oweals/busybox.git
7 years agogetopt32: move support for "always treat first arg as option" to users (tar/ar)
Denys Vlasenko [Fri, 4 Aug 2017 14:46:17 +0000 (16:46 +0200)]
getopt32: move support for "always treat first arg as option" to users (tar/ar)

Now getopt() never leaks (and never performs) any xmalloc's.

function                                             old     new   delta
ar_main                                              522     556     +34
tar_main                                             986    1014     +28
getopt32                                            1458    1350    -108
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/1 up/down: 62/-108)           Total: -46 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agogetopt32: factor out code to treat all args as options
Denys Vlasenko [Fri, 4 Aug 2017 14:23:42 +0000 (16:23 +0200)]
getopt32: factor out code to treat all args as options

Working towards making getopt32() xmalloc-free

function                                             old     new   delta
make_all_argv_opts                                     -      58     +58
top_main                                             914     912      -2
getopt32                                            1517    1458     -59
------------------------------------------------------------------------------
(add/remove: 2/0 grow/shrink: 0/2 up/down: 58/-61)             Total: -3 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agotweak NOFORK_NOEXEC.lst
Denys Vlasenko [Fri, 4 Aug 2017 14:01:39 +0000 (16:01 +0200)]
tweak NOFORK_NOEXEC.lst

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: BASH_XTRACEFD bashism
Denys Vlasenko [Fri, 4 Aug 2017 13:24:49 +0000 (15:24 +0200)]
ash: BASH_XTRACEFD bashism

Based on patch by Johannes Schindelin <johannes.schindelin@gmx.de>

function                                             old     new   delta
evalcommand                                         1447    1500     +53

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: remove no-longer-used variable
Johannes Schindelin [Fri, 4 Aug 2017 10:16:46 +0000 (12:16 +0200)]
ash: remove no-longer-used variable

As of 035486c75 (ash: significant overhaul of redirect saving logic,
2017-07-31), the sv_pos variable is no longer used (just assigned to,
with no further effect).

Let's just remove it.

Signed-off-by: Johannes Schindelin <johannes.schindelin@gmx.de>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: INT_OFF/INT_ON around run_nofork_applet()
Denys Vlasenko [Fri, 4 Aug 2017 12:50:03 +0000 (14:50 +0200)]
ash: INT_OFF/INT_ON around run_nofork_applet()

function                                             old     new   delta
evalcommand                                         1441    1447      +6

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agosheel: improve comments on signal handling
Denys Vlasenko [Fri, 4 Aug 2017 12:28:16 +0000 (14:28 +0200)]
sheel: improve comments on signal handling

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agomodutils: make them NOEXEC except depmod
Denys Vlasenko [Fri, 4 Aug 2017 00:56:39 +0000 (02:56 +0200)]
modutils: make them NOEXEC except depmod

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agomicrocom: require exactly one non-option
Denys Vlasenko [Fri, 4 Aug 2017 00:01:45 +0000 (02:01 +0200)]
microcom: require exactly one non-option

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: add a fixme comment at run_nofork_applet
Denys Vlasenko [Thu, 3 Aug 2017 17:56:29 +0000 (19:56 +0200)]
ash: add a fixme comment at run_nofork_applet

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agonofork: fix a bug uncovered by hush testsuite (forgotten fflush)
Denys Vlasenko [Thu, 3 Aug 2017 17:52:47 +0000 (19:52 +0200)]
nofork: fix a bug uncovered by hush testsuite (forgotten fflush)

function                                             old     new   delta
run_nofork_applet                                    280     287      +7

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agogetopt: make it NOEXEC
Denys Vlasenko [Thu, 3 Aug 2017 17:30:21 +0000 (19:30 +0200)]
getopt: make it NOEXEC

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agonew NOFORKs: pwdx,kill[all5],ttysize,realpath,readlink NOEXECs: date,resize
Denys Vlasenko [Thu, 3 Aug 2017 17:00:01 +0000 (19:00 +0200)]
new NOFORKs: pwdx,kill[all5],ttysize,realpath,readlink NOEXECs: date,resize

function                                             old     new   delta
run_nofork_applet                                    258     280     +22
readlink_main                                        112     123     +11
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 33/0)               Total: 33 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoconfig: FEDORA_COMPAT option (so far only tweaks uname)
Denys Vlasenko [Thu, 3 Aug 2017 01:46:14 +0000 (03:46 +0200)]
config: FEDORA_COMPAT option (so far only tweaks uname)

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agonew NOFORKs: clear, nproc, tty, uname, arch, unlink, which
Denys Vlasenko [Thu, 3 Aug 2017 01:29:32 +0000 (03:29 +0200)]
new NOFORKs: clear, nproc, tty, uname, arch, unlink, which

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: do not accept "if() { echo; }" function def
Denys Vlasenko [Wed, 2 Aug 2017 17:44:05 +0000 (19:44 +0200)]
hush: do not accept "if() { echo; }" function def

function                                             old     new   delta
parse_stream                                        2634    2692     +58
msg_and_die_if_script                                  -      21     +21
syntax_error_unexpected_ch                            41      46      +5
syntax_error_at                                       14      18      +4
die_if_script                                         31      28      -3
setup_redirects                                      319     308     -11
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 3/2 up/down: 88/-14)             Total: 74 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoline editing: make read_line_input() not take timeout param
Denys Vlasenko [Wed, 2 Aug 2017 15:27:28 +0000 (17:27 +0200)]
line editing: make read_line_input() not take timeout param

It's almost always -1.

function                                             old     new   delta
read_line_input                                     3902    3912     +10
new_line_input_t                                      24      31      +7
pgetc                                                583     585      +2
save_command_ps_at_cur_history                        80      78      -2
read_line                                             76      74      -2
fgetc_interactive                                    246     244      -2
addLines                                              84      82      -2
doCommands                                          2226    2222      -4
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 3/5 up/down: 19/-12)              Total: 7 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: make SIGINT handling visually less confusing
Denys Vlasenko [Wed, 2 Aug 2017 14:52:12 +0000 (16:52 +0200)]
hush: make SIGINT handling visually less confusing

$ echo $$
18448
$ echo $? <wait here, run "kill -INT 18448" in other shell><press enter>
   <=== NOTHING??
$

That empty line does not look right. After this patch:

$ echo $$
18448
$ echo $? <wait here, run "kill -INT 18448" in other shell><press enter>
^C
$

function                                             old     new   delta
fgetc_interactive                                    245     246      +1

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoNOFORK fixes
Denys Vlasenko [Wed, 2 Aug 2017 14:37:39 +0000 (16:37 +0200)]
NOFORK fixes

"rm -i FILE" and "yes" can now be interrupted by ^C in hush.
This also now works:

$ usleep 19999999
^C
$ echo $?
130

function                                             old     new   delta
run_pipe                                            1668    1711     +43
pseudo_exec_argv                                     312     321      +9
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 52/0)               Total: 52 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agodo not use `a' quoting style in comments
Denys Vlasenko [Wed, 2 Aug 2017 12:26:33 +0000 (14:26 +0200)]
do not use `a' quoting style in comments

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoconfig: do not use `a' quoting in help texts
Denys Vlasenko [Wed, 2 Aug 2017 12:12:48 +0000 (14:12 +0200)]
config: do not use `a' quoting in help texts

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: align --login code with dash
Denys Vlasenko [Mon, 31 Jul 2017 17:43:47 +0000 (19:43 +0200)]
ash: align --login code with dash

Upstream commit:

    Date: Sun, 13 Jul 2008 22:34:50 +0800
    [OPTIONS] Added support for -l

    This patch adds support for the -l option (login shell) as required
    by the LSB.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
It's a bit bigger, but gets rid of one global variable

function                                             old     new   delta
options                                              554     576     +22

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agotest: fix 4-argument case
Denys Vlasenko [Mon, 31 Jul 2017 17:20:43 +0000 (19:20 +0200)]
test: fix 4-argument case

Upstream dash commit:

    Date: Sun, 13 Jul 2008 19:20:10 +0800
    Fixed 3,4-argument cases for test per POSIX

    ----- Forwarded message from Gerrit Pape <pape@smarden.org> -----

    Subject: Bug#455828: dash: 4-argument test "test \( ! -e \)" yields an error
    Date: Fri, 28 Dec 2007 08:53:29 +0000
    From: Gerrit Pape <pape@smarden.org>
    To: Vincent Lefevre <vincent@vinc17.org>, 455828@bugs.debian.org

    On Thu, Dec 27, 2007 at 06:23:20PM +0100, Vincent Lefevre wrote:
    > On 2007-12-27 16:00:06 +0000, Gerrit Pape wrote:
    > > On Wed, Dec 12, 2007 at 02:18:47AM +0100, Vincent Lefevre wrote:
    > > > According to POSIX[*], "test \( ! -e \)" is a 4-argument test and is
    > > > here equivalent to "test ! -e". But dash (like ksh93 and bash) yields
    > > > an error:
    > > >
    > > > $ test \( ! -e \) || echo $?
    > > > test: 1: closing paren expected
    > > > 2
    > > > $ test ! -e || echo $?
    > > > 1
    > >
    > > Hi Vincent,
    > >
    > > the -e switch to test takes an argument, a pathname.
    >
    > According to POSIX, in both above examples, "-e" is *not* a switch,
    > just a string.
    >
    >   test \( ! -e \)
    >
    > means: return true if the string "-e" is empty, otherwhise return false.
    > The error in dash is that it incorrectly thinks that "-e" is a switch in
    > this context.

    I see, you're right.  Thanks, Gerrit.

    ----- End forwarded message -----

    This patch hard-codes the 3,4-argument cases in the way required by
    POSIX.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
function                                             old     new   delta
test_main                                            370     421     +51

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: remove redundant "G_flag_return_in_progress = -1"
Denys Vlasenko [Mon, 31 Jul 2017 16:06:07 +0000 (18:06 +0200)]
hush: remove redundant "G_flag_return_in_progress = -1"

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: fix "true | func_with_return" not allowing return.
Denys Vlasenko [Mon, 31 Jul 2017 16:02:28 +0000 (18:02 +0200)]
hush: fix "true | func_with_return" not allowing return.

function                                             old     new   delta
pseudo_exec_argv                                     305     312      +7

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: fix display of ">&-" redirect in job strings
Denys Vlasenko [Mon, 31 Jul 2017 12:32:18 +0000 (14:32 +0200)]
ash: fix display of ">&-" redirect in job strings

function                                             old     new   delta
cmdtxt                                               558     569     +11

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: functions have priority over builtins (!)
Denys Vlasenko [Mon, 31 Jul 2017 03:27:09 +0000 (05:27 +0200)]
hush: functions have priority over builtins (!)

function                                             old     new   delta
pseudo_exec_argv                                     291     305     +14
run_pipe                                            1560    1555      -5
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/1 up/down: 14/-5)               Total: 9 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: if STANDALONE, close interactive fd for NOEXECed children
Denys Vlasenko [Mon, 31 Jul 2017 02:54:53 +0000 (04:54 +0200)]
hush: if STANDALONE, close interactive fd for NOEXECed children

function                                             old     new   delta
pseudo_exec_argv                                     291     305     +14

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash,hush: ">&10" redirects to script/tty fds should not work
Denys Vlasenko [Mon, 31 Jul 2017 02:32:06 +0000 (04:32 +0200)]
ash,hush: ">&10" redirects to script/tty fds should not work

The fact that shell has open fds to tty and/or scripts should be
unobservable, if possible. In particular, if redirect tries to dup
one of them via ">&script_fd", it's better to pretend that script_fd
is closed, and thus redirect fails with EBADF.

Fixes these two testcase failures:
ash-redir/redir_to_bad_fd.tests
hush-redir/redir_to_bad_fd3.tests

function                                             old     new   delta
redirect                                            1018    1129    +111
setup_redirects                                      250     359    +109
readtoken1                                          2651    2655      +4
cmdloop                                              185     187      +2
changepath                                           194     195      +1
save_fd_on_redirect                                  203     194      -9
evaltree                                             501     484     -17
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 5/2 up/down: 227/-26)           Total: 201 bytes
   text    data     bss     dec     hex filename
 914553     485    6848  921886   e111e busybox_old
 914754     485    6848  922087   e11e7 busybox_unstripped

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: significant overhaul of redirect saving logic
Denys Vlasenko [Mon, 31 Jul 2017 02:09:19 +0000 (04:09 +0200)]
ash: significant overhaul of redirect saving logic

New code is similar to what hush is doing.
Make CLOSED to -1: same as dash.
popredir() loses "restore" parameter: same as dash.
COPYFD_RESTORE bit is no longer necessary.

This change fixes this interactive bug:

$ ls -l /proc/$$/fd 10>&-
ash: can't set tty process group: Bad file descriptor
ash: can't set tty process group: Bad file descriptor
[1]+  Done(2)                    ls -l /proc/${\$}/fd 10>&4294967295

function                                             old     new   delta
unwindredir                                           29      27      -2
tryexec                                              154     152      -2
evaltree                                             503     501      -2
evalcommand                                         1369    1367      -2
cmdloop                                              187     185      -2
redirect                                            1029    1018     -11
popredir                                             153     123     -30
need_to_remember                                      36       -     -36
is_hidden_fd                                          68       -     -68
------------------------------------------------------------------------------
(add/remove: 0/2 grow/shrink: 0/7 up/down: 0/-155)           Total: -155 bytes
   text    data     bss     dec     hex filename
 914572     485    6848  921905   e1131 busybox_old
 914553     485    6848  921886   e111e busybox_unstripped

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: massage redirect code to be slightly more like ash
Denys Vlasenko [Sun, 30 Jul 2017 21:34:04 +0000 (23:34 +0200)]
hush: massage redirect code to be slightly more like ash

function                                             old     new   delta
save_fd_on_redirect                                    -     203    +203
xdup_CLOEXEC_and_close                                 -      75     +75
setup_redirects                                      245     250      +5
xdup_and_close                                        72       -     -72
save_fds_on_redirect                                 221       -    -221
------------------------------------------------------------------------------
(add/remove: 2/2 grow/shrink: 1/0 up/down: 283/-293)          Total: -10 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: remove REDIR_SAVEFD2
Denys Vlasenko [Sun, 30 Jul 2017 14:51:05 +0000 (16:51 +0200)]
ash: remove REDIR_SAVEFD2

function                                             old     new   delta
evalcommand                                         1364    1369      +5
redirect                                            1055    1014     -41
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/1 up/down: 5/-41)             Total: -36 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: if !ENABLE_ASH_EXPAND_PRMT, disable PSSYNTAX code
Denys Vlasenko [Sat, 29 Jul 2017 20:58:44 +0000 (22:58 +0200)]
ash: if !ENABLE_ASH_EXPAND_PRMT, disable PSSYNTAX code

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [REDIR] Fix popredir on abnormal exit from built-in
Denys Vlasenko [Sat, 29 Jul 2017 20:51:52 +0000 (22:51 +0200)]
ash: [REDIR] Fix popredir on abnormal exit from built-in

Upstream commit:

    Date: Thu, 27 May 2010 15:03:46 +0800
    [REDIR] Fix popredir on abnormal exit from built-in

    Just like the poplocalvar problem recently fixed, redirections
    can also be leaked in case of an abnormal exit.  This patch fixes
    it using the same method as poplocalvar, by storing the previous
    redirection state and restoring to that point.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [PARSER] Add FAKEEOFMARK for expandstr
Denys Vlasenko [Sat, 29 Jul 2017 19:12:29 +0000 (21:12 +0200)]
ash: [PARSER] Add FAKEEOFMARK for expandstr

Upstream commit:

    Date: Thu, 27 Dec 2007 13:54:16 +1100
    [PARSER] Add FAKEEOFMARK for expandstr

    Previously expandstr used the string "" to indicate that it needs to be
    treated just like a here-doc except that there is no terminator.  However,
    the string "" is in fact a valid here-doc terminator so now that we deal
    with it correctly expandstr no longer works in the presence of new-lines
    in the prompt.

    This patch introduces the FAKEEOFMARK macro which does not equal any
    real EOF marker but is distinct from the NULL pointer which is used to
    indicate non-here-doc contexts.

    Thanks to Markus Triska for reporting this regression.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Unfortunately, I did not find the failing example for this old fix.

I also tweaked the code which was added by this commit:
"
Date: Mon Sep 24 18:30:02 2007 +0000
ash: fix prompt expansion (Natanael Copa <natanael.copa@gmail.com>)
"
since other parts of code do expect expandstr() to use DQSYNTAX, not PSSYNTAX.

function                                             old     new   delta
parse_stream                                        2609    2634     +25
setprompt_if                                         128     133      +5
read_profile                                          32      37      +5
evalcommand                                         1334    1339      +5
expandstr                                            122     120      -2
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 4/1 up/down: 40/-2)              Total: 38 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: fix handling of empty heredoc EOF marker
Denys Vlasenko [Sat, 29 Jul 2017 18:43:26 +0000 (20:43 +0200)]
hush: fix handling of empty heredoc EOF marker

function                                             old     new   delta
parse_stream                                        2609    2634     +25

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [PARSER] Removed noexpand/length check on eofmark
Denys Vlasenko [Sat, 29 Jul 2017 17:57:28 +0000 (19:57 +0200)]
ash: [PARSER] Removed noexpand/length check on eofmark

Upstream comment:

Date: Sun, 11 Nov 2007 14:21:23 +0800
[PARSER] Removed noexpand/length check on eofmark

On Tue, Oct 30, 2007 at 04:23:35AM +0000, Oleg Verych wrote:
>
> } 8<<""
> ======================

Actually this (the empty delim) only works with dash by accident.
I've tried bash and pdksh and they both terminate on the first
empty line which is what you would expect rather than EOF.  The
real Korn shell does something completely different.

I've fixed this in dash to conform to bash/pdksh.

> In [0] it's stated, that delimiter isn't evaluated (expanded), only
> quoiting must be checked. That if() seems to be completely bogus.

OK I agree.  The reason it was there is because the parser would
have already replaced the dollar sign by an internal representation.

I've fixed it properly with this patch.

Test case:

        cat <<- $a
                OK
        $a

        cat <<- ""
                OK

        echo OK

Old result:

        dash: Syntax error: Illegal eof marker for << redirection
        OK

        echo OK

New result:

        OK
        OK
        OK

function                                             old     new   delta
parsefname                                           227     152     -75
readtoken1                                          2819    2651    -168
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-243)           Total: -243 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: "Undo all redirections" comment is wrong, delete it
Denys Vlasenko [Sat, 29 Jul 2017 16:54:53 +0000 (18:54 +0200)]
ash: "Undo all redirections" comment is wrong, delete it

No code changes.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agolibbb: remove vdprintf
Ron Yorston [Sat, 29 Jul 2017 13:48:00 +0000 (14:48 +0100)]
libbb: remove vdprintf

Two reasons:

* it's only built for GNU C Library versions < 2 (released 1997)

* it isn't used anywhere in BusyBox

Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agops: tweak outdated comment, no code changes
Denys Vlasenko [Sat, 29 Jul 2017 00:27:12 +0000 (02:27 +0200)]
ps: tweak outdated comment, no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agops: implement -o sid
Denys Vlasenko [Sat, 29 Jul 2017 00:20:17 +0000 (02:20 +0200)]
ps: implement -o sid

function                                             old     new   delta
func_sid                                               -      29     +29
out_spec                                             320     340     +20
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 1/0 up/down: 49/0)               Total: 49 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agostandalone shell / prefer_applets: fix "exe" in comm fields
Denys Vlasenko [Sat, 29 Jul 2017 00:19:01 +0000 (02:19 +0200)]
standalone shell / prefer_applets: fix "exe" in comm fields

function                                             old     new   delta
main                                                  92     106     +14

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: make tryexec(cmd) parameter const char
Denys Vlasenko [Fri, 28 Jul 2017 23:20:53 +0000 (01:20 +0200)]
ash: make tryexec(cmd) parameter const char

Fewer casts this way.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoshell: make standalone shell tab-complete "busybox"
Denys Vlasenko [Fri, 28 Jul 2017 22:59:24 +0000 (00:59 +0200)]
shell: make standalone shell tab-complete "busybox"

function                                             old     new   delta
busybox_main                                           -     624    +624
packed_usage                                       31758   31777     +19
applet_names                                        2638    2646      +8
applet_main                                         1528    1532      +4
applet_install_loc                                   191     192      +1
run_applet_and_exit                                  681      78    -603
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 4/1 up/down: 656/-603)           Total: 53 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: support platforms that don't have '%m' printf specifier
Ron Yorston [Thu, 27 Jul 2017 12:53:39 +0000 (13:53 +0100)]
ash: support platforms that don't have '%m' printf specifier

The '%m' conversion specifier prints an error message based on the
current value of 'errno'.  It is available in the GNU C library,
Cygwin (since 2012), uClibc and musl.

It is not available in various BSDs, BSD-derived systems (MacOS,
Android) or Microsoft Windows.

Use a symbol defined in platform.h to control how error messages
can be formatted to display the 'errno' message.  On platforms that
support it use '%m'; on other platforms use '%s' and strerror().

On platforms that have '%m' there is essentially no change in the
size of the binary.  Otherwise:

function                                             old     new   delta
redirect                                            1287    1310     +23
xtcsetpgrp                                            27      44     +17
dup2_or_raise                                         34      51     +17
setinputfile                                         267     275      +8
.rodata                                           163379  163371      -8
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 4/1 up/down: 65/-8)              Total: 57 bytes

Signed-off-by: Ron Yorston <rmy@pobox.com>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: less hackish implementation of evaltreenr()
Denys Vlasenko [Fri, 28 Jul 2017 13:28:33 +0000 (15:28 +0200)]
ash: less hackish implementation of evaltreenr()

Defining a function alias with __attribute__ ((alias("evaltree"),__noreturn__))
is not that usual, and clang had a bug which made it misunderstand
this construct.

Switch to:
ALWAYS_INLINE NORETURN evaltreenr() { evaltree(); unreachable(); }

Older gcc's do not know unreachable(), on them we pay the price of having
a few extra calls to abort():

function                                             old     new   delta
evalsubshell                                         151     156      +5
evalpipe                                             357     362      +5
argstr                                              1141    1144      +3

On newer gcc, code size does not change.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agounzip: fix regression on big-endian machines
Natanael Copa [Tue, 25 Jul 2017 18:44:50 +0000 (20:44 +0200)]
unzip: fix regression on big-endian machines

This fixes a regression which was introduced with commit 2a0867a5
("unzip: optional support for bzip2 and lzma") and causes unzip to exit
with error when extracting archives:

  unzip: unsupported method 2048

Signed-off-by: Natanael Copa <ncopa@alpinelinux.org>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoconfig: more tweaking of help texts
Denys Vlasenko [Thu, 27 Jul 2017 12:45:25 +0000 (14:45 +0200)]
config: more tweaking of help texts

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agordate: tweak comments, no code changes
Denys Vlasenko [Thu, 27 Jul 2017 12:31:59 +0000 (14:31 +0200)]
rdate: tweak comments, no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoinetd: improve --helpt text and config help text.
Denys Vlasenko [Thu, 27 Jul 2017 11:34:51 +0000 (13:34 +0200)]
inetd: improve --helpt text and config help text.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agotweak examples/var_service/*
Denys Vlasenko [Thu, 27 Jul 2017 10:53:20 +0000 (12:53 +0200)]
tweak examples/var_service/*

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoed: fix "\n" removal in command line; make "w" set "dirty = 0"
Denys Vlasenko [Thu, 27 Jul 2017 10:34:56 +0000 (12:34 +0200)]
ed: fix "\n" removal in command line; make "w" set "dirty = 0"

function                                             old     new   delta
doCommands                                          2184    2226     +42
getNum                                               345     343      -2
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/1 up/down: 42/-2)              Total: 40 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoed: code shrink
Denys Vlasenko [Thu, 27 Jul 2017 09:58:25 +0000 (11:58 +0200)]
ed: code shrink

function                                             old     new   delta
findString                                           117     115      -2
skip_blank                                            16       -     -16
getNum                                               369     345     -24
doCommands                                          2448    2183    -265
------------------------------------------------------------------------------
(add/remove: 0/1 grow/shrink: 0/3 up/down: 0/-307)           Total: -307 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoed: fix --help and reorder functions, no code changes
Denys Vlasenko [Thu, 27 Jul 2017 09:17:15 +0000 (11:17 +0200)]
ed: fix --help and reorder functions, no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoswapon: do not use FEATURE_MOUNT_LABEL, have your own FEATURE_SWAPONOFF_LABEL
Denys Vlasenko [Thu, 27 Jul 2017 08:58:08 +0000 (10:58 +0200)]
swapon: do not use FEATURE_MOUNT_LABEL, have your own FEATURE_SWAPONOFF_LABEL

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoconfig: trim/improve item names and help texts, take 2
Denys Vlasenko [Thu, 27 Jul 2017 08:53:09 +0000 (10:53 +0200)]
config: trim/improve item names and help texts, take 2

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoconfig: trim/improve item names and help texts.
Denys Vlasenko [Thu, 27 Jul 2017 00:59:13 +0000 (02:59 +0200)]
config: trim/improve item names and help texts.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: force inlining of a trivial function
Denys Vlasenko [Wed, 26 Jul 2017 21:08:36 +0000 (23:08 +0200)]
ash: force inlining of a trivial function

function                                             old     new   delta
bltinlookup                                            5       -      -5

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: revert previous implementation of "A=1 A=2 B=$A cmd" code
Denys Vlasenko [Wed, 26 Jul 2017 21:03:21 +0000 (23:03 +0200)]
ash: revert previous implementation of "A=1 A=2 B=$A cmd" code

Reverts this:

    commit 0e6f661e23d358cca104c24f8438d0ec64df32f1
    Date:   Fri Feb 15 15:02:15 2008 +0000
    ash: handle "A=1 A=2 B=$A; echo $B". closes bug 947.

A different fix from upstream has been imported by previous six commits.

Last seven commits, cumulative:

function                                             old     new   delta
poplocalvars                                           -     314    +314
mklocal                                                -     288    +288
pushlocalvars                                          -      48     +48
evalcommand                                         1372    1408     +36
unwindlocalvars                                        -      22     +22
ash_main                                            1022    1029      +7
setvar                                               167     172      +5
localvar_stack                                         -       4      +4
setvareq                                             303     302      -1
evalcase                                             271     269      -2
subevalvar                                          1202    1198      -4
localvars                                              4       -      -4
cmdenviron                                             4       -      -4
expandarg                                            984     973     -11
evalvar                                              589     574     -15
argstr                                              1164    1141     -23
dotcmd                                               335     303     -32
bltinlookup                                           51       5     -46
varvalue                                             709     596    -113
evalfun                                              456     270    -186
localcmd                                             364      44    -320
------------------------------------------------------------------------------
(add/remove: 5/2 grow/shrink: 3/11 up/down: 724/-761)         Total: -37 bytes
   text    data     bss     dec     hex filename
 915353     485    6888  922726   e1466 busybox_old
 915320     485    6880  922685   e143d busybox_unstripped

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [VAR] Fix loss of variables when hash collides
Denys Vlasenko [Wed, 26 Jul 2017 18:06:48 +0000 (20:06 +0200)]
ash: [VAR] Fix loss of variables when hash collides

Upstream commit:

    Date: Tue, 6 Jul 2010 17:40:53 +0800
    [VAR] Fix loss of variables when hash collides

    Brian Koropoff reported that the new var patches broke the following
    script:

    #!/bin/dash
    GDM_LANG="bar"
    OPTION="foo"
    unset GDM_LANG
    # OPTION has mysteriously become unset
    echo "$OPTION"

    He correctly diagnosed this as a result of removing all variables
    in the hash chain preceding the one that should be removed in
    setvareq.

    He also provided a patch to fix this.

    This patch is based on his but without keeping the original vpp.
    As a result, we now store new variables at the end of the hash
    chain instead of the beginning.

    To make this work, setvareq/setvar now returns the vp pointer
    modified.  In case they're used to unset a variable the pointer
    returned is undefined.  This is because mklocal needs it and
    used to get it by assuming that the new variable always appear
    at the beginning of the chain.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [VAR] Do not poplocalvars prematurely on regular utilities
Denys Vlasenko [Wed, 26 Jul 2017 18:00:08 +0000 (20:00 +0200)]
ash: [VAR] Do not poplocalvars prematurely on regular utilities

Upstream commit:

    Date: Thu, 27 May 2010 11:50:19 +0800
    [VAR] Do not poplocalvars prematurely on regular utilities

    The recent cmdenviron removal broke regular utilities by calling
    poplocalvars too early.  This patch fixes that by postponing the
    poplocalvars for regular utilities until they have completed.

    In order to ensure that local still works, it is now a special
    built-in.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [VAR] Fix poplocalvar on abnormal exit from function
Denys Vlasenko [Wed, 26 Jul 2017 17:55:31 +0000 (19:55 +0200)]
ash: [VAR] Fix poplocalvar on abnormal exit from function

Upstream commit:

    Date: Thu, 27 May 2010 11:32:55 +0800
    [VAR] Fix poplocalvar on abnormal exit from function

    The new localvar code broke the abnormal exit from functions
    and built-ins by not restoring the original localvar state.

    This patch fixes this by storing the previous localvar state so
    that we always unwind correctly in case of an abnormal exit.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [VAR] Replace cmdenviron with localvars
Denys Vlasenko [Wed, 26 Jul 2017 17:53:11 +0000 (19:53 +0200)]
ash: [VAR] Replace cmdenviron with localvars

Upstream commit:

    Date: Wed, 26 May 2010 18:54:19 +0800
    [VAR] Replace cmdenviron with localvars

    This patch replaces the cmdenviron mechanism for temporary command
    variables with the localvars mechanism used by functions.

    This reduces code size, and more importantly, makes the variable
    assignment take effect immediately as required by POSIX.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [VAR] Fix poplocalvar leak
Denys Vlasenko [Wed, 26 Jul 2017 17:25:40 +0000 (19:25 +0200)]
ash: [VAR] Fix poplocalvar leak

Upstream commit:

    Date: Tue, 25 May 2010 18:14:32 +0800
    [VAR] Fix poplocalvar leak

    When a variable is marked as local, we set VSTRFIXED on its vp
    recored.  However, poplocalvar never clears this flag for variables
    that were unset to begin with.  Thus if you ever made an unset
    variable local, it would get the VSTRFIXED bit and stick around
    forever.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [VAR] Add localvars nesting
Denys Vlasenko [Wed, 26 Jul 2017 17:22:34 +0000 (19:22 +0200)]
ash: [VAR] Add localvars nesting

Upstream commit:

    Date: Mon, 24 May 2010 15:31:27 +0800
    [VAR] Add localvars nesting

    This patch adds localvars nesting infrastructure so we can reuse
    the localvars mechanism for command evaluation.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: alloc slightly smaller buffer in cvtnum(); faster unsetvar()
Denys Vlasenko [Wed, 26 Jul 2017 12:38:19 +0000 (14:38 +0200)]
ash: alloc slightly smaller buffer in cvtnum(); faster unsetvar()

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: fix redir_leak.tests if STANDALONE=y
Denys Vlasenko [Wed, 26 Jul 2017 11:42:53 +0000 (13:42 +0200)]
ash: fix redir_leak.tests if STANDALONE=y

If STANDALONE and we run a NOEXEC applet, saved copies of redirected fds
were visible for the child. They have CLOEXEC bit, yes, but we do not exec
in this case.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoshell: remove ${#+} tests, it is not a valid construct
Denys Vlasenko [Tue, 25 Jul 2017 22:30:02 +0000 (00:30 +0200)]
shell: remove ${#+} tests, it is not a valid construct

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: fix ${##}, ${#?}, ${#!} handling
Denys Vlasenko [Tue, 25 Jul 2017 22:07:27 +0000 (00:07 +0200)]
hush: fix ${##}, ${#?}, ${#!} handling

function                                             old     new   delta
parse_dollar                                         786     820     +34
expand_one_var                                      1579    1592     +13
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 47/0)               Total: 47 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [SHELL] Optimize dash -c "command" to avoid a fork
Denys Vlasenko [Tue, 25 Jul 2017 18:31:14 +0000 (20:31 +0200)]
ash: [SHELL] Optimize dash -c "command" to avoid a fork

Upstream commit:

    Date: Thu, 7 Jul 2011 13:58:48 +0800
    [SHELL] Optimize dash -c "command" to avoid a fork

    On Sun, Apr 10, 2011 at 07:36:49AM +0000, Jonathan Nieder wrote:
    > From: Jilles Tjoelker <jilles@stack.nl>
    > Date: Sat, 13 Jun 2009 16:17:45 -0500
    >
    > This change only affects strings passed to -c, when the -s option is
    > not used.
    >
    > Use the EV_EXIT flag to inform the eval machinery that the string
    > being passed is the entirety of input.  This way, a fork may be
    > omitted in many special cases.
    >
    > If there are empty lines after the last command, the evalcmd will not
    > see the end early enough and forks will not be omitted. The same thing
    > seems to happen in bash.
    >
    > Example:
    >   sh -c 'ps lT'
    > No longer shows a shell process waiting for ps to finish.
    >
    > [jn: ported from FreeBSD SVN r194128.  Bugs are mine.]
    >
    > Signed-off-by: Jonathan Nieder <jrnieder@gmail.com>

    Instead of detecting EOF using the input layer, I'm going to
    use the parser instead.  In either case, we always have to read
    ahead in order to complete the parsing of the previous node.
    Therefore we always know whether there is more to come, except
    in the case where we see a newline/semicolon or similar.

    For the purposes of sh -c, this should be sufficient.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
function                                             old     new   delta
evalstring                                           190     224     +34
ash_main                                            1014    1022      +8
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/0 up/down: 42/0)               Total: 42 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: allow "trap NUM [SIG]..." syntax
Denys Vlasenko [Tue, 25 Jul 2017 18:06:17 +0000 (20:06 +0200)]
ash: allow "trap NUM [SIG]..." syntax

While at it, make get_signum() return -1 for numeric strings >= NSIG.

function                                             old     new   delta
trapcmd                                              292     306     +14
get_signum                                           295     300      +5
builtin_trap                                         413     412      -1
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/1 up/down: 19/-1)              Total: 18 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [REDIR] Replace GPL noclobberopen code with the FreeBSD version
Denys Vlasenko [Tue, 25 Jul 2017 15:47:48 +0000 (17:47 +0200)]
ash: [REDIR] Replace GPL noclobberopen code with the FreeBSD version

Upstream commit:

    Date: Thu, 10 Mar 2011 16:52:13 +0800
    [REDIR] Replace GPL noclobberopen code with the FreeBSD version

    Replace noclobberopen() from bash with the FreeBSD code for noclobber
    opens.

    This also reduces code size by eliminating an unnecessary check.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
function                                             old     new   delta
changepath                                           192     194      +2
localcmd                                             366     364      -2
expmeta                                              521     517      -4
redirect                                            1210    1135     -75
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 1/3 up/down: 2/-81)             Total: -79 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: sync up with dash with respect to redirection escaping
Denys Vlasenko [Tue, 25 Jul 2017 15:37:57 +0000 (17:37 +0200)]
ash: sync up with dash with respect to redirection escaping

We fixed the problem differently than they. Let's not deviate.

Upstream commit:

    Date: Thu, 27 May 2010 20:07:29 +1000
    [EXPAND] Fix corruption of redirections with byte 0x81

    In other ash variants, a partial implementation of ksh-like cmd >file*
    adds and removes CTLESC bytes ('\x81') in redirection filenames,
    preserving 8-bit transparency. Long ago, dash removed the code to add
    the CTLESC bytes, but not the code to remove them, causing corruption of
    filenames containing CTLESC. This commit removes the code to remove the
    CTLESC bytes.

    The CTLESC byte occurs frequently in UTF-8 encoded non-Latin text.

    This bug has been reported various times to Ubuntu and Debian (e.g.
    Launchpad Ubuntu #422298). This patch is the same as the one submitted
    by Alexander Korolkov in Ubuntu #422298.

Signed-off-by: Jilles Tjoelker <jilles@stack.nl>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
function                                             old     new   delta
changepath                                           194     192      -2
expandarg                                           1000     984     -16
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-18)             Total: -18 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: [VAR] Move unsetvar functionality into setvareq
Denys Vlasenko [Tue, 25 Jul 2017 14:29:36 +0000 (16:29 +0200)]
ash: [VAR] Move unsetvar functionality into setvareq

Upstream commit:

    Date: Tue, 25 May 2010 20:55:05 +0800
    [VAR] Move unsetvar functionality into setvareq

    This patch moves the unsetvar code into setvareq so that we can
    no have a pathological case of an unset variable hanging around
    unless it has a bit pinning it like VEXPORT.

Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
function                                             old     new   delta
setvareq                                             227     303     +76
expmeta                                              517     521      +4
localcmd                                             364     366      +2
unsetcmd                                              96      76     -20
unsetvar                                             129       7    -122
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 3/2 up/down: 82/-142)           Total: -60 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: import param_expand_default.tests from hush
Denys Vlasenko [Tue, 25 Jul 2017 13:25:07 +0000 (15:25 +0200)]
ash: import param_expand_default.tests from hush

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: treat ${#?} as "length of $?"
Denys Vlasenko [Tue, 25 Jul 2017 13:18:57 +0000 (15:18 +0200)]
hush: treat ${#?} as "length of $?"

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: add a few tests from hush-vars/*
Denys Vlasenko [Tue, 25 Jul 2017 12:55:05 +0000 (14:55 +0200)]
ash: add a few tests from hush-vars/*

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agovolume_id: enable minix detection
Denys Vlasenko [Tue, 25 Jul 2017 12:22:08 +0000 (14:22 +0200)]
volume_id: enable minix detection

function                                             old     new   delta
volume_id_probe_minix                                  -      87     +87
fs2                                                   64      68      +4
------------------------------------------------------------------------------
(add/remove: 2/0 grow/shrink: 1/0 up/down: 91/0)               Total: 91 bytes

Patch by wdlkmpx <wdlkmpx@gmail.com>

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: fix comment, no code changes
Denys Vlasenko [Mon, 24 Jul 2017 18:03:24 +0000 (20:03 +0200)]
ash: fix comment, no code changes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: copy three tests from hush_test/hush-signals/*
Denys Vlasenko [Mon, 24 Jul 2017 17:42:46 +0000 (19:42 +0200)]
ash: copy three tests from hush_test/hush-signals/*

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoshell: some additions to *sh-misc/* tests
Denys Vlasenko [Mon, 24 Jul 2017 16:51:40 +0000 (18:51 +0200)]
shell: some additions to *sh-misc/* tests

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agotar: postpone creation of symlinks with "suspicious" targets. Closes 8411
Denys Vlasenko [Mon, 24 Jul 2017 15:20:13 +0000 (17:20 +0200)]
tar: postpone creation of symlinks with "suspicious" targets. Closes 8411

function                                             old     new   delta
data_extract_all                                     968    1038     +70
tar_main                                             952     986     +34
scan_tree                                            258     262      +4
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 3/0 up/down: 108/0)             Total: 108 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoudhcpc: downgrade "MAC X:X:X:X:X:X" message to log2 level
Denys Vlasenko [Mon, 24 Jul 2017 11:37:38 +0000 (13:37 +0200)]
udhcpc: downgrade "MAC X:X:X:X:X:X" message to log2 level

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: add most of hush process subst tests
Denys Vlasenko [Mon, 24 Jul 2017 10:49:49 +0000 (12:49 +0200)]
ash: add most of hush process subst tests

ash passes these.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: fix a case when redirect to a closed fd #1 is not restoring (closing) it
Denys Vlasenko [Mon, 24 Jul 2017 10:42:17 +0000 (12:42 +0200)]
hush: fix a case when redirect to a closed fd #1 is not restoring (closing) it

function                                             old     new   delta
setup_redirects                                      200     245     +45
append_squirrel                                        -      41     +41
save_fds_on_redirect                                 256     221     -35
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 1/1 up/down: 86/-35)             Total: 51 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoudhcpc[6]: downgrade "adapter index N" messages to log2 level
Denys Vlasenko [Mon, 24 Jul 2017 10:06:53 +0000 (12:06 +0200)]
udhcpc[6]: downgrade "adapter index N" messages to log2 level

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoudhcpc[6]: fix messages referring to select() while we use poll()
Denys Vlasenko [Mon, 24 Jul 2017 10:01:28 +0000 (12:01 +0200)]
udhcpc[6]: fix messages referring to select() while we use poll()

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: add all hush parsing tests to ast tests
Denys Vlasenko [Mon, 24 Jul 2017 00:49:56 +0000 (02:49 +0200)]
ash: add all hush parsing tests to ast tests

All pass.

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agohush: use mempcpy where useful
Denys Vlasenko [Mon, 24 Jul 2017 00:17:05 +0000 (02:17 +0200)]
hush: use mempcpy where useful

function                                             old     new   delta
o_addblock                                            58      42     -16
expand_one_var                                      1618    1555     -63
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 0/2 up/down: 0/-79)             Total: -79 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: fix for last commit
Denys Vlasenko [Sun, 23 Jul 2017 19:55:40 +0000 (21:55 +0200)]
ash: fix for last commit

"mempcpy(q, s, len) + len" is obviously no good :(

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: use mempcpy() in more places
Denys Vlasenko [Sun, 23 Jul 2017 19:46:02 +0000 (21:46 +0200)]
ash: use mempcpy() in more places

Most changes are taken from dash.

function                                             old     new   delta
single_quote                                         127     129      +2
stack_nputstr                                         28      29      +1
path_advance                                         209     202      -7
rmescapes                                            346     308     -38
------------------------------------------------------------------------------
(add/remove: 0/0 grow/shrink: 2/2 up/down: 3/-45)             Total: -42 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agozcip: fix slow environment leak
Denys Vlasenko [Sat, 22 Jul 2017 01:04:20 +0000 (03:04 +0200)]
zcip: fix slow environment leak

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agocrond: do not assume setenv() does not leak
Denys Vlasenko [Sat, 22 Jul 2017 00:25:47 +0000 (02:25 +0200)]
crond: do not assume setenv() does not leak

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agolibbb: avoid malloc/free in bb_unsetenv()
Denys Vlasenko [Sat, 22 Jul 2017 00:15:17 +0000 (02:15 +0200)]
libbb: avoid malloc/free in bb_unsetenv()

function                                             old     new   delta
bb_unsetenv                                           55      83     +28

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agocrond: allow $SHELL and starting user's shell override DEFAULT_SHELL. Closes 6458
Denys Vlasenko [Fri, 21 Jul 2017 23:47:19 +0000 (01:47 +0200)]
crond: allow $SHELL and starting user's shell override DEFAULT_SHELL. Closes 6458

function                                             old     new   delta
crond_main                                          1134    1149     +15
fork_job                                             453     454      +1

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: use mempcpy() where appropriate
Denys Vlasenko [Fri, 21 Jul 2017 16:51:29 +0000 (18:51 +0200)]
ash: use mempcpy() where appropriate

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agobasename: do not use argc
Denys Vlasenko [Fri, 21 Jul 2017 16:41:46 +0000 (18:41 +0200)]
basename: do not use argc

function                                             old     new   delta
basename_main                                        153     148      -5

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agopgrep: fix "pgrep -v -P1"; also allow matching of comm - closes 10086
Denys Vlasenko [Fri, 21 Jul 2017 15:43:14 +0000 (17:43 +0200)]
pgrep: fix "pgrep -v -P1"; also allow matching of comm - closes 10086

function                                             old     new   delta
pgrep_main                                           662     720     +58

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoping: fix help text to show what parameter -p takes
Denys Vlasenko [Fri, 21 Jul 2017 14:07:13 +0000 (16:07 +0200)]
ping: fix help text to show what parameter -p takes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
7 years agoash: suppress readonly1.tests false positive; add readonly0.tests
Denys Vlasenko [Fri, 21 Jul 2017 11:37:25 +0000 (13:37 +0200)]
ash: suppress readonly1.tests false positive; add readonly0.tests

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