matches coreutils behavior.  We don't read the fd, and truncation
does not need blocking.

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

truncate: always set mode when opening file to avoid fortify errors

Busybox crashes due to no mode being given when opening:
$ ./busybox truncate -s 1M foo
*** invalid open64 call: O_CREAT without mode ***: ./busybox terminated
======= Backtrace: =========
/lib/x86_64-linux-gnu/libc.so.6(+0x7338f)[0x7f66d921338f]
/lib/x86_64-linux-gnu/libc.so.6(__fortify_fail+0x5c)[0x7f66d92aac9c]
/lib/x86_64-linux-gnu/libc.so.6(+0xeb6aa)[0x7f66d928b6aa]
./busybox[0x4899f9]
======= Memory map: ========
00400000-004d0000 r-xp 00000000 00:1a 137559                             /home/ari/busybox/busybox
006cf000-006d0000 r--p 000cf000 00:1a 137559                             /home/ari/busybox/busybox
006d0000-006d1000 rw-p 000d0000 00:1a 137559                             /home/ari/busybox/busybox
006d1000-006d4000 rw-p 00000000 00:00 0
014e7000-01508000 rw-p 00000000 00:00 0                                  [heap]
7f66d8f8a000-7f66d8fa0000 r-xp 00000000 08:07 1579008                    /lib/x86_64-linux-gnu/libgcc_s.so.1
7f66d8fa0000-7f66d919f000 ---p 00016000 08:07 1579008                    /lib/x86_64-linux-gnu/libgcc_s.so.1
7f66d919f000-7f66d91a0000 rw-p 00015000 08:07 1579008                    /lib/x86_64-linux-gnu/libgcc_s.so.1
7f66d91a0000-7f66d935b000 r-xp 00000000 08:07 1578994                    /lib/x86_64-linux-gnu/libc-2.19.so
7f66d935b000-7f66d955a000 ---p 001bb000 08:07 1578994                    /lib/x86_64-linux-gnu/libc-2.19.so
7f66d955a000-7f66d955e000 r--p 001ba000 08:07 1578994                    /lib/x86_64-linux-gnu/libc-2.19.so
7f66d955e000-7f66d9560000 rw-p 001be000 08:07 1578994                    /lib/x86_64-linux-gnu/libc-2.19.so
7f66d9560000-7f66d9565000 rw-p 00000000 00:00 0
7f66d9565000-7f66d966a000 r-xp 00000000 08:07 1579020                    /lib/x86_64-linux-gnu/libm-2.19.so
7f66d966a000-7f66d9869000 ---p 00105000 08:07 1579020                    /lib/x86_64-linux-gnu/libm-2.19.so
7f66d9869000-7f66d986a000 r--p 00104000 08:07 1579020                    /lib/x86_64-linux-gnu/libm-2.19.so
7f66d986a000-7f66d986b000 rw-p 00105000 08:07 1579020                    /lib/x86_64-linux-gnu/libm-2.19.so
7f66d986b000-7f66d988e000 r-xp 00000000 08:07 1578981                    /lib/x86_64-linux-gnu/ld-2.19.so
7f66d9a64000-7f66d9a67000 rw-p 00000000 00:00 0
7f66d9a8a000-7f66d9a8d000 rw-p 00000000 00:00 0
7f66d9a8d000-7f66d9a8e000 r--p 00022000 08:07 1578981                    /lib/x86_64-linux-gnu/ld-2.19.so
7f66d9a8e000-7f66d9a8f000 rw-p 00023000 08:07 1578981                    /lib/x86_64-linux-gnu/ld-2.19.so
7f66d9a8f000-7f66d9a90000 rw-p 00000000 00:00 0
7ffc47761000-7ffc47782000 rw-p 00000000 00:00 0                          [stack]
7ffc477ab000-7ffc477ad000 r-xp 00000000 00:00 0                          [vdso]
ffffffffff600000-ffffffffff601000 r-xp 00000000 00:00 0                  [vsyscall]
Aborted (core dumped)
$

Fix this by simply always setting the mode, as it doesn't hurt even
when O_CREAT is not specified.

This bug is a regression introduced in fc3e40e, as xopen(), which
was originally used, would automatically set the mode.

Signed-off-by: Ari Sundholm <ari@tuxera.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

login: add commented-out PAM double password avoidance from BZ 4003

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

nandwrite: implement -n (read/write without ecc)

Implement -n (read/write without ecc).

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

blkdiscard: new applet

function                                             old     new   delta
blkdiscard_main                                        -     264    +264

Signed-off-by: Ari Sundholm <ari@tuxera.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: suppress a compilation warning

Reported by gcc (Debian 5.3.1-4) 5.3.1 20151219

shell/ash.c: In function 'evaltree':
shell/ash.c:8432:19: warning: logical not is only applied to the left hand side of comparison

Signed-off-by: Cristian Ionescu-Idbohrn <cristian.ionescu-idbohrn@axis.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

mount: support "nofail" option. Closes 8551

function                                             old     new   delta
singlemount                                         1045    1060     +15
mount_option_str                                     338     345      +7
mount_it_now                                         355     361      +6
mount_options                                        172     176      +4
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 4/0 up/down: 32/0)               Total: 32 bytes

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

iproute: support "scope". Closes 8561

function                                             old     new   delta
iproute_modify                                      1051    1120     +69

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

swaponoff: fix compile-time warning

  CC      util-linux/swaponoff.o
cc1: warnings being treated as errors
util-linux/swaponoff.c: In function 'swap_enable_disable':
util-linux/swaponoff.c:100: warning: passing argument 1 of 'resolve_mount_spec' from incompatible pointer type
make[1]: *** [util-linux/swaponoff.o] Error 1

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

chpasswd: support -c argument and respect DEFAULT_PASSWD_ALGO

Signed-off-by: Pascal Bach <pascal.bach@siemens.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

include/platform.h: Remove extra #endif introduced in 6df9612.

It causes the compilation to fail.

Signed-off-by: Ari Sundholm <ari@tuxera.com>
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

Resolve linker issues with Android API 21 (dprintf, tcdrain)

Signed-off-by: Chris Renshaw <osm0sis@outlook.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

blkid: add bcache support

URL: https://bugs.gentoo.org/508596
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

swapon/swapoff: refine the -e (ifexists) option

The -e option should only apply to swapon, and it should swallow all
errors/warnings when the device does not exist.  So delete the flag
from the swapoff patch and unify the check in the swapoff path.

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

trylink: include LDFLAGS when checking linkage

The user might be including options in their LDFLAGS (like -fuse-ld=gold)
that change the behavior of the linker and thus change the results of the
flag tests.  Make sure we include the user's LDFLAGS when running these
tests so we filter out flags that will fail when used later on.

URL: https://bugs.gentoo.org/499712
Signed-off-by: Mike Frysinger <vapier@gentoo.org>

modutils: fix build error with !DEPMOD

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

modutils: merge module_entry and module_info to common

This merges the in-memory module info structures of modprobe
and depmod. This allows sharing hashing by modulename code
improving depmod runtime with almost factor of 2x.

function                                             old     new   delta
get_or_add_modentry                                    -      17     +17
do_modprobe                                          590     601     +11
moddb_get_or_create                                    -      10     +10
load_modules_dep                                     195     205     +10
moddb_get                                              -       7      +7
add_probe                                             81      78      -3
modprobe_main                                        721     714      -7
depmod_main                                          553     543     -10
config_file_action                                   434     421     -13
helper_get_module                                    160     144     -16
parse_module                                         343     320     -23
order_dep_list                                       105      82     -23
------------------------------------------------------------------------------
(add/remove: 3/0 grow/shrink: 2/7 up/down: 55/-95)            Total: -40 bytes

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

TODO file: remove mpstat, iostat, powertop - we have them now

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

ash: add support for bash 'function' keyword

Where the POSIX shell allows functions to be defined as:

   name () compound-command [ redirections ]

bash adds the alternative syntax:

   function name [()] compound-command [ redirections ]

Implement this in ash's bash compatibility mode.  Most compound
commands work (for/while/until/if/case/[[]]/{}); one exception is:

   function f (echo "no way!")

The other two variants work:

   f() (echo "ok")
   function f() (echo "also ok")

function                                             old     new   delta
parse_command                                       1555    1744    +189
tokname_array                                        232     240      +8
.rodata                                           155612  155566     -46
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/1 up/down: 197/-46)           Total: 151 bytes

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

hush-misc/func_args1.tests: remove "UNFIXED BUG", it does not fail

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

ash: copy function tests from hush testsuite

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

i2c_tools: suppress "'blen' may be used uninitialized in this function"

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

i2cdump: display the numeric value for block read ioctl() errors

This makes busybox i2cdump compatible with the upstream version, which
also displays the numeric error value in case of a block read failure.

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

i2cdump: bail-out if block read fails

We should bail-out if i2c_smbus_read_block_data() or
i2c_smbus_read_i2c_block_data() return 0 or less. Add the missing check
for the former and fix the existing for the latter.

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

i2cdump: use I2C block mode for the 'i' mode parameter

Currently we're calling i2c_smbus_read_block_data() for both 'i' and 's'
mode parameters. If the bus doesn't support SMBus block mode, then the
i2c access ioctl() fails. Make i2cdump behave compatibly with upstream
version by calling i2c_smbus_read_i2c_block_data() for I2C block.

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

i2cdump: don't read block data in non-block modes

We currently read data twice in byte mode. Add a check to avoid calling
i2c_smbus_read_i2c_block_data() if we're not in I2C or SMBus block mode.

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

i2cdetect: coding style: add a space after 'for'

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

i2cdetect: fix address skipping in auto mode

If the bus doesn't support SMBus Quick Write or Receive Byte commands
and we're running in auto mode all addresses will be skipped resulting
in an empty table being printed.

This is caused by not restoring the auto mode after it's been changed
for certain address ranges - we need an additional variable to hold the
temporary state.

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

ifupdown: use -x hostname:NAME with udhcpc

The -H NAME is deprecated in udhcpc. See commit
2017d48c0d70bef8768efb42909e605ea8eb5a21

Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

[g]unzip: fix recent breakage.

Also, do emit error message we so painstakingly pass from gzip internals

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

ash: allow popredir to be called if the stack is empty

If /tmp/test.sh is a script that tries to run a second script which
happens to be non-executable this:

   command . /tmp/test.sh

causes a seg fault.

This is because clearredir is called in the error path to clear the
stack of redirections.  The normal path then calls popredir, but popredir
fails when the stack is empty.

Reported-by: Bastian Bittorf <bittorf@bluebottle.com>
Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: fix EXEXEC status clobbering

evalcommand always clobbers the exit status in case of an EXEXEC
which means that exec always fails with exit status 2 regardless
of what it actually returns.

This patch adds the missing check for EXEXEC so that the correct
exit status is preserved.  It causes the test ash-misc/exec.tests
to succeed.

Based on commit 7f68426 in dash git, by Herbert Xu.

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

ash: add tests for failures of the exec and command builtins

The exec builtin should return an exit status of 127 if the command
can't be found.  It doesn't:  it returns 2.

If the command builtin is used to source a script that runs a second
script that doesn't exist ash should issue an error.  Instead it seg
faults.

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

stat: make -f optional. This allows to build stat for non-linux systems

function                                             old     new   delta
packed_usage                                       30761   30706     -55

Based on the patch by Ron Yorston.

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

lineedit: FEATURE_REVERSE_SEARCH should not depend on SAVEHISTORY

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

lineedit: search applets as well as PATH for tab completion

In standalone shell mode search the applet table as well as PATH
when tab completing a command.

Use a stupid linear search:  we're also about to read all the
directories on PATH so efficiency isn't a big concern.

function                                             old     new   delta
add_match                                              -      53     +53
complete_cmd_dir_file                                687     724     +37
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 1/0 up/down: 90/0)               Total: 90 bytes

Signed-off-by: Ron Yorston <rmy@frippery.org>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

busybox: alter help message in standalone shell mode

Signed-off-by: Ron Yorston <rmy@frippery.org>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

hush: copy for.tests from ash testsuite

This test passes for hush

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

ash: simplify EOF/newline handling in list parser

Processing of here documents in ash has had a couple of breakages
which are now the subject of tests.  This commit should fix both.

It is based on the following commit in dash git by Herbert Xu:

   <7c245aa> [PARSER] Simplify EOF/newline handling in list parser

(See git://git.kernel.org/pub/scm/utils/dash/dash.git)

Reported-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

Revert "ash: fix a SEGV case in an invalid heredoc" xxx

This reverts commit 7e66102f762a7d80715f0c7e5925433256b78cee but
leaves the test in place as it's still valid.

Reported-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ash: fix error during recursive processing of here document

Save the value of the checkkwd flag to prevent it being clobbered
during recursion.

Based on commit ec2c84d from git://git.kernel.org/pub/scm/utils/dash/dash.git
by Herbert Xu.

function                                             old     new   delta
readtoken                                            190     203     +13
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/0 up/down: 13/0)               Total: 13 bytes

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

ash tests: make test print test name before it is run

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

ash: only allow local variables in functions

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

ash: respect -p flag when command builtin is run with -v/-V

The command builtin should only check the default path, not $PATH,
when the -p flag is used along with -v/-V.

Based on commits 65ae84b (by Harald van Dijk) and 29ee27d (by Herbert
Xu) from git://git.kernel.org/pub/scm/utils/dash/dash.git).

function                                             old     new   delta
commandcmd                                            72      87     +15
describe_command                                     437     450     +13
typecmd                                               84      86      +2
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 3/0 up/down: 30/0)               Total: 30 bytes

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

ash: fix command -- crash

busybox sh -c 'command --' segfaults because parse_command_args
returns a pointer to a null pointer.

Based on commit 18071c7 from git://git.kernel.org/pub/scm/utils/dash/dash.git
by Gerrit Pape.

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

ash: save a few bytes in code to parse case statements

Based on commit 49b82fc from git://git.kernel.org/pub/scm/utils/dash/dash.git
by Herbert Xu.

function                                             old     new   delta
parse_command                                       1563    1555      -8
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-8)               Total: -8 bytes

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

ash: allow newline after variable name in for loop

Newline is a valid delimiter between the variable name and `in`
keyword in for loops.

Based on commit 22e8fb4 from git://git.kernel.org/pub/scm/utils/dash/dash.git
by Herbert Xu.

function                                             old     new   delta
parse_command                                       1568    1563      -5
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-5)               Total: -5 bytes

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

ash: add test for issue with here document

This used to work but doesn't now:

foo () {
cat <<EOF && { echo "$1" ; }
$1
EOF
}

foo "bar"

Reported-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

modinfo: fix "-F firmware", add "intree" field display

function                                             old     new   delta
shortcuts                                              -      52     +52
modinfo                                              317     330     +13
display                                               77      87     +10
packed_usage                                       30752   30761      +9
modinfo_main                                         351     345      -6
static.shortcuts                                      48       -     -48
------------------------------------------------------------------------------
(add/remove: 1/1 grow/shrink: 3/1 up/down: 84/-54)             Total: 30 bytes

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

lzop: eliminate variable, use "int" as return type

Based on patch by Maxin B. John <maxin.john@intel.com>

function                                             old     new   delta
pack_lzop                                            870     859     -11

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

fbsplash: use virtual y size in mmap size calculations

The virtual y can be larger - and we can be even writing there since
we are taking into account the y offset. Avoids possible crash.
But use it only if set, seems it is not set if virtual area is not
allocated (though, often fbcon allocates some scrollback area).

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

gzip: add support for --no-name long option

Add support for --no-name long option. Just silently ignore it
like the short -n option.

This allows to use busybox gzip with Lynx browser.

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

gunzip: add support for long options

Add support for long options.

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

unzip: test for bad archive SEGVing

function                                             old     new   delta
huft_build                                          1296    1300      +4

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

tr: support octal ranges

now, we can do printf "a\tb\tcdef\n" | ./busybox tr -d "\1-\14b-e"
af

and bonus, we save some bytes.

function                                             old     new   delta
expand                                               718     699     -19
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-19)             Total: -19
bytes

Signed-off-by: Richard Genoud <richard.genoud@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

ifupdown: pass interface device name for ipv6 route commands

IPv6 routes need the device argument for link-local routes, or they
cannot be used at all. E.g. "gateway fe80::def" seems to be used in
some places, but kernel refuses to insert the route unless device
name is explicitly specified in the route addition.

Signed-off-by: Timo Teräs <timo.teras@iki.fi>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

tweak defconfig

MONOTONIC_SYSCALL=y by default

FEATURE_LAST_SMALL is gone: now FEATURE_LAST_FANCY is a "bool", not a "choice".

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

inetd: make FEATURE_INETD_RPC off by default

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

libiproute: use if_nametoindex

Saves 87 bytes.  Assuming, of course, all platforms have it.

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

xargs: make -I imply -r

Make -I imply -r (GNU findutils seems to do the same).

Fixes the following bug:

        $ echo -n | xargs -I% echo %
        Segmentation fault

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

dumpleases: string reuse

   text    data     bss     dec     hex filename
 926254     906   17160  944320   e68c0 busybox_old
 926242     906   17160  944308   e68b4 busybox_unstripped

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

dumpleases: new option -d to show time in seconds

function                                             old     new   delta
dumpleases_main                                      493     534     +41
static.dumpleases_longopts                            31      41     +10
packed_usage                                       30777   30752     -25
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/1 up/down: 51/-25)             Total: 26 bytes

Signed-off-by: Isaac Dunham <ibid.ag@gmail.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>

zcip: ...ad suppress the warning

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

zcip: do not query current time if we won't use the result

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