Please share your BSD Tips and Tricks and we can publish a selection of them on the aggregator page: https://www.moritz.systems/bsd-tips-and-tricks/ Enjoy!

"Find the largest files in the current directory" by Kamil Rytarowski find . -type f -exec stat -f "%z %N" {} \; | sort -r -n -k1 if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } https://www.moritz.systems/bsd-tips-and-tricks/find-largest-files-in-the-current-directory/ #FreeBSD #NetBSD #OpenBSD

@kamil ...and with some eye-candy 😉 I'm really liking this broot tool.

EDIT: removed stray ; in the c function. kamil 3 selections: Remove duplicate lines in files (eg. rc.conf files, shell history files): $ cat ~/bin/nuniq.sh #!/bin/sh # # nuniq.sh: Remove duplicate lines, leaving only the last one. # (based on histsort.awk from "GAWK: Effective AWK Programming") exec awk ' { lines[NR] = $0 if ($0 in array) delete lines[array[$0]] array[$0] = NR } END { for (i = 1; i <= NR; i++) if (i in lines) print lines[i] }' "$@" if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } Show diff output/patch files in colour: $ cat ~/bin/cpatch.sh #!/bin/sh # # Colour unified diffs. org=$(tput sgr0) # Reset attrs. old=$(tput setf 4 bold) # Bold Red new=$(tput setf 6 bold) # Bold Yellow exec sed -e "s/^+.*$/${new}&${org}/; s/^-.*$/${old}&${org}/" -- "$@" if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } Use as: diff -urN DIR1 DIR2 | cpatch.sh OR cpatch.sh a.patch b.patch ... Emulate clear in restricted constrained environments which don't have a clear and the associated termcap / terminfo database and library--eg. an OpenBSD USB install stick 😠: c() { printf '\033[H\033[J' } if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } Works on all ANSI-/VT100-compatible terminals.

To reduce the cycle of asking people for permission. Pasting here the tips and tricks means that I can publish them on the website freely, without asking and with the credit for submitter.

A tip which is trivial, but overlooked: Use -C to specify the directory in which make should run, as in: make -C /usr/pkgsrc/devel/git-base install clean clean-depends if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); }

MVMPIYTGA! Make viewing man pages in your terminal great again! alias man='w3mman'

rvp Is this form OK: alias clr='printf "\033[H\033[J"' ? I needed to delete the semicolon character as otherwise it was printed in the termnial (NetBSD, ksh).

[unknown] Use clear instead of hardcoding the sequence to stay terminal independent. If you need it multiple times and don't want to call en external binary each time: $ CLEAR=$(clear) # or tput clear $ printf "$CLEAR" | hexdump -c 0000000 033 [ H 033 [ J 0000006 if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); }

kamil I needed to delete the semicolon character I swear that wasn't there when I typed it in. It's these semicolons I tell you--they sneak right in somehow. Thank you for eradicating the beast. I made the clear work-alike a function because a) I prefer functions, and b) some shells don't have an alias command. However, alias is in POSIX, so go right ahead with them. As for permissions, etc: All code/scripts I write and post on this board can be considered to be in the public-domain. Use as you wish. Code & scripts I did not write, I will try to attribute correctly, and for these, you'll have to use your own judgement. Some sed(1) one-liners: The sed(1) "list" command is very useful when dealing with unprintable (and invisible!) chars. I use it all the time because the other methods of dealing with these pesky characters is not very convenient: hexdump(1) is great, but, it doesn't preserve line endings; and vis(1) is a BSD-only thing. $ printf 'MS-DOS line ending\r\n' | sed -n l MS-DOS line ending\r$ $ printf 'Non-breaking space: '$'\u00A0''\n' | LC_ALL=C sed -n l Non-breaking space: \302\240$ if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } Insert thousands-separators into numbers. The -h option to commands like du(1), df(1), etc. are nice and all, but, sometimes I want to see whole numbers in outputs. Then separators are useful to make sense of large numbers. #!/usr/bin/sed -Ef # # Insert US-style thousands-separators (commas) into numbers. : loop s/(^|[[:blank:]]+)([0-9]{1,})([0-9]{3})([0-9,]*)([[:blank:]]+|$)/\1\2,\3\4\5/ t loop if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } Use as: du -a /some/dir | comma , or netstat -b -w 1 -I alc0 | comma , etc. Number lines in file. #!/bin/sh # # Simulate nl(1) or cat(1) -n exec sed -e = -- ${1+"$1"} | sed -nEe 'N;s/([[:digit:]]+)./\1'$'\t''/;p' if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } NetBSD doesn't come with a tree(1) command so here's one in sed(1): #!/bin/sh # # Print directory tree exec find -- ${1:-.} | sed -Ee '2,$s:[^/]*/: :g; s: ([^ ]+)$:|_ \1:' if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } Note: it doesn't draw a fully-connected tree. Drawing a full tree is very hairy--and, completely out of our self-imposed one-liners category. Reverse-print a file--aka tac(1): #!/bin/sh # # Implement tac(1) using sed(1). exec sed -ne '1h; 2,${x;H;}; ${x;p;}' -- "$@" if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); } Same thing in ed(1); using the standard idiom: #!/bin/sh # # Implement tac(1) using ed(1). : ${1?"Usage: ${0##*/} FILE"} exec ed -s -- "$1" <<\EoF g/^.*$/m0 %p EoF if(window.hljsLoader && !document.currentScript.parentNode.hasAttribute('data-s9e-livepreview-onupdate')) { window.hljsLoader.highlightBlocks(document.currentScript.parentNode); }

qik A tip which is trivial, but overlooked: Use -C to specify the directory in which make should run Yes, just like patch -d /some/dir ...

Moritz Systems | BSD Tips and Tricks

JuvenalUrbino

kamil Set your favorite pager

env PAGER=less man 1 intro

Worth mentioning the $LESS environmental variable, which defines less(1) default behavior:
This makes less clear the screen by default, turns on case-insensitive search, and sets a nice bottom prompt displaying fundamental built-in keys

LESS='-ci -P \spacebar\:pg_ahead  b\:pg_back  /\:search_ahead  \?\:search_back  h\:help  q\:quit'; export LESS

yeti

JuvenalUrbino Have a look at w3mman.

rvp

yeti w3mman is nice, but, it needs perl

JuvenalUrbino

script to automatically configure multiscreen setup:

#!/bin/sh
# Grab the list of enabled outputs
OUTPUTS=$(xrandr | \
    sed -n 's/^\(^[A-Z0-9-]*\) connected [0-9x]*\(+0+0\)*.*/\1\2/p')
# Example:
#  VGA1+0+0
#  DVI1

ARGS="--auto"
for output in $OUTPUTS; do
    # Use the name of the first output to determine the order we
    # should use. Consecutive calls of this script will therefore
    # reverse the order.
    case "$output:$prevoutput" in
	*+0+0:)
	    DIRECTION=left-of
	    ;;
	*:)
	    DIRECTION=right-of
	    ;;
    esac

# Autoconfigure this output
ARGS="$ARGS --output ${output%%+*} --auto"

# And put it at left or at right of the previous output
[ -n "$prevoutput" ] && \
ARGS="$ARGS --output ${output%%+*} --$DIRECTION $prevoutput"

prevoutput=${output%%+*}
done

# Call xrandr
xrandr $ARGS

JuvenalUrbino

set default number of make jobs depending on HW_NCPUONLINE integer

CPUS=$(sysctl -a |grep hw\.ncpuonline|sed 's/[^0-9]*//g')
alias make='make -j$CPUS'

yeti

JuvenalUrbino Depending on local PATH settings users might not have sysctl in their PATH.

And what's wrong with this sysctl call?

~$ /sbin/sysctl -n hw.ncpuonline
24

It seems to save ... | grep ... | sed ... and might even be short enough to remember.

Maybe this is an idea?

export MAKEFLAGS="-j $(/sbin/sysctl -n hw.ncpuonline)"

rvp

JuvenalUrbino I usually set make -j N to N = hw.ncpuonline - 1: You need at least one CPU free to kill run-away jobs--eg. compiling GCC.

JuvenalUrbino

yeti thanks so much. that was indeed a bit of horrid a mess I had come up with.
My bad habit of piping sysctl -a through grep instead of specifying the exact call on the command line is though to kill and sometimes I forget about it. It's sort of cat|grep bad practice

yeti /sbin/sysctl -n hw.ncpuonline

oh, I must admit I didn't know sysctl -n were a thing, how stupid. This indeed saves a lot of sed.

yeti Maybe this is an idea?
export MAKEFLAGS="-j $(/sbin/sysctl -n hw.ncpuonline)"

definitely 😆

yeti

I sometimes iterate over adding stuff to previously proofread commands for postprocessing their output. Something like "coward's complex commandline composition". Explosive like C4! ;-D

$ pkg_info

Ok... we know the "package-version description" lines...

$ pkg_info | awk '$0=$1'

Only shows the "package-version" part.

$ pkg_info | awk '$0=$1' | xargs echo pkg_admin set rebuild=yes

...proofread and if happy with it either change the echo to sudo or pipe it into sudo sh.

Adding postprocessing to previously proofread commands really keeps adrenaline levels low.

And nah! I do not rebuild all packages often. It just looked like a nice example.

oui

CPU sets to build software - useful for CPUs using "big.LITTLE" architecture where some cores are faster than others.

E.g. on my RockPRO64 that has 2 big cores up to 2.0 GHz and 4 smaller ones at 1.4 GHz:

Find out which cores are faster (Cortex-A72) and slower (Cortex-A53):

dmesg | grep cpu
[     1.000000] cpu0 at cpus0: Arm Cortex-A53 r0p4 (v8-A), id 0x0
[     1.000000] cpu1 at cpus0: Arm Cortex-A53 r0p4 (v8-A), id 0x1
[     1.000000] cpufreqdt1 at cpu1
[     1.000000] cpu2 at cpus0: Arm Cortex-A53 r0p4 (v8-A), id 0x2
[     1.000000] cpufreqdt2 at cpu2
[     1.000000] cpu3 at cpus0: Arm Cortex-A53 r0p4 (v8-A), id 0x3
[     1.000000] cpufreqdt3 at cpu3
[     1.000000] cpu4 at cpus0: Arm Cortex-A72 r0p2 (v8-A), id 0x100
[     1.000000] cpufreqdt4 at cpu4
[     1.000000] cpu5 at cpus0: Arm Cortex-A72 r0p2 (v8-A), id 0x101

List CPU sets

psrset
system processor set 0: processor(s) 0 1 2 3 4 5

Create a new CPU set with the 2 faster cores (4 & 5)

psrset -c 4 5
1

List CPU sets (note the newly created CPU set with id 1)

psrset
system processor set 0: processor(s) 0 1 2 3
user processor set 1: processor(s) 4 5

We can now build software only using the 2 faster cores and leaving the 4 slower cores available for user input etc.

psrset -e 1 make install clean -j 2

Don't forget to delete the newly created CPU set after building as it removes the specified cores away from the default CPU set:

psrset -d 1

yeti

I'm playing with some NetBSD systems where I only use home baked PkgSrc packages and was looking for a way to automate creating the pkg_summary.gz file.

After some digging I've put ...

##
## build pkg_summary.gz after making a binary package
##
release-package-lock:
        ( cd ${PACKAGES}/All && pkg_bin_summary | gzip -9 > pkg_summary.gz )

... in the 1st section protected with ...

.ifdef BSD_PKG_MK
## active while building packages
...
.else # ifdef BSD_PKG_MK
## here be the rest (build.sh stuff and friends)
...
.endif

That's the 1st time I dared to put a target into mk.conf. It even could be a one liner (use ';' to join the lines) and that feels really like minimal invasive. Maybe there is a better way, but until then, I might be happy with this hack.

Better ideas someone?

rvp

yeti Better ideas someone?

Pull in the system Makefiles using include in yours?

JuvenalUrbino

per process temporary shell history:

HISTFILE="$HOME/.hist_$$"; export HISTFILE
if [ ! -e $HISTFILE ]; then
    touch $HISTFILE
fi  

trap 'rm -f $HISTFILE' EXIT

To bring up history I also use:

h (){ fc -l 1 | sort | uniq; }

rvp

Remove duplicate and nonexistent directories from PATH

#!/usr/bin/awk -f
#
# Remove duplicate and nonexistent directories from PATH
#
# Dir. names must not contain newlines or `:'; everything 
# else should be OK.
#
# Usage: export PATH="$(repath "$PATH")" (sh)  OR
#        export PATH="$(repath)"               OR
#        setenv PATH "`repath "$PATH"`"  (csh) OR
#        setenv PATH "`repath`"

# Quote `arg' for shell
function q(arg, s)
{
	s = arg
	gsub(/'/, "'\\''", s)	# first, turn every ' -> '\''
	return "'"s"'"		# then, single-quote the whole thing
}
BEGIN {
	path = ARGC == 2 ? ARGV[1] : ENVIRON["PATH"]
	sep = ":"
	n = split(path, P, sep)
	for (i = 1; i <= n; i++) {
		dir = P[i] == "" ? "." : P[i]
		if (H[dir]++ == 0 &&
		    system(sprintf("test -d %s", q(dir))) == 0)
			printf("%s%s", c++ == 0 ? "" : sep, dir)
	}
	if (c) printf("\n");
}

To use this script, put something like this towards the end of your ~/.profile or ~/.shrc--after all PATHs have been set:

export PATH="$(repath "$PATH")"

csh users should use this in their ~/.login or ~/.cshrc files:

setenv PATH "`repath "$PATH"`"

JuvenalUrbino

Script to set up a user-owned tmp directory with restricted access. Good to be sourced inside .xinitrc | .xsession

#!/bin/sh
USER=$(whoami)
GROUP=$(id -ng $USER)
mkdir /tmp/${USER}-runtime 
chown $USER:$GROUP /tmp/${USER}-runtime
chmod -R 0700 /tmp/${USER}-runtime
XDG_RUNTIME_DIR=/tmp/${USER}-runtime; export XDG_RUNTIME_DIR

Script to create temporary user cache on volatile memory.

# Check if ~/.cache is a symbolic link to a directory owned by the user.
# If the link's target is not a directory or does not exist, create a
# directory owned by the user under /tmp, and make ~/.cache a symbolic
# link to it.
#
uid=$(id -u)
if [ -L ~/.cache ]; then
	if [ -d ~/.cache ]; then
		ownerid=$(stat -L -f %u ~/.cache)
		if [ "${ownerid}" != "${uid}" ]; then
			recreate_cache=1
		fi
	else
		recreate_cache=1
	fi
	if [ "${recreate_cache}" = "1" ]; then
		user=$(whoami)
		unlink ~/.cache
		cachedir=$(mktemp -d /tmp/${user}.XXXXXX)
		ln -s ${cachedir} ~/.cache
	fi
fi

rvp

JuvenalUrbino This is pretty much what I do on OpenBSD for $XDG_RUNTIME_DIR--except that I use install.

JuvenalUrbino

Small script to quickly get a list of onboard PCI devices names and IDs

/usr/sbin/pcictl pci0 list \
 | awk -F: '{print "pcictl pci0 dump -b "$1" -d "$2}' \
 | sort -u| sh | grep -i 'device name' | cut -d ':' -f 2

On action:

thinkpad_R51:~ $ ids                                                                                            

 82855GM Host-Hub Controller (0x3580)
 82855GM GMCH Integrated Graphics Device (0x3582)
 82801DB USB UHCI Controller (0x24c2)
 82801BAM Hub-PCI Bridge (0x2448)
 82801DB LPC Interface Bridge (0x24cc)
 PCI1510 PCI-CardBus Bridge (0xac56)
 PRO/Wireless LAN 2200BG Mini-PCI Adapter (0x4220)

kamil

leave(1) - remind you when you have to leave

$ leave +0001
Alarm set for Mon Jan 25 21:16. (pid 6979)
Just one more minute!
$

One minute later…

$ Time to leave!

https://www.moritz.systems/bsd-tips-and-tricks/leave/

rvp

kamil Another golden oldie: lock - lock the current terminal until password is input

$ lock -np
lock: /dev/pts/3 on x202e.localdomain.  no timeout.
time now is Mon Jan 25 23:51:20 UTC 2021
Password:

Input password to unlock, or use only -p for a 15 min. timeout.

rvp

JuvenalUrbino Some useful X settings:

Very nice--I use most of 'em myself.

kamil leave(1) - remind you when you have to leave

I'm not saying leave is not useful--glad that some folks still remember it exists, but, I prefer at for this. It's specified by POSIX; you can enter arbitrary commands; you can log off the original terminal you set it on--or even reboot--and come back elsewhere and it'll still email you; you can queue multiple reminders:

$ at 19:30
echo 'Enough! Time to go home.'
Ctrl-D
Job 87 will be executed using /bin/sh
$ atq
Date                 Owner              Queue    Job
19:00:00 2021-01-31  rvp                c        87
$

JuvenalUrbino

Some useful X settings:

xset s off                                  # disable saver (using xlock) 
xset +dpms                                  # powersaving
xset dpms 1200 1500 1800                    # screen blanking
xset b 0                                    # mute bell
xset m 55/20 4                              # mouse acceleration 
xset r rate 250 40                          # faster keypresses
xset +fp  ${HOME}/.local/share/fonts/       # user fontpath
xhost +local:root > /dev/null 2>&1          # privileged X server access 
xsetroot -cursor_name top_left_arrow        # left pointer as cursor
xsetroot -mod 16 16 -fg steelblue -bg black # grid background 
setxkbmap -model pc105 -layout "it,us" -option terminate:ctrl_alt_bksp

« Previous Page Next Page »