avoid MAIL* environment variables to save a few bytes in install media

ksh(1) MAIL, MAILCHECK, MAILPATH mbox handling is useless in the installer.

OK miod deraadt

Omit version in SMALL ksh builds

No need for KSH_VERSION and its PS1 esacape sequences in installer shells.
Save some bits and clean up what(1) output on ramdisk kernels.

OK deraadt

When system calls indicate an error they return -1, not some arbitrary
value < 0. errno is only updated in this case. Change all (most?)
callers of syscalls to follow this better, and let's see if this strictness
helps us in the future.

When evaluating an arithmetical expression, for example inside $(()),
never do substitution (neither parameter, nor command, nor arithmetic,
nor tilde substitution) on the values of any variables encountered
inside the expression, but do recursively perform arithmetical
evaluation of subexpressions as required. This makes behaviour
more consistent, without hindering any behaviour promised in the
manual page.

A quirk originally reported by Andy Chu <andychup at gmail dot com>
was that in the past, when encountering an array index, the shell
would not only do evaluation, but also substitution on the array
index, even though substitution would not be done on the expression
in general.

tobias@ contributed to initial efforts of understanding the quirk.
patch tested in a bulk build by naddy@
"please commit" deraadt@

Fix the case where the recursion detection isn't reset when the command is
interrupted.

Lots of back and forth with anton@
OK jca@, tb@, anton@

Use a very regular call pattern to pledge, so that we can continue to
grep and compare the use in all programs..

Include "id" in pledge (for setres[ug]id, setgroups) if the shell is
privileged and remove it when dropping privileges (set +p), setting a
flag to make sure we don't do it again.

ok deraadt millert

Export the PWD and OLDPWD shell variables as per POSIX.
Previously, these would only be exported if they were present
in the environment when the shell started. OK deraadt@ anton@ kn@

remove the alias type='whence -v' and replace it with
a buildin command, that just calls into c_whence().
This makes type look like the buildin in other shells and makes
things like system("'type' 'git'"); work.

With lots of suggestions and feedback from anton@, kn@ and jca@.
ok kn@ jca@

Support 64 bit integers on 32 bit architectures.

No binary change on amd64 and there should be no differences on any
other 64 bit architecture either (because long = int64_t).

ok cheloha, tb

Favor usage of __func__ in warning/error messages. Some of them referred to the
wrong function and fix the rest for consistency.
Diff from Michael W. Bombardieri with some cosmetic cleanup applied.

ok benno@ tb@

Introduce internal_warningf() and mark internal_errorf() as noreturn

This helps tools like scan-build, and follows the example of warningf()
and errorf(). ok anton@

Stop pretending we support building ksh without EDIT/HISTORY support

ok anton@

unifdef BRACE_EXPAND; ok jca@

unifdef JOBS support

Prompted by a mail from Klemens Nanni, who also had the same diff.
ok deraadt@ millert@

Include hostname in shell prompts by default

With tmux, ssh and vmd, we tend to open shells on many different hosts
simultaneously and the default prompts '$ ' and '# ' for {,k}sh as well
as '% ' and '# ' for csh become dangerous: it's very easy to issue a
command on the wrong host.

This can easily be avoided by displaying the hostname in the prompt.
Everything beyond "hostname{$,#,%} " is going to be a matter of taste,
so we left it at that. If you use an FQDN, only the first part (the
output of 'hostname -s') will be printed.

requested by and ok deraadt; mostly positive feedback many
ok anton, brynet, bcallah and others

Drop the unused second argument from set_prompt(). It used to be used for
early special casing of ! and !! in the PS1 expansion. This was removed
from set_prompt() as part of the implementaion of the character count
toggles \[ and \] back in 2004.

ok jca

Retire old behavior of requiring root prompt to contain # or \!

Requested by akoshibe and phessler
ok phessler@ anton@ jca@

No need to inspect LC_CTYPE:
* It was only used in vi input mode.
* No functional change in case of set -o vi-show8.
* No functional change if LC_CTYPE is set to UTF-8.
* More robust for the default of LC_CTYPE=C on a UTF-8 xterm.

Behaviour changes for the combination LC_CTYPE=C with set +o vi-show8
on non-UTF-8 terminals, but that combination is useless and dangerous
with or without this patch. If you want to edit individual raw non-ASCII
non-UTF-8 bytes on the shell command line, always use set -o vi-show8.
(Besides, i doubt that you actually want to do that; better use a real
hex editor in the first place.)

OK czarkoff@.

Partial UTF-8 line editing support for ksh(1) vi input mode;
so far, it covers these commands: a h i l x /
This is an isu8cont()-based hack similar in style to what i did
in emacs input mode, but less elegant and slightly more intrusive
because the vi mode code is much more ugly and less straightforward
than the emacs mode code. This one required partial rewrites of
a few helper functions, and comments were added while there.

This is not perfect, but hopefully reduces people's cursing
until a more rigorous solution can be devised (much) later.
Some polishing may be useful in tree, in particular adding
utf8cont() support to a few missing commands.

Mostly written shortly after Christmas 2015.
Reminded by and OK czarkoff@.
Feedback, partial review and testing, no longer any objection by martijn@.
Feedback and testing by tb@.
Also read fine to nicm@.

Mark PPID read-only as per the manual. From Anthony Coulter.

remove CSRG BSD-licensed mknod builtin code which was previously used to
accelerate install/upgrade time MAKEDEV runs. That allows ramdisk ksh
to be pledged, without needing "dpath". We'll solve the performance issues
a different way (soon).
ok otto espie natano tb tobiasu rpe

rename global "e" to genv to avoid accidental shadowing and aliasing.
ok millert nicm tb

Tweak comment: we now always make a copy of argv.

ok halex@

unconditionally duplicate the argv array on initialization, to make it
resilient against being altered by a subsequent shift operation

tweak and ok semarie@

Move system headers from sh.h to those files that actually need them.

ok mmcc@ a while ago

exit() after perror() for pledge failure. Perhaps this got introduced
as a test idiom, either when pledge was young or during the transition
to strings.... dunno

Add uses of areallocarray(). mksh and Bitrig ksh already have similar
functions. With help from Theo Buehler.

ok nicm@

'fc -e -' is deprecated in favor of 'fc -s', so update the built-in alias

ok millert@

Final removal of EXTERN.

ok nicm@

Penultimate commit to remove EXTERN.

ok nicm@

Move string.h include from sh.h to the files that use it.

ok nicm@

More removal of EXTERN.

ok nicm@

Apply style(9) to header includes.

ok nicm@

Move more EXTERN-defined globals from sh.h.

ok nicm@

Move a system header include from the global header (sh.h) into the
files that need it. No binary change.

"This looks fine" -nicm@

wrap a long line

Implement real "flock" request and add it to userland programs that
use pledge and file locking. OK deraadt@

normalize a few more tame request orderings, to help review

ksh can run with pledge "stdio rpath wpath cpath getpw fattr proc exec tty"
if the mknod builtin is disabled. It looks like a lot of abilities, but
hey, this is a shell. can't open sockets or do other nasty stuff though.
(we'll leave the mknod builtin enabled on the install media for now; there
is work happening to regain the MAKEDEV performance in a different way)
discussions with otto & millert in particular

Last of the (thing *)0 -> NULL, from Michael McConville. No binary
change.

Remove unnecessary casts, from Michael McConville. No binary change.

correct spelling of NULL from (char *)0. from Michael McConville.
ok md5

Replace newline and space defines by "\n" and " " directly, from Michael
McConville. ok millert

no need to check for null argv anymore. from Martijn van Duren
ok deraadt

If we hit multiple errors while unwinding we'll end up running code that
assumes a pointer is valid when it has been free'd.

This is convoluted as ksh has it's own allocator and uses long jumps.
Set the pointer to NULL after the quitenv() call in unwind() in case we
later hit a long jump in unwind().

Found with afl a while back.

ok tedu@ deraadt@

remove trailing whitespaces; use tabs instead of spaces where appropriate;
no binary changes.

Add a proper suspend builtin that saves/restores the tty and pgrp
as needed instead of an alias that just sends SIGSTOP. Login shells
may be suspended if they are not running in an orphan process group.
OK guenther@ jmc@

Run any pending traps before calling the EXIT or ERR traps when -e
is set. Fixes a bug where we would not run the signal trap if,
for example, ^C was pressed and -e was set. OK espie@

fake a sigwinch after each job, so if the terminal changes size,
we'll notice and update

Avoid modifying argv when building argv for $* and $@ since it will
affect ps output. This can happen when command line options are
specified, e.g. "sh -c command". Based on a diff from espie@
OK espie@

backout, Todd's version may be cleaner, but it's also wrong.

sh -c should not munge argv[].
This fixes ps -ww output.

joint work by millert@ and me@.

okay otto@, deraadt@ "feel free to commit my version" millert@

beter determination if the shell is restricted; from Alf Schlichting.
ok jasper@ millert@

properly keep track of the line number after a trap.
this fixes pr 6195. diff from manuel giraud, thanks.

ok kili@

pass "xerrok" status across the execution call stack to more closely
match what both POSIX and ksh.1 already describe in regards to set
-e/errexit's behavior in determining when to exit from nonzero return
values.

specifically, the truth values tested as operands to `&&' and `||', as
well as the resulting compound expression itself, along with the truth
value resulting from a negated command (i.e. a pipeline prefixed `!'),
should not make the shell exit when -e is in effect.

issue reported by matthieu.
testing matthieu, naddy.
ok miod (earlier version), otto.
man page ok jmc.

ksh starting as root will refuse to import a PS1 prompt unless it
contains a '#' character. Make it also accept the \$ PS1 sequence
(which is supposed to automagically be $ or # depending on the user
running the shell) unless ksh has been started in sh-compatibility
mode.

pointed out by millert@ when I asked why my SUDO_PS1 was not being
honoured; ok millert@

Only process ENV when in interactive mode as SU says. Started by
tedu to fix make -j, SU thing spotted by naddy@; ok millert@ @tedu
jmc@ for the man page bit.

Only check if last three letters are "/sh" if string is at least
three characters long.

OK moritz@.

fix lint comments, no functional changes; ok ray

fix a few name clashes with libc; found by lint. ok deraadt@

remove unused variables and functions

lots of indentation cleanup, now ksh is readable like our other code.
double checked to make sure no binaries change, and eyed by niallo

if argv[0] not set, do not set it to "pdksh", but to "ksh"

Set \u in prompt expansion to the right value, while avoiding getpw* calls,
which might be very inconvenient when the yp server is not available.

ok deraadt@ millert@

Fix a use-after-free, that causs core dumps if a shell is killed
running with strict malloc.conf options. Problem spotted by hshoexer@;
fix by me with some help from millert@.

ok millert@ hshoexer@ krw@ deraadt@

Use stdbool.h instead of rolling our own bools.

Ansification plus some minor knf. No binary change on i386 and
sparc64, binary change in lex.o on macppc due to introduction of
dopprompt() prototype. ok millert@

deregister

When invoked as sh set SH_VERSION, not KSH_VERSION like the man page says.

Get rid of #ifdef KSH since we don't care about building a V7 style sh and
the #ifdef KSH code is required to make a POSIX sh. From Matthias Kilian

Remove unused OS dependent #ifdef blocks, #defines and macro abstraction.
First step in making the ksh code easier to read. From Matthias Kilian

Don't send SIGINT or SIGTERM to the entire processes group when the
shell receives SIGINT/SIGTERM unless the shell is the process group
leader. Fixes PR 3820. OK danh@ sturm@ miod@

do not push ksh into stupid-random mode; leave it in arc4random() mode.
oops. spotted by weerd@weirdnet.nl

typo; from ray at cyth dot net (PR 3632);

protos

two fixes; help from tedu & tdeval

spelling fixes
ok millert@

typos; from Brian Poole

knf

do not reset nonblock if shell is not interactive, from crossd@cs.rpi.edu
(i should have commited this about 6 months ago, ok millert@).

Better fix for PR #921 (main.c rev 1.18) from the pdksh cvs tree.

Fix a core dump caused by a signal race; yval@hackers.co.il

Turn <tab> completion on by default for people used to bash, tcsh, etc...

o Set default line edit mode to emacs (VISUAL/EDITOR or user can override)
o Spell environment correctly

patches from pdksh 5.2.13.11

disable newgrp alias since OpenBSD doesn't have a newgrp(1)

sync with pdksh-unstable-5.2.13.6

bug fixes from pdksh-unstable-5.2.13.5; some of which we already had.

Bug fixes from pdksh-unstable-5.2.13.4, including "official" versions of
some that we had already fixed locally.
o typeset -f FUNC doesn't print follows command (and expression) substitutions.
o when re-allocating memory, too much may be copied from old memory.
o set -o printed some options sans names.
o emacs mode: <esc>. in very fist command causes core dump.
o pdksh dumps core after a cd command.
o typeset -i reports on array elements that have no value
(at&t ksh reports on array base name - no index).
o ulimit -ctn unlimittttted kills shell (resource exceeded).
o ". /dev/null" says access denied.
o flag field in aliases incorrectly changed (all flags set instead of
clearing ISSET) in exec.c(flushcom).
o ${#array[*]} prints largest index instead of number of (set) elements
in an array (ksh88 does the former).
o sys_siglist[] doesn't always have NSIG non-null entries...

pdksh-5.2.13 + local changes

do not quitenv() too early, or "s" goes away due to it's temporary nature;
fix found in freebsd ports tree.. original by ejc@bazzle.com

i am bored enough to fix terminal space/tab uglies

back out

(foo *)0 -> NULL

Add FSH (set -o sh), initialize it if we're /bin/sh, and add the first use:
don't set $_ if we're non-interactive.

Update to 5.2.12.

Integrate pdksh 5.2.9.

update to pdksh-5.2.8

branches: 1.1.1;
Initial revision

Omit version in SMALL ksh builds

No need for KSH_VERSION and its PS1 esacape sequences in installer shells.
Save some bits and clean up what(1) output on ramdisk kernels.

OK deraadt

When system calls indicate an error they return -1, not some arbitrary
value < 0. errno is only updated in this case. Change all (most?)
callers of syscalls to follow this better, and let's see if this strictness
helps us in the future.

When evaluating an arithmetical expression, for example inside $(()),
never do substitution (neither parameter, nor command, nor arithmetic,
nor tilde substitution) on the values of any variables encountered
inside the expression, but do recursively perform arithmetical
evaluation of subexpressions as required. This makes behaviour
more consistent, without hindering any behaviour promised in the
manual page.

A quirk originally reported by Andy Chu <andychup at gmail dot com>
was that in the past, when encountering an array index, the shell
would not only do evaluation, but also substitution on the array
index, even though substitution would not be done on the expression
in general.

tobias@ contributed to initial efforts of understanding the quirk.
patch tested in a bulk build by naddy@
"please commit" deraadt@

Fix the case where the recursion detection isn't reset when the command is
interrupted.

Lots of back and forth with anton@
OK jca@, tb@, anton@

Use a very regular call pattern to pledge, so that we can continue to
grep and compare the use in all programs..

Include "id" in pledge (for setres[ug]id, setgroups) if the shell is
privileged and remove it when dropping privileges (set +p), setting a
flag to make sure we don't do it again.

ok deraadt millert

Export the PWD and OLDPWD shell variables as per POSIX.
Previously, these would only be exported if they were present
in the environment when the shell started. OK deraadt@ anton@ kn@

remove the alias type='whence -v' and replace it with
a buildin command, that just calls into c_whence().
This makes type look like the buildin in other shells and makes
things like system("'type' 'git'"); work.

With lots of suggestions and feedback from anton@, kn@ and jca@.
ok kn@ jca@

Support 64 bit integers on 32 bit architectures.

No binary change on amd64 and there should be no differences on any
other 64 bit architecture either (because long = int64_t).

ok cheloha, tb

Favor usage of __func__ in warning/error messages. Some of them referred to the
wrong function and fix the rest for consistency.
Diff from Michael W. Bombardieri with some cosmetic cleanup applied.

ok benno@ tb@

Introduce internal_warningf() and mark internal_errorf() as noreturn

This helps tools like scan-build, and follows the example of warningf()
and errorf(). ok anton@

Stop pretending we support building ksh without EDIT/HISTORY support

ok anton@

unifdef BRACE_EXPAND; ok jca@

unifdef JOBS support

Prompted by a mail from Klemens Nanni, who also had the same diff.
ok deraadt@ millert@

Include hostname in shell prompts by default

With tmux, ssh and vmd, we tend to open shells on many different hosts
simultaneously and the default prompts '$ ' and '# ' for {,k}sh as well
as '% ' and '# ' for csh become dangerous: it's very easy to issue a
command on the wrong host.

This can easily be avoided by displaying the hostname in the prompt.
Everything beyond "hostname{$,#,%} " is going to be a matter of taste,
so we left it at that. If you use an FQDN, only the first part (the
output of 'hostname -s') will be printed.

requested by and ok deraadt; mostly positive feedback many
ok anton, brynet, bcallah and others

Drop the unused second argument from set_prompt(). It used to be used for
early special casing of ! and !! in the PS1 expansion. This was removed
from set_prompt() as part of the implementaion of the character count
toggles \[ and \] back in 2004.

ok jca

Retire old behavior of requiring root prompt to contain # or \!

Requested by akoshibe and phessler
ok phessler@ anton@ jca@

No need to inspect LC_CTYPE:
* It was only used in vi input mode.
* No functional change in case of set -o vi-show8.
* No functional change if LC_CTYPE is set to UTF-8.
* More robust for the default of LC_CTYPE=C on a UTF-8 xterm.

Behaviour changes for the combination LC_CTYPE=C with set +o vi-show8
on non-UTF-8 terminals, but that combination is useless and dangerous
with or without this patch. If you want to edit individual raw non-ASCII
non-UTF-8 bytes on the shell command line, always use set -o vi-show8.
(Besides, i doubt that you actually want to do that; better use a real
hex editor in the first place.)

OK czarkoff@.

Partial UTF-8 line editing support for ksh(1) vi input mode;
so far, it covers these commands: a h i l x /
This is an isu8cont()-based hack similar in style to what i did
in emacs input mode, but less elegant and slightly more intrusive
because the vi mode code is much more ugly and less straightforward
than the emacs mode code. This one required partial rewrites of
a few helper functions, and comments were added while there.

This is not perfect, but hopefully reduces people's cursing
until a more rigorous solution can be devised (much) later.
Some polishing may be useful in tree, in particular adding
utf8cont() support to a few missing commands.

Mostly written shortly after Christmas 2015.
Reminded by and OK czarkoff@.
Feedback, partial review and testing, no longer any objection by martijn@.
Feedback and testing by tb@.
Also read fine to nicm@.

Mark PPID read-only as per the manual. From Anthony Coulter.

remove CSRG BSD-licensed mknod builtin code which was previously used to
accelerate install/upgrade time MAKEDEV runs. That allows ramdisk ksh
to be pledged, without needing "dpath". We'll solve the performance issues
a different way (soon).
ok otto espie natano tb tobiasu rpe

rename global "e" to genv to avoid accidental shadowing and aliasing.
ok millert nicm tb

Tweak comment: we now always make a copy of argv.

ok halex@

unconditionally duplicate the argv array on initialization, to make it
resilient against being altered by a subsequent shift operation

tweak and ok semarie@

Move system headers from sh.h to those files that actually need them.

ok mmcc@ a while ago

exit() after perror() for pledge failure. Perhaps this got introduced
as a test idiom, either when pledge was young or during the transition
to strings.... dunno

Add uses of areallocarray(). mksh and Bitrig ksh already have similar
functions. With help from Theo Buehler.

ok nicm@

'fc -e -' is deprecated in favor of 'fc -s', so update the built-in alias

ok millert@

Final removal of EXTERN.

ok nicm@

Penultimate commit to remove EXTERN.

ok nicm@

Move string.h include from sh.h to the files that use it.

ok nicm@

More removal of EXTERN.

ok nicm@

Apply style(9) to header includes.

ok nicm@

Move more EXTERN-defined globals from sh.h.

ok nicm@

Move a system header include from the global header (sh.h) into the
files that need it. No binary change.

"This looks fine" -nicm@

wrap a long line

Implement real "flock" request and add it to userland programs that
use pledge and file locking. OK deraadt@

normalize a few more tame request orderings, to help review

ksh can run with pledge "stdio rpath wpath cpath getpw fattr proc exec tty"
if the mknod builtin is disabled. It looks like a lot of abilities, but
hey, this is a shell. can't open sockets or do other nasty stuff though.
(we'll leave the mknod builtin enabled on the install media for now; there
is work happening to regain the MAKEDEV performance in a different way)
discussions with otto & millert in particular

Last of the (thing *)0 -> NULL, from Michael McConville. No binary
change.

Remove unnecessary casts, from Michael McConville. No binary change.

correct spelling of NULL from (char *)0. from Michael McConville.
ok md5

Replace newline and space defines by "\n" and " " directly, from Michael
McConville. ok millert

no need to check for null argv anymore. from Martijn van Duren
ok deraadt

If we hit multiple errors while unwinding we'll end up running code that
assumes a pointer is valid when it has been free'd.

This is convoluted as ksh has it's own allocator and uses long jumps.
Set the pointer to NULL after the quitenv() call in unwind() in case we
later hit a long jump in unwind().

Found with afl a while back.

ok tedu@ deraadt@

remove trailing whitespaces; use tabs instead of spaces where appropriate;
no binary changes.

Add a proper suspend builtin that saves/restores the tty and pgrp
as needed instead of an alias that just sends SIGSTOP. Login shells
may be suspended if they are not running in an orphan process group.
OK guenther@ jmc@

Run any pending traps before calling the EXIT or ERR traps when -e
is set. Fixes a bug where we would not run the signal trap if,
for example, ^C was pressed and -e was set. OK espie@

fake a sigwinch after each job, so if the terminal changes size,
we'll notice and update

Avoid modifying argv when building argv for $* and $@ since it will
affect ps output. This can happen when command line options are
specified, e.g. "sh -c command". Based on a diff from espie@
OK espie@

backout, Todd's version may be cleaner, but it's also wrong.

sh -c should not munge argv[].
This fixes ps -ww output.

joint work by millert@ and me@.

okay otto@, deraadt@ "feel free to commit my version" millert@

beter determination if the shell is restricted; from Alf Schlichting.
ok jasper@ millert@

properly keep track of the line number after a trap.
this fixes pr 6195. diff from manuel giraud, thanks.

ok kili@

pass "xerrok" status across the execution call stack to more closely
match what both POSIX and ksh.1 already describe in regards to set
-e/errexit's behavior in determining when to exit from nonzero return
values.

specifically, the truth values tested as operands to `&&' and `||', as
well as the resulting compound expression itself, along with the truth
value resulting from a negated command (i.e. a pipeline prefixed `!'),
should not make the shell exit when -e is in effect.

issue reported by matthieu.
testing matthieu, naddy.
ok miod (earlier version), otto.
man page ok jmc.

ksh starting as root will refuse to import a PS1 prompt unless it
contains a '#' character. Make it also accept the \$ PS1 sequence
(which is supposed to automagically be $ or # depending on the user
running the shell) unless ksh has been started in sh-compatibility
mode.

pointed out by millert@ when I asked why my SUDO_PS1 was not being
honoured; ok millert@

Only process ENV when in interactive mode as SU says. Started by
tedu to fix make -j, SU thing spotted by naddy@; ok millert@ @tedu
jmc@ for the man page bit.

Only check if last three letters are "/sh" if string is at least
three characters long.

OK moritz@.

fix lint comments, no functional changes; ok ray

fix a few name clashes with libc; found by lint. ok deraadt@

remove unused variables and functions

lots of indentation cleanup, now ksh is readable like our other code.
double checked to make sure no binaries change, and eyed by niallo

if argv[0] not set, do not set it to "pdksh", but to "ksh"

Set \u in prompt expansion to the right value, while avoiding getpw* calls,
which might be very inconvenient when the yp server is not available.

ok deraadt@ millert@

Fix a use-after-free, that causs core dumps if a shell is killed
running with strict malloc.conf options. Problem spotted by hshoexer@;
fix by me with some help from millert@.

ok millert@ hshoexer@ krw@ deraadt@

Use stdbool.h instead of rolling our own bools.

Ansification plus some minor knf. No binary change on i386 and
sparc64, binary change in lex.o on macppc due to introduction of
dopprompt() prototype. ok millert@

deregister

When invoked as sh set SH_VERSION, not KSH_VERSION like the man page says.

Get rid of #ifdef KSH since we don't care about building a V7 style sh and
the #ifdef KSH code is required to make a POSIX sh. From Matthias Kilian

Remove unused OS dependent #ifdef blocks, #defines and macro abstraction.
First step in making the ksh code easier to read. From Matthias Kilian

Don't send SIGINT or SIGTERM to the entire processes group when the
shell receives SIGINT/SIGTERM unless the shell is the process group
leader. Fixes PR 3820. OK danh@ sturm@ miod@

do not push ksh into stupid-random mode; leave it in arc4random() mode.
oops. spotted by weerd@weirdnet.nl

typo; from ray at cyth dot net (PR 3632);

protos

two fixes; help from tedu & tdeval

spelling fixes
ok millert@

typos; from Brian Poole

knf

do not reset nonblock if shell is not interactive, from crossd@cs.rpi.edu
(i should have commited this about 6 months ago, ok millert@).

Better fix for PR #921 (main.c rev 1.18) from the pdksh cvs tree.

Fix a core dump caused by a signal race; yval@hackers.co.il

Turn <tab> completion on by default for people used to bash, tcsh, etc...

o Set default line edit mode to emacs (VISUAL/EDITOR or user can override)
o Spell environment correctly

patches from pdksh 5.2.13.11

disable newgrp alias since OpenBSD doesn't have a newgrp(1)

sync with pdksh-unstable-5.2.13.6

bug fixes from pdksh-unstable-5.2.13.5; some of which we already had.

Bug fixes from pdksh-unstable-5.2.13.4, including "official" versions of
some that we had already fixed locally.
o typeset -f FUNC doesn't print follows command (and expression) substitutions.
o when re-allocating memory, too much may be copied from old memory.
o set -o printed some options sans names.
o emacs mode: <esc>. in very fist command causes core dump.
o pdksh dumps core after a cd command.
o typeset -i reports on array elements that have no value
(at&t ksh reports on array base name - no index).
o ulimit -ctn unlimittttted kills shell (resource exceeded).
o ". /dev/null" says access denied.
o flag field in aliases incorrectly changed (all flags set instead of
clearing ISSET) in exec.c(flushcom).
o ${#array[*]} prints largest index instead of number of (set) elements
in an array (ksh88 does the former).
o sys_siglist[] doesn't always have NSIG non-null entries...

pdksh-5.2.13 + local changes

do not quitenv() too early, or "s" goes away due to it's temporary nature;
fix found in freebsd ports tree.. original by ejc@bazzle.com

i am bored enough to fix terminal space/tab uglies

back out

(foo *)0 -> NULL

Add FSH (set -o sh), initialize it if we're /bin/sh, and add the first use:
don't set $_ if we're non-interactive.

Update to 5.2.12.

Integrate pdksh 5.2.9.

update to pdksh-5.2.8

branches: 1.1.1;
Initial revision

When system calls indicate an error they return -1, not some arbitrary
value < 0. errno is only updated in this case. Change all (most?)
callers of syscalls to follow this better, and let's see if this strictness
helps us in the future.

When evaluating an arithmetical expression, for example inside $(()),
never do substitution (neither parameter, nor command, nor arithmetic,
nor tilde substitution) on the values of any variables encountered
inside the expression, but do recursively perform arithmetical
evaluation of subexpressions as required. This makes behaviour
more consistent, without hindering any behaviour promised in the
manual page.

A quirk originally reported by Andy Chu <andychup at gmail dot com>
was that in the past, when encountering an array index, the shell
would not only do evaluation, but also substitution on the array
index, even though substitution would not be done on the expression
in general.

tobias@ contributed to initial efforts of understanding the quirk.
patch tested in a bulk build by naddy@
"please commit" deraadt@

Fix the case where the recursion detection isn't reset when the command is
interrupted.

Lots of back and forth with anton@
OK jca@, tb@, anton@

Use a very regular call pattern to pledge, so that we can continue to
grep and compare the use in all programs..

Include "id" in pledge (for setres[ug]id, setgroups) if the shell is
privileged and remove it when dropping privileges (set +p), setting a
flag to make sure we don't do it again.

ok deraadt millert

Export the PWD and OLDPWD shell variables as per POSIX.
Previously, these would only be exported if they were present
in the environment when the shell started. OK deraadt@ anton@ kn@

remove the alias type='whence -v' and replace it with
a buildin command, that just calls into c_whence().
This makes type look like the buildin in other shells and makes
things like system("'type' 'git'"); work.

With lots of suggestions and feedback from anton@, kn@ and jca@.
ok kn@ jca@

Support 64 bit integers on 32 bit architectures.

No binary change on amd64 and there should be no differences on any
other 64 bit architecture either (because long = int64_t).

ok cheloha, tb

Favor usage of __func__ in warning/error messages. Some of them referred to the
wrong function and fix the rest for consistency.
Diff from Michael W. Bombardieri with some cosmetic cleanup applied.

ok benno@ tb@

Introduce internal_warningf() and mark internal_errorf() as noreturn

This helps tools like scan-build, and follows the example of warningf()
and errorf(). ok anton@

Stop pretending we support building ksh without EDIT/HISTORY support

ok anton@

unifdef BRACE_EXPAND; ok jca@

unifdef JOBS support

Prompted by a mail from Klemens Nanni, who also had the same diff.
ok deraadt@ millert@

Include hostname in shell prompts by default

With tmux, ssh and vmd, we tend to open shells on many different hosts
simultaneously and the default prompts '$ ' and '# ' for {,k}sh as well
as '% ' and '# ' for csh become dangerous: it's very easy to issue a
command on the wrong host.

This can easily be avoided by displaying the hostname in the prompt.
Everything beyond "hostname{$,#,%} " is going to be a matter of taste,
so we left it at that. If you use an FQDN, only the first part (the
output of 'hostname -s') will be printed.

requested by and ok deraadt; mostly positive feedback many
ok anton, brynet, bcallah and others

Drop the unused second argument from set_prompt(). It used to be used for
early special casing of ! and !! in the PS1 expansion. This was removed
from set_prompt() as part of the implementaion of the character count
toggles \[ and \] back in 2004.

ok jca

Retire old behavior of requiring root prompt to contain # or \!

Requested by akoshibe and phessler
ok phessler@ anton@ jca@

No need to inspect LC_CTYPE:
* It was only used in vi input mode.
* No functional change in case of set -o vi-show8.
* No functional change if LC_CTYPE is set to UTF-8.
* More robust for the default of LC_CTYPE=C on a UTF-8 xterm.

Behaviour changes for the combination LC_CTYPE=C with set +o vi-show8
on non-UTF-8 terminals, but that combination is useless and dangerous
with or without this patch. If you want to edit individual raw non-ASCII
non-UTF-8 bytes on the shell command line, always use set -o vi-show8.
(Besides, i doubt that you actually want to do that; better use a real
hex editor in the first place.)

OK czarkoff@.

Partial UTF-8 line editing support for ksh(1) vi input mode;
so far, it covers these commands: a h i l x /
This is an isu8cont()-based hack similar in style to what i did
in emacs input mode, but less elegant and slightly more intrusive
because the vi mode code is much more ugly and less straightforward
than the emacs mode code. This one required partial rewrites of
a few helper functions, and comments were added while there.

This is not perfect, but hopefully reduces people's cursing
until a more rigorous solution can be devised (much) later.
Some polishing may be useful in tree, in particular adding
utf8cont() support to a few missing commands.

Mostly written shortly after Christmas 2015.
Reminded by and OK czarkoff@.
Feedback, partial review and testing, no longer any objection by martijn@.
Feedback and testing by tb@.
Also read fine to nicm@.

Mark PPID read-only as per the manual. From Anthony Coulter.

remove CSRG BSD-licensed mknod builtin code which was previously used to
accelerate install/upgrade time MAKEDEV runs. That allows ramdisk ksh
to be pledged, without needing "dpath". We'll solve the performance issues
a different way (soon).
ok otto espie natano tb tobiasu rpe

rename global "e" to genv to avoid accidental shadowing and aliasing.
ok millert nicm tb

Tweak comment: we now always make a copy of argv.

ok halex@

unconditionally duplicate the argv array on initialization, to make it
resilient against being altered by a subsequent shift operation

tweak and ok semarie@

Move system headers from sh.h to those files that actually need them.

ok mmcc@ a while ago

exit() after perror() for pledge failure. Perhaps this got introduced
as a test idiom, either when pledge was young or during the transition
to strings.... dunno

Add uses of areallocarray(). mksh and Bitrig ksh already have similar
functions. With help from Theo Buehler.

ok nicm@

'fc -e -' is deprecated in favor of 'fc -s', so update the built-in alias

ok millert@

Final removal of EXTERN.

ok nicm@

Penultimate commit to remove EXTERN.

ok nicm@

Move string.h include from sh.h to the files that use it.

ok nicm@

More removal of EXTERN.

ok nicm@

Apply style(9) to header includes.

ok nicm@

Move more EXTERN-defined globals from sh.h.

ok nicm@

Move a system header include from the global header (sh.h) into the
files that need it. No binary change.

"This looks fine" -nicm@

wrap a long line

Implement real "flock" request and add it to userland programs that
use pledge and file locking. OK deraadt@

normalize a few more tame request orderings, to help review

ksh can run with pledge "stdio rpath wpath cpath getpw fattr proc exec tty"
if the mknod builtin is disabled. It looks like a lot of abilities, but
hey, this is a shell. can't open sockets or do other nasty stuff though.
(we'll leave the mknod builtin enabled on the install media for now; there
is work happening to regain the MAKEDEV performance in a different way)
discussions with otto & millert in particular

Last of the (thing *)0 -> NULL, from Michael McConville. No binary
change.

Remove unnecessary casts, from Michael McConville. No binary change.

correct spelling of NULL from (char *)0. from Michael McConville.
ok md5

Replace newline and space defines by "\n" and " " directly, from Michael
McConville. ok millert

no need to check for null argv anymore. from Martijn van Duren
ok deraadt

If we hit multiple errors while unwinding we'll end up running code that
assumes a pointer is valid when it has been free'd.

This is convoluted as ksh has it's own allocator and uses long jumps.
Set the pointer to NULL after the quitenv() call in unwind() in case we
later hit a long jump in unwind().

Found with afl a while back.

ok tedu@ deraadt@

remove trailing whitespaces; use tabs instead of spaces where appropriate;
no binary changes.

Add a proper suspend builtin that saves/restores the tty and pgrp
as needed instead of an alias that just sends SIGSTOP. Login shells
may be suspended if they are not running in an orphan process group.
OK guenther@ jmc@

Run any pending traps before calling the EXIT or ERR traps when -e
is set. Fixes a bug where we would not run the signal trap if,
for example, ^C was pressed and -e was set. OK espie@

fake a sigwinch after each job, so if the terminal changes size,
we'll notice and update

Avoid modifying argv when building argv for $* and $@ since it will
affect ps output. This can happen when command line options are
specified, e.g. "sh -c command". Based on a diff from espie@
OK espie@

backout, Todd's version may be cleaner, but it's also wrong.

sh -c should not munge argv[].
This fixes ps -ww output.

joint work by millert@ and me@.

okay otto@, deraadt@ "feel free to commit my version" millert@

beter determination if the shell is restricted; from Alf Schlichting.
ok jasper@ millert@

properly keep track of the line number after a trap.
this fixes pr 6195. diff from manuel giraud, thanks.

ok kili@

pass "xerrok" status across the execution call stack to more closely
match what both POSIX and ksh.1 already describe in regards to set
-e/errexit's behavior in determining when to exit from nonzero return
values.

specifically, the truth values tested as operands to `&&' and `||', as
well as the resulting compound expression itself, along with the truth
value resulting from a negated command (i.e. a pipeline prefixed `!'),
should not make the shell exit when -e is in effect.

issue reported by matthieu.
testing matthieu, naddy.
ok miod (earlier version), otto.
man page ok jmc.

ksh starting as root will refuse to import a PS1 prompt unless it
contains a '#' character. Make it also accept the \$ PS1 sequence
(which is supposed to automagically be $ or # depending on the user
running the shell) unless ksh has been started in sh-compatibility
mode.

pointed out by millert@ when I asked why my SUDO_PS1 was not being
honoured; ok millert@

Only process ENV when in interactive mode as SU says. Started by
tedu to fix make -j, SU thing spotted by naddy@; ok millert@ @tedu
jmc@ for the man page bit.

Only check if last three letters are "/sh" if string is at least
three characters long.

OK moritz@.

fix lint comments, no functional changes; ok ray

fix a few name clashes with libc; found by lint. ok deraadt@

remove unused variables and functions

lots of indentation cleanup, now ksh is readable like our other code.
double checked to make sure no binaries change, and eyed by niallo

if argv[0] not set, do not set it to "pdksh", but to "ksh"

Set \u in prompt expansion to the right value, while avoiding getpw* calls,
which might be very inconvenient when the yp server is not available.

ok deraadt@ millert@

Fix a use-after-free, that causs core dumps if a shell is killed
running with strict malloc.conf options. Problem spotted by hshoexer@;
fix by me with some help from millert@.

ok millert@ hshoexer@ krw@ deraadt@

Use stdbool.h instead of rolling our own bools.

Ansification plus some minor knf. No binary change on i386 and
sparc64, binary change in lex.o on macppc due to introduction of
dopprompt() prototype. ok millert@

deregister

When invoked as sh set SH_VERSION, not KSH_VERSION like the man page says.

Get rid of #ifdef KSH since we don't care about building a V7 style sh and
the #ifdef KSH code is required to make a POSIX sh. From Matthias Kilian

Remove unused OS dependent #ifdef blocks, #defines and macro abstraction.
First step in making the ksh code easier to read. From Matthias Kilian

Don't send SIGINT or SIGTERM to the entire processes group when the
shell receives SIGINT/SIGTERM unless the shell is the process group
leader. Fixes PR 3820. OK danh@ sturm@ miod@

do not push ksh into stupid-random mode; leave it in arc4random() mode.
oops. spotted by weerd@weirdnet.nl

typo; from ray at cyth dot net (PR 3632);

protos

two fixes; help from tedu & tdeval

spelling fixes
ok millert@

typos; from Brian Poole

knf

do not reset nonblock if shell is not interactive, from crossd@cs.rpi.edu
(i should have commited this about 6 months ago, ok millert@).

Better fix for PR #921 (main.c rev 1.18) from the pdksh cvs tree.

Fix a core dump caused by a signal race; yval@hackers.co.il

Turn <tab> completion on by default for people used to bash, tcsh, etc...

o Set default line edit mode to emacs (VISUAL/EDITOR or user can override)
o Spell environment correctly

patches from pdksh 5.2.13.11

disable newgrp alias since OpenBSD doesn't have a newgrp(1)

sync with pdksh-unstable-5.2.13.6

bug fixes from pdksh-unstable-5.2.13.5; some of which we already had.

Bug fixes from pdksh-unstable-5.2.13.4, including "official" versions of
some that we had already fixed locally.
o typeset -f FUNC doesn't print follows command (and expression) substitutions.
o when re-allocating memory, too much may be copied from old memory.
o set -o printed some options sans names.
o emacs mode: <esc>. in very fist command causes core dump.
o pdksh dumps core after a cd command.
o typeset -i reports on array elements that have no value
(at&t ksh reports on array base name - no index).
o ulimit -ctn unlimittttted kills shell (resource exceeded).
o ". /dev/null" says access denied.
o flag field in aliases incorrectly changed (all flags set instead of
clearing ISSET) in exec.c(flushcom).
o ${#array[*]} prints largest index instead of number of (set) elements
in an array (ksh88 does the former).
o sys_siglist[] doesn't always have NSIG non-null entries...

pdksh-5.2.13 + local changes

do not quitenv() too early, or "s" goes away due to it's temporary nature;
fix found in freebsd ports tree.. original by ejc@bazzle.com

i am bored enough to fix terminal space/tab uglies

back out

(foo *)0 -> NULL

Add FSH (set -o sh), initialize it if we're /bin/sh, and add the first use:
don't set $_ if we're non-interactive.

Update to 5.2.12.

Integrate pdksh 5.2.9.

update to pdksh-5.2.8

branches: 1.1.1;
Initial revision

When evaluating an arithmetical expression, for example inside $(()),
never do substitution (neither parameter, nor command, nor arithmetic,
nor tilde substitution) on the values of any variables encountered
inside the expression, but do recursively perform arithmetical
evaluation of subexpressions as required. This makes behaviour
more consistent, without hindering any behaviour promised in the
manual page.

A quirk originally reported by Andy Chu <andychup at gmail dot com>
was that in the past, when encountering an array index, the shell
would not only do evaluation, but also substitution on the array
index, even though substitution would not be done on the expression
in general.

tobias@ contributed to initial efforts of understanding the quirk.
patch tested in a bulk build by naddy@
"please commit" deraadt@

Fix the case where the recursion detection isn't reset when the command is
interrupted.

Lots of back and forth with anton@
OK jca@, tb@, anton@

Use a very regular call pattern to pledge, so that we can continue to
grep and compare the use in all programs..

Include "id" in pledge (for setres[ug]id, setgroups) if the shell is
privileged and remove it when dropping privileges (set +p), setting a
flag to make sure we don't do it again.

ok deraadt millert

Export the PWD and OLDPWD shell variables as per POSIX.
Previously, these would only be exported if they were present
in the environment when the shell started. OK deraadt@ anton@ kn@

remove the alias type='whence -v' and replace it with
a buildin command, that just calls into c_whence().
This makes type look like the buildin in other shells and makes
things like system("'type' 'git'"); work.

With lots of suggestions and feedback from anton@, kn@ and jca@.
ok kn@ jca@

Support 64 bit integers on 32 bit architectures.

No binary change on amd64 and there should be no differences on any
other 64 bit architecture either (because long = int64_t).

ok cheloha, tb

Favor usage of __func__ in warning/error messages. Some of them referred to the
wrong function and fix the rest for consistency.
Diff from Michael W. Bombardieri with some cosmetic cleanup applied.

ok benno@ tb@

Introduce internal_warningf() and mark internal_errorf() as noreturn

This helps tools like scan-build, and follows the example of warningf()
and errorf(). ok anton@

Stop pretending we support building ksh without EDIT/HISTORY support

ok anton@

unifdef BRACE_EXPAND; ok jca@

unifdef JOBS support

Prompted by a mail from Klemens Nanni, who also had the same diff.
ok deraadt@ millert@

Include hostname in shell prompts by default

With tmux, ssh and vmd, we tend to open shells on many different hosts
simultaneously and the default prompts '$ ' and '# ' for {,k}sh as well
as '% ' and '# ' for csh become dangerous: it's very easy to issue a
command on the wrong host.

This can easily be avoided by displaying the hostname in the prompt.
Everything beyond "hostname{$,#,%} " is going to be a matter of taste,
so we left it at that. If you use an FQDN, only the first part (the
output of 'hostname -s') will be printed.

requested by and ok deraadt; mostly positive feedback many
ok anton, brynet, bcallah and others

Drop the unused second argument from set_prompt(). It used to be used for
early special casing of ! and !! in the PS1 expansion. This was removed
from set_prompt() as part of the implementaion of the character count
toggles \[ and \] back in 2004.

ok jca

Retire old behavior of requiring root prompt to contain # or \!

Requested by akoshibe and phessler
ok phessler@ anton@ jca@

No need to inspect LC_CTYPE:
* It was only used in vi input mode.
* No functional change in case of set -o vi-show8.
* No functional change if LC_CTYPE is set to UTF-8.
* More robust for the default of LC_CTYPE=C on a UTF-8 xterm.

Behaviour changes for the combination LC_CTYPE=C with set +o vi-show8
on non-UTF-8 terminals, but that combination is useless and dangerous
with or without this patch. If you want to edit individual raw non-ASCII
non-UTF-8 bytes on the shell command line, always use set -o vi-show8.
(Besides, i doubt that you actually want to do that; better use a real
hex editor in the first place.)

OK czarkoff@.

Partial UTF-8 line editing support for ksh(1) vi input mode;
so far, it covers these commands: a h i l x /
This is an isu8cont()-based hack similar in style to what i did
in emacs input mode, but less elegant and slightly more intrusive
because the vi mode code is much more ugly and less straightforward
than the emacs mode code. This one required partial rewrites of
a few helper functions, and comments were added while there.

This is not perfect, but hopefully reduces people's cursing
until a more rigorous solution can be devised (much) later.
Some polishing may be useful in tree, in particular adding
utf8cont() support to a few missing commands.

Mostly written shortly after Christmas 2015.
Reminded by and OK czarkoff@.
Feedback, partial review and testing, no longer any objection by martijn@.
Feedback and testing by tb@.
Also read fine to nicm@.

Mark PPID read-only as per the manual. From Anthony Coulter.

remove CSRG BSD-licensed mknod builtin code which was previously used to
accelerate install/upgrade time MAKEDEV runs. That allows ramdisk ksh
to be pledged, without needing "dpath". We'll solve the performance issues
a different way (soon).
ok otto espie natano tb tobiasu rpe

rename global "e" to genv to avoid accidental shadowing and aliasing.
ok millert nicm tb

Tweak comment: we now always make a copy of argv.

ok halex@

unconditionally duplicate the argv array on initialization, to make it
resilient against being altered by a subsequent shift operation

tweak and ok semarie@

Move system headers from sh.h to those files that actually need them.

ok mmcc@ a while ago

exit() after perror() for pledge failure. Perhaps this got introduced
as a test idiom, either when pledge was young or during the transition
to strings.... dunno

Add uses of areallocarray(). mksh and Bitrig ksh already have similar
functions. With help from Theo Buehler.

ok nicm@

'fc -e -' is deprecated in favor of 'fc -s', so update the built-in alias

ok millert@

Final removal of EXTERN.

ok nicm@

Penultimate commit to remove EXTERN.

ok nicm@

Move string.h include from sh.h to the files that use it.

ok nicm@

More removal of EXTERN.

ok nicm@

Apply style(9) to header includes.

ok nicm@

Move more EXTERN-defined globals from sh.h.

ok nicm@

Move a system header include from the global header (sh.h) into the
files that need it. No binary change.

"This looks fine" -nicm@

wrap a long line

Implement real "flock" request and add it to userland programs that
use pledge and file locking. OK deraadt@

normalize a few more tame request orderings, to help review

ksh can run with pledge "stdio rpath wpath cpath getpw fattr proc exec tty"
if the mknod builtin is disabled. It looks like a lot of abilities, but
hey, this is a shell. can't open sockets or do other nasty stuff though.
(we'll leave the mknod builtin enabled on the install media for now; there
is work happening to regain the MAKEDEV performance in a different way)
discussions with otto & millert in particular

Last of the (thing *)0 -> NULL, from Michael McConville. No binary
change.

Remove unnecessary casts, from Michael McConville. No binary change.

correct spelling of NULL from (char *)0. from Michael McConville.
ok md5

Replace newline and space defines by "\n" and " " directly, from Michael
McConville. ok millert

no need to check for null argv anymore. from Martijn van Duren
ok deraadt

If we hit multiple errors while unwinding we'll end up running code that
assumes a pointer is valid when it has been free'd.

This is convoluted as ksh has it's own allocator and uses long jumps.
Set the pointer to NULL after the quitenv() call in unwind() in case we
later hit a long jump in unwind().

Found with afl a while back.

ok tedu@ deraadt@

remove trailing whitespaces; use tabs instead of spaces where appropriate;
no binary changes.

Add a proper suspend builtin that saves/restores the tty and pgrp
as needed instead of an alias that just sends SIGSTOP. Login shells
may be suspended if they are not running in an orphan process group.
OK guenther@ jmc@

Run any pending traps before calling the EXIT or ERR traps when -e
is set. Fixes a bug where we would not run the signal trap if,
for example, ^C was pressed and -e was set. OK espie@

fake a sigwinch after each job, so if the terminal changes size,
we'll notice and update

Avoid modifying argv when building argv for $* and $@ since it will
affect ps output. This can happen when command line options are
specified, e.g. "sh -c command". Based on a diff from espie@
OK espie@

backout, Todd's version may be cleaner, but it's also wrong.

sh -c should not munge argv[].
This fixes ps -ww output.

joint work by millert@ and me@.

okay otto@, deraadt@ "feel free to commit my version" millert@

beter determination if the shell is restricted; from Alf Schlichting.
ok jasper@ millert@

properly keep track of the line number after a trap.
this fixes pr 6195. diff from manuel giraud, thanks.

ok kili@

pass "xerrok" status across the execution call stack to more closely
match what both POSIX and ksh.1 already describe in regards to set
-e/errexit's behavior in determining when to exit from nonzero return
values.

specifically, the truth values tested as operands to `&&' and `||', as
well as the resulting compound expression itself, along with the truth
value resulting from a negated command (i.e. a pipeline prefixed `!'),
should not make the shell exit when -e is in effect.

issue reported by matthieu.
testing matthieu, naddy.
ok miod (earlier version), otto.
man page ok jmc.

ksh starting as root will refuse to import a PS1 prompt unless it
contains a '#' character. Make it also accept the \$ PS1 sequence
(which is supposed to automagically be $ or # depending on the user
running the shell) unless ksh has been started in sh-compatibility
mode.

pointed out by millert@ when I asked why my SUDO_PS1 was not being
honoured; ok millert@

Only process ENV when in interactive mode as SU says. Started by
tedu to fix make -j, SU thing spotted by naddy@; ok millert@ @tedu
jmc@ for the man page bit.

Only check if last three letters are "/sh" if string is at least
three characters long.

OK moritz@.

fix lint comments, no functional changes; ok ray

fix a few name clashes with libc; found by lint. ok deraadt@

remove unused variables and functions

lots of indentation cleanup, now ksh is readable like our other code.
double checked to make sure no binaries change, and eyed by niallo

if argv[0] not set, do not set it to "pdksh", but to "ksh"

Set \u in prompt expansion to the right value, while avoiding getpw* calls,
which might be very inconvenient when the yp server is not available.

ok deraadt@ millert@

Fix a use-after-free, that causs core dumps if a shell is killed
running with strict malloc.conf options. Problem spotted by hshoexer@;
fix by me with some help from millert@.

ok millert@ hshoexer@ krw@ deraadt@

Use stdbool.h instead of rolling our own bools.

Ansification plus some minor knf. No binary change on i386 and
sparc64, binary change in lex.o on macppc due to introduction of
dopprompt() prototype. ok millert@

deregister

When invoked as sh set SH_VERSION, not KSH_VERSION like the man page says.

Get rid of #ifdef KSH since we don't care about building a V7 style sh and
the #ifdef KSH code is required to make a POSIX sh. From Matthias Kilian

Remove unused OS dependent #ifdef blocks, #defines and macro abstraction.
First step in making the ksh code easier to read. From Matthias Kilian

Don't send SIGINT or SIGTERM to the entire processes group when the
shell receives SIGINT/SIGTERM unless the shell is the process group
leader. Fixes PR 3820. OK danh@ sturm@ miod@

do not push ksh into stupid-random mode; leave it in arc4random() mode.
oops. spotted by weerd@weirdnet.nl

typo; from ray at cyth dot net (PR 3632);

protos

two fixes; help from tedu & tdeval

spelling fixes
ok millert@

typos; from Brian Poole

knf

do not reset nonblock if shell is not interactive, from crossd@cs.rpi.edu
(i should have commited this about 6 months ago, ok millert@).

Better fix for PR #921 (main.c rev 1.18) from the pdksh cvs tree.

Fix a core dump caused by a signal race; yval@hackers.co.il

Turn <tab> completion on by default for people used to bash, tcsh, etc...

o Set default line edit mode to emacs (VISUAL/EDITOR or user can override)
o Spell environment correctly

patches from pdksh 5.2.13.11

disable newgrp alias since OpenBSD doesn't have a newgrp(1)

sync with pdksh-unstable-5.2.13.6

bug fixes from pdksh-unstable-5.2.13.5; some of which we already had.

Bug fixes from pdksh-unstable-5.2.13.4, including "official" versions of
some that we had already fixed locally.
o typeset -f FUNC doesn't print follows command (and expression) substitutions.
o when re-allocating memory, too much may be copied from old memory.
o set -o printed some options sans names.
o emacs mode: <esc>. in very fist command causes core dump.
o pdksh dumps core after a cd command.
o typeset -i reports on array elements that have no value
(at&t ksh reports on array base name - no index).
o ulimit -ctn unlimittttted kills shell (resource exceeded).
o ". /dev/null" says access denied.
o flag field in aliases incorrectly changed (all flags set instead of
clearing ISSET) in exec.c(flushcom).
o ${#array[*]} prints largest index instead of number of (set) elements
in an array (ksh88 does the former).
o sys_siglist[] doesn't always have NSIG non-null entries...

pdksh-5.2.13 + local changes

do not quitenv() too early, or "s" goes away due to it's temporary nature;
fix found in freebsd ports tree.. original by ejc@bazzle.com

i am bored enough to fix terminal space/tab uglies

back out

(foo *)0 -> NULL

Add FSH (set -o sh), initialize it if we're /bin/sh, and add the first use:
don't set $_ if we're non-interactive.

Update to 5.2.12.

Integrate pdksh 5.2.9.

update to pdksh-5.2.8

branches: 1.1.1;
Initial revision

Fix the case where the recursion detection isn't reset when the command is
interrupted.

Lots of back and forth with anton@
OK jca@, tb@, anton@

Use a very regular call pattern to pledge, so that we can continue to
grep and compare the use in all programs..

Include "id" in pledge (for setres[ug]id, setgroups) if the shell is
privileged and remove it when dropping privileges (set +p), setting a
flag to make sure we don't do it again.

ok deraadt millert

Export the PWD and OLDPWD shell variables as per POSIX.
Previously, these would only be exported if they were present
in the environment when the shell started. OK deraadt@ anton@ kn@

remove the alias type='whence -v' and replace it with
a buildin command, that just calls into c_whence().
This makes type look like the buildin in other shells and makes
things like system("'type' 'git'"); work.

With lots of suggestions and feedback from anton@, kn@ and jca@.
ok kn@ jca@

Support 64 bit integers on 32 bit architectures.

No binary change on amd64 and there should be no differences on any
other 64 bit architecture either (because long = int64_t).

ok cheloha, tb

Favor usage of __func__ in warning/error messages. Some of them referred to the
wrong function and fix the rest for consistency.
Diff from Michael W. Bombardieri with some cosmetic cleanup applied.

ok benno@ tb@

Introduce internal_warningf() and mark internal_errorf() as noreturn

This helps tools like scan-build, and follows the example of warningf()
and errorf(). ok anton@

Stop pretending we support building ksh without EDIT/HISTORY support

ok anton@

unifdef BRACE_EXPAND; ok jca@

unifdef JOBS support

Prompted by a mail from Klemens Nanni, who also had the same diff.
ok deraadt@ millert@

Include hostname in shell prompts by default

With tmux, ssh and vmd, we tend to open shells on many different hosts
simultaneously and the default prompts '$ ' and '# ' for {,k}sh as well
as '% ' and '# ' for csh become dangerous: it's very easy to issue a
command on the wrong host.

This can easily be avoided by displaying the hostname in the prompt.
Everything beyond "hostname{$,#,%} " is going to be a matter of taste,
so we left it at that. If you use an FQDN, only the first part (the
output of 'hostname -s') will be printed.

requested by and ok deraadt; mostly positive feedback many
ok anton, brynet, bcallah and others

Drop the unused second argument from set_prompt(). It used to be used for
early special casing of ! and !! in the PS1 expansion. This was removed
from set_prompt() as part of the implementaion of the character count
toggles \[ and \] back in 2004.

ok jca

Retire old behavior of requiring root prompt to contain # or \!

Requested by akoshibe and phessler
ok phessler@ anton@ jca@

No need to inspect LC_CTYPE:
* It was only used in vi input mode.
* No functional change in case of set -o vi-show8.
* No functional change if LC_CTYPE is set to UTF-8.
* More robust for the default of LC_CTYPE=C on a UTF-8 xterm.

Behaviour changes for the combination LC_CTYPE=C with set +o vi-show8
on non-UTF-8 terminals, but that combination is useless and dangerous
with or without this patch. If you want to edit individual raw non-ASCII
non-UTF-8 bytes on the shell command line, always use set -o vi-show8.
(Besides, i doubt that you actually want to do that; better use a real
hex editor in the first place.)

OK czarkoff@.

Partial UTF-8 line editing support for ksh(1) vi input mode;
so far, it covers these commands: a h i l x /
This is an isu8cont()-based hack similar in style to what i did
in emacs input mode, but less elegant and slightly more intrusive
because the vi mode code is much more ugly and less straightforward
than the emacs mode code. This one required partial rewrites of
a few helper functions, and comments were added while there.

This is not perfect, but hopefully reduces people's cursing
until a more rigorous solution can be devised (much) later.
Some polishing may be useful in tree, in particular adding
utf8cont() support to a few missing commands.

Mostly written shortly after Christmas 2015.
Reminded by and OK czarkoff@.
Feedback, partial review and testing, no longer any objection by martijn@.
Feedback and testing by tb@.
Also read fine to nicm@.

Mark PPID read-only as per the manual. From Anthony Coulter.

remove CSRG BSD-licensed mknod builtin code which was previously used to
accelerate install/upgrade time MAKEDEV runs. That allows ramdisk ksh
to be pledged, without needing "dpath". We'll solve the performance issues
a different way (soon).
ok otto espie natano tb tobiasu rpe

rename global "e" to genv to avoid accidental shadowing and aliasing.
ok millert nicm tb

Tweak comment: we now always make a copy of argv.

ok halex@

unconditionally duplicate the argv array on initialization, to make it
resilient against being altered by a subsequent shift operation

tweak and ok semarie@

Move system headers from sh.h to those files that actually need them.

ok mmcc@ a while ago

exit() after perror() for pledge failure. Perhaps this got introduced
as a test idiom, either when pledge was young or during the transition
to strings.... dunno

Add uses of areallocarray(). mksh and Bitrig ksh already have similar
functions. With help from Theo Buehler.

ok nicm@

'fc -e -' is deprecated in favor of 'fc -s', so update the built-in alias

ok millert@

Final removal of EXTERN.

ok nicm@

Penultimate commit to remove EXTERN.

ok nicm@

Move string.h include from sh.h to the files that use it.

ok nicm@