meena it needing root and writing to /usr/obj won't do anything to the system that's currently running, the host system, right?

and which dir has release/?

meena, That deployment article is pretty interesting. :-)

polyex, Another article you might find inspirational is forums.freebsd.org/threads/howto-freebsd-zfs-madness.31662 see the posting "6.3. Import Boot Environment from Other Machine".

Title: HOWTO: FreeBSD ZFS Madness | The FreeBSD Forums

pkg-provides came through in a pinch. self-built liquidsoap needed a particular lib. found one as part of another package. bam

i need to learn how to just get a hold of that single file and put it where freebsd expects it to be... still.. it's a start

meena k i made the freebsd repo. make -j8 buildworld buildkernel KERNELCONF=GENERIC and i looked in release/scripts like you said but it doesn't have mkisoimages.sh like the downloadable .iso has. what i'm doing wrong pls?

oh wow it's in release/amd64!!

wait i cloned the repo, built it, but there's no /usr dir. only /usr.bin and /usr.sbin. where am i going to put my distribution.txz for scripted bsdinstall from this data then?

can't run release/amd64/mkisoimages.sh because it complains boot/loader.efi no such file or dir

Is anyone familiar with www/rubygems-rails? Hoping someone can point me to where the $PATH environment is configured for gems, e.g. where PATH=/usr/local/lib/ruby/gems/3.1/bin:/sbin:/bin:/usr/sbin:/usr/bin set?

After some analysis it looks like my question has now transformed: What is the recommended way to set $PATH for a service/all services (e.g. nginx)? Specifically to add /usr/local/sbin:/usr/local/bin

/quit

batrachian: you can set it in login.conf, but they don't all run under the same User/Group/Class, so i don't really know

i am mostly used to using ubuntu for development. is there a good blog post/other resource on migrating from ubuntu to freebsd?

meena, I don't think login.conf applies, most/all services are configured to run as users configured as "nologin". e.g. nginx runs as user 'www' by default.

razetime: I recommend reading the FreeBSD Handbook. Don't try do connect Ubuntu with FreeBSD. It'll be much easier that way

razetime: the handbook isn't that big btw. it sounds more daunting than it is :]

i'll go through it and come back with questions

batrachian: man-dev.freebsd.org/rc.subr.8 see ${name}_limits and ${name}_login_class

Title: rc.subr(8) - FreeBSD Manual Pages (Dev)

batrachian: and if reading that doesn't convince you, you can also read /etc/rc.subr itself, and cry yourself to sleep

that's what I do anyway, every time i have to read that code

batrachian: nologin is just set as program login executes by default when you login as that user, I would hope that login.conf applies to services as it is the way to configure stuff for users (not sure if I ever want to look into rc.subr)

meena, nimaje ok, mea culpa. I can see on this system one daemon account using login.conf class 'daemon'. So hopefully the ~/.login.conf will also work for 'nologin' accounts using the home directory set in master.passwd?

well, that is often /nonexistent as services shouldn't write to their "home"

Hmm, not going to work by the look of it. Just saw the comments in /etc/login.conf above the daemon class entry, basically says path will be clobbered by rc, service and cron.

i'm trying to go the most direct path from cloning freebsd-src to a scripted bsdinstall iso. so i make buildworld/kernel, but there's no etc/usr/ dir to copy my distributable.txz into usr/freebsd-dist/. and release/amd64/mkisoimages.sh errors boot/loader.efi no such file or dir. what's wrong?

So far then the only option is to export PATH in the /usr/local/etc/rc.d/ script.

hm, rc and service hopefully get the default from login.conf and then only add their additional stuff, afaik in cron you can only set it, but hopefully it gets the default from login.conf as well, so that not setting it in crontab would work (well, but that's just what I hope, no idea how it is implemented)

man 8 service has it explicitly in ENVIRONMENT that HOME and PATH are set to what they are at boot time as per /etc/rc, which for PATH is without /usr/local/

yeah, the rc.d script itself, but for the command run hopefully the settings in login.conf apply

Ok, could be, I'll give it a whirl!

batrachian: i think you'll have to modify /etc/login.conf for this stuff

I have a very early boot rc.d script that updates rc.conf settings

is there any way to tell rc to reload its state (settings/vars) mid-run?

currently we're using an rc.local hack

yes!

I documented it!

oohhhh

where? where the heck is it

dch: freshbsd.org/freebsd/src/commit/098…75c75d5678172eb9dd8fbc89ef61c05c979

Title: FreeBSD / src / 0981275 / rc(8): document how to reload rc's configuration - FreshBSD

SIGALRM

kill -SIGALRM $RC_PID

see also, github.com/canonical/cloud-init/blo…loudinit/distros/freebsd.py#L40-L52

Title: cloud-init/cloudinit/distros/freebsd.py at abbcfc00f3ee6c65b43fbb73198f2cf9f52054a1 · canonical/cloud-init · GitHub

thanks!

used here: github.com/canonical/cloud-init/blo…cloudinit/config/cc_rsyslog.py#L374

Title: cloud-init/cloudinit/config/cc_rsyslog.py at abbcfc00f3ee6c65b43fbb73198f2cf9f52054a1 · canonical/cloud-init · GitHub

that is indeed a useful feature

glad I asked about it instead of just going lol watever

this says MFC after 1 week, but i doesn't look like it's been MFC'd

freshbsd.org/freebsd/src/commit/098…75c75d5678172eb9dd8fbc89ef61c05c979 maybe someone who still has a commit bit could push that down to stable branches

Title: FreeBSD / src / 0981275 / rc(8): document how to reload rc's configuration - FreshBSD

Cuz that "feature" has been around for a long ass time.

debdrup: are you able to MFC that ^^^ ?

dch: sorry, I really don't have the bandwidth right now :(

debdrup: thats ofc fine!

Also, we're in the middle of a major release, so I'm not sure now would be the best time. I can't remember the exact points for the un-freezes, but I wouldn't wanna step on any toes.

given the time of commit, it should already be in 14, just needs to go down to 13 and 12

fair enough

meena: btw this works. very handy indeed.

dch: of course it does.

dch: how MFC my commits!

what about tea

actually, never mind, don't throw tea at me

too late!

I hope it was iced tea

great, now all we need is some cooking skills

Lots of EC2 related commits in the releng/14.0 branch. Is FreeBSD making into an AWS EC2 AMI?

freebsd has had amis for years

raym: it's been there forever. but we're getting better support

I see

Hey, I'm pretty new to jails and I'm trying to install bind to a jail. I'm using nullfs so I'm not sure what the issue actually is?

This is my jail.conf: bsd.to/bJAz/raw

Title: bJAz

samip537[l]: what's the problem?

Hi. I am happily using the i915kms drivers for my intel/nvidia setup. Now trying to install a software that requires the nvidia-drivers package installed and the module loaded. Will this two interfere with each other? Is it a bad idea?

how can I tell which driver is providing the virtual terminals? none of syscons, sc, or vt show up in dmesg

sysctl kern.vty

thanks

for some reason I got it in my head that that wasn't definitive, but re-reading the man pages it seems that wasn't too accurate

search for VT in /var/run/dmesg.boot, too

"VT(efifb): resolution 640x480" ... "VT: Replacing driver "efifb" with new "fb"."

VT doesn't mean the vt driver though, right? it just means virtual terminals in general?

no, it means the vt driver in this case

the old syscons driver doesn't do those messages

is there a way to change the video mode of VTs when using the vt driver? vidcontrol doesn't seem to want to work (as suggested in vidcontrol(1) that it might not) but the video mode does change on startup to show 266x75 text so it seems there is some mechanism somewhere to do it

I'm not sure if it matters, but this is while using i915kms.ko and X11 on another terminal

once the drm driver is loaded, the video mode is controlled by that; if you want to change the text size, you do it mainly by changing the vt font size rather than the resolution

is there a man page or something for i915kms.ko? I didn't find anything

I'm running xfce (on a Pi4 if that matters). How can I allow the user to halt or reboot instead of simply looking out? I'm using startx as I cannot seem to run xdm or any other without the screen going blank.

xdm ought to work if X works at all?

the screen going blank effect also works in VirtualBox, grr... # but I don't need that fixed, I need the Pi users to be able to reboot

RhodiumToad: you would think

members of the operator group can run the shutdown command

oh? lemme try that

(they can also do stuff like read (but not write) raw disks)

the reason why I'm trying to change the mode is because I attached an external monitor to my laptop and now when I switch to a text VT (non X11) the framebuffer doesn't reach the bottom or right side of the screen

presumably because the external monitor has a higher resolution than the internal one

how is X handling it?

for X I just used xrandr to tell it to fix things up. currently I'm just doing xrandr --output eDP-1 --off to disable the internal monitor in X, then restart my window manager to get it to re-read the new dimensions that it should place UI elements, then call the windows back to the visible area, because they would be off the screen by then

I think I could also set the external monitor to --primary and then the later WM steps too, but that's not what I'm doing right now

otherwise X is working fine, at least no worse than without the external monitor

there was one hiccup yesterday when the first VT stopped displaying new characters, but a restart fixed that. I was running that VT as text mode at the time, with X11 started from a different VT

I tried to reset(1) from a different VT but for some reason that gave an error. I forgot which error

I haven't really played with video modes in text

there's probably a better way that involves X starting up with the correct resolution for the external monitor, but I haven't figured that out yet

I think the VT text modes are really only technically video because the drm driver is involved, but maybe I'm wrong

maybe they really are true text modes

the vt driver doesn't use text mode unless you force it do, and you haven't

its intended mode of operation is in a graphics mode, i.e. backed by some kind of framebuffer (whether vga, efi, or drm)

but it would be nice if I could set X11 and the text VTs to use the same resolution so that VT switches are snappier

yeah it did appear to be a graphical mode to me, just by the looks of things

but "text mode" emulated on to of that

*top

emulated in the sense that the vt driver renders characters on the screen using a character-cell font, yes

yes precisely

so basically I have two goals: 1) be able to use all the real estate available on my external monitor when in text VT modes, and 2) use the same resolution for X11 and VT text modes for faster VT switching. (2) is really just a nice-to-have though

so, if I'm in the operator group, why don't I have permission to run /sbin/reboot ? I'm also in wheel

operators have to use shutdown -r

/sbin/reboot can use dangerous options

(and hence is reserved to root)

being in wheel just allows you to su to root if you know the root pw (or have configured sudo to allow it)

OK, that works, thanks

good enough, thanks... they can run startx, and either in an exterm or after they log out they can run my aliased fore reboot and halt, problem solved... I can deploy the Pi Monday :-)

man, sorry for the typos, I'm leaning over and juggling two keyboards :-/

we all speak Typo fluently

Anyone happen to have an newer (like 10 years?) nvidia card on a system where X isn't loaded? Does loading nvidia-modeset load move nvidia klds?

nvidia-modeset.ko loads nvidia.ko but that's all, as far as I know - but that's not the newest card or driver

nvidia-driver-470 only has those two .kos

my card is a GT 710

Ok... someone is claiming their card is running at high temps until X is loaded.

I'd assume just have the kld loaded would hand any GPU temp management.

Hard to know. the nvidia-driver package also includes user-mode code for X

xorg/modules/drivers/nvidia_drv.so and the nvidia versions of the GL libs

I see nvidia-settings has fan control so maybe... Never seen this in Linux and sounds like a bug in the FreeBSD driver.

However, it seems the FreeBSD driver forum on nvidia devel is more or less user posts... user solves.

Bit of a different exp. with the Linux side of the forum.

Hello. How do I use lurch?

i did pkg install lurch

it seems to be a plugin for libpurple?

RhodiumToad: that should give me an ability to use OMEMO

V-T60: does pkg query %Fp lurch indicate any installed documentation?

Are there ever situations anymore where `freebsd-update install` needs to be run in single user mode?

Wait, is "freebsd-update install" supposed to be run in single user mode? (All these times, since 12.x, have not been doing that)

No mention of "single user" in man.freebsd.org/cgi/man.cgi?query=f…+and+Ports&arch=default&format=html

Title: freebsd-update

hmm

mount --libxo=json and =xml appear to have issues with non-unicode mountpoints

Freaky: that makes sense, at least for JSON: "A string is a sequence of zero or more Unicode characters, wrapped in double quotes, using backslash escapes." — json.org/json-en.html

Title: JSON

which makes it rather unsuitable for encoding paths

❯ mount --libxo=json |jq

jq: parse error: Invalid string: control characters from U+0000 through U+001F must be escaped at line 1, column 11694

XML mode mangles it instead

and XML should just be using <?xml… encoding="blahblah"?>

XML replaces my dodgy bytes with a question mark

So, basically, libxo has to, effectively convert non-Unicode to Unicode, and for XML it needs to specify the correct encoding

does just happen with mount, or do you have other ways of reproducing it?

what's your locale?

❯ procstat --libxo=json pwdx $$ |jq

jq: parse error: Invalid string: control characters from U+0000 through U+001F must be escaped at line 1, column 105

en_GB.UTF-8, not that I can imagine it possibly matters

"/tmp/bork/foo\nbar\tbaz\xc3\x28boop" is my mountpoint

I'd quite like to be able to reliably extract that from.. something

Linux mountinfo encodes it

what encoding is "/tmp/bork/foo\nbar\tbaz\xc3\x28boop" ?

it's unix filename encoding

a bag of bytes that don't contain NULL

what does it expand to?

Also, it looks fairly Unicode to me

it's very not

"/tmp/bork/foo\nbar\tbaz\xc3\x28boop".scrub # => "/tmp/bork/foo\nbar\tbaz�(boop"

that's literally what I'm testing

I could have just grabbed some /dev/random output and had a reasonable chance of it being a valid path

is anyone still using an actual phys terminal? I am sort of looking into getting one just for the fun of it

am mostly leaning towards the DEC vt520 color - because it makes it possible to do away with the crt and attach a normal flatscreen vga monitor

phys terminals <3

sure - but you can say that about a lot of old systems ;-)

the server I currently have has an ilo - ssh capable so no real need for it

i used to have a wyse terminal,

it sure beat hyperterm for what i was doing with it

they where also good - and a cheaper alternative to a real DEC

probably still are

they must have made millions of them, to every bank, supermarket, library etc etc - but a refurbed one sure costs premium today

not to mention a keyboard to match

any Makefile ninjas are awake? :P

cdrmack: what is your question?

sometimes I'm scared that we don't actually need build systems, it's just the matrix keeping us busy

first of all it's not ports related, just regular makefile question

I use depend target to call mkdep to generate .depend file with my *.c <-> *.h relations

however when I modify *.h file make doesn't recompile related files even though these are visible inside .depend

so I wonder what I'm doing wrong

are you using bmake or some other make?

just a regular make, the one which is on FreeBSD out of the box

do you include that .depend file?

make is supposed to read it automatically according to the manpage

nimaje, according to the man pages makefile looks for this file automatically, default name is .depend

yup

ah, ok

what does the output of make -dp look like

should I submit it somewhere?

any paste site, e.g. dpaste.org

or bsd.to if that's up

or you can do make -dp 2>&1 | nc termbin.com 9999

either way you have to tell us the resulting url

need to run bash, 2>&1 doesn't work on tcsh it seems

csh / tcsh is horribly lame when it comes to redirections

nah, on bash it stores only the last line too

huh?

one sec

anyway, make -dp |& nc termbin.com 9999 might work in csh

apologies, doing my best but I'm newbie

Makefile: bsd.to/wV0h

Title: dpaste/wV0h (Plain Text)

Folder structure is really simple (I'm learning). main.c, foo.c and foo.h

ok, so it definitely reads the .depend

now try make -dm

touch the .h file first, ideally

it's -dm after touch .h

what's the Makefile?

I sent the link above, on bsd.to

Title: dpaste/wV0h (Plain Text)

using find to populate $(SRCS) is kind of bad practice, but I don't think it should cause this

I have tried to find a better way but failed. I will need sth like that when I will introduce some subdirectories for src files.

what's strange is

that even when I add foo.c : foo.h

it won't detect the change and recompile

:|

got it

the problem is that your SRCS contains ./foo.c and not foo.c

och, so that shell cmd was problematic after all

when you have subdirs it's usual to have a file in each subdir which is a makefile fragment which contains info about the files in that subdir

any recommendations how to scan for all src files? I searched for few hours today and that was the best - not GNU Make - way

how do I add all *.c files in the current dir though?

by hand?

if you insist on using find, do it something like this:

RAWSRCS!= find ${SRCDIR} -maxdepth 1 -name '*.c'

SRCS= ${RAWSRCS:S:^./::}

but it is conventional to list all the source files explicitly in the makefile, not search for them

gotcha, thanks a lot for you time

and help

since I'm learning I want to follow the best practices, thanks for sharing

to sum up - one Makefile per directory and add files manually, right?

just don't copy ideas from the src tree or the ports makefiles, those are all horrible

:D

avoid recursive make if you can.

do you know where I can read more about what you suggested? having parts of makefile in subdirs?

note that you can .include other files in the makefile, and you can even do that in a loop over your subdirs

one more question and I'm done for today, what's your opinion about mkdep? i've read in few places that ppl prefer calling cc with -MMD to generate these dependencies and then .include inside Makefile

there was a "recursive make considered harmful" article kicking around a long time ago, it might still exist, but iirc it may be specific to GNU make

I see

I've generally preferred the compiler option, but it may be more work

okay

thanks once again

have a good day/night!