Feels a little like overkill to run one program in a jail, but I added memory recently…

It's only files on disk, right? If it is really just one program then the memory usage won't be any different than if it were not in a jail.

And if the files on disk were clones of a snapshot then there wouldn't be significantly more disk space used either.

Nah, it's tiny.

Im kinda, stuck where i am with the mpt0 issue, not a lot of people having the same issue and i cant seem to really figure out.. why its happening? So, i have a plan

im gonna switch over to ubuntu again, i have a spare disk so i can just pause freebsd basically, then ill see if the wiring and the drive and everything work still. See if i can narrow it down to just the mpt driver, or my hardware failed just now

hi, I have a weird case of stuck src build of databases/mysql-server80 on 13.1 releng. It consistently hangs on what seems to be the last file in the build, for at least 4 hours. pastebin.com/KJf2935g

Title: PID USERNAME THR PRI NICE SIZE RES STATE C TIME WCPU COMMAND - Pastebin.com

Do you see no CPU or IO activity for the compilation? Have you checked whatever logs that might be generated in the port build directory?

the build was using 2 jobs but is now down to just one cc job burning 1 core at 100%. there is no activity, the build disk has spun down in the meantime. if it's actually calculating anything, it's being done in ram

the file itself is a tiny helper file, so there is no reason why it should be taking so long. Unless it's doing some additional work while hiding behind that commandline

exit

ultramage: did you enable any extra optimisations?

ultramage: mimicking what the others have asked plus did pressing ctrl+t produce any info whilst the cc appears to hang ?

didn't we have a wiki page detailing the differences between MacOS and FreeBSD? I can't seem to find it

developer.apple.com/library/archive…/doc/uid/TP30000905-CH214-TPXREF101 i don't think anyone tested this website on anything other than safari

Title: BSD Overview

meena: works_for_me(TM) (on macos using microsoft edge :-D)

otis: maybe it's just a Firefox mobile thing then

in the grand scheme of Semver, is 13.1 a major or a minor release?

Yes.

it's been at 100% for 12 hours now, I can safely call it stuck now. I haven't dealt with a malfunctioning compiler like this before, so I'm not sure how to go about it.

my only tweaks are CPUTYPE?=native and MAKE_JOBS_NUMBER=2. same with the kernel/world build. My only changes to the port makefile were switching libicu dependency from ports to bundled.

mi_switch+0xc2 sleepq_catch_signals+0x2e3 sleepq_timedwait_sig+0x13 _cv_timedwait_sig_sbt+0x10f seltdwait+0x75 kern_poll+0x52b sys_poll+0x51 amd64_syscall+0x112 fast_syscall_common+0xf8

I can break in with gdb if needed

I remember that when this happened in the previous attempt, I ctrl+c'd and tried resuming the build. Then it would hang on some other file. I will try that again if this attempt goes nowhere

hm so the syscall is [select]. So I guess it's busy-waiting on a pipe?

oh that's printing the status of the 'make' process that's waiting on the foreground. the background 'c++' process looks like this pastebin.com/Qj9trzaE

Title: [Switching to LWP 100269 of process 38470]0x00000000044230c7 in operator!= () - Pastebin.com

Is there any way to have persistent shell command history (i.e., not just for the current session) when using /bin/sh as root's shell?

runOnMachineFunction -> allocatePhysRegs -> selectOrSplit -> selectOrSplitImpl -> tryLastChanceRecoloring -> tryRecoloringCandidates is the common part on all the call stacks I paused on. The levels above differ every time. I sometimes see that it's recursing.

msiism, if you check the file on your home dir .sh_history

you'll find it there.

Unfortunately, I don't have that file. Also, sh(1) doesn't mention it, as far as I can see.

you can see it with ls -a

is a hidden file.

Yeah, but it's not there.

I mean, sh(1) should mention that file if that is where histroy is stored, right?

I have it on my dir.

Okay, what shell are you using?

sh

Like /bin/sh?

yes

Weird.

Helper: What version of FreeBSD are you running?

current

Oh, as it seems, I'll have to use fc to get persistent history.

freshbsd.org/freebsd/src/commit/c4a…2633d975bd0813afca6b8e23ead29d80e82 you know, i would've expected a bigger difference

Title: FreeBSD / src / c4a4b2633d975bd0813afca6b8e23ead29d80e82 - FreshBSD

Well, as far as I can see here on 13.1, /bin/sh just doesn't offer persistent command history, at leats not out of the box.

debdrup: that's a one line change

meena: you comment remind me of this ftp.openbsd.org/pub/OpenBSD/songs/song62.ogg

So, my machine just rebooted because of a kernel panic. It also told me where the relevant log file is. But then motd used most of the screen and the info was gone.

Someting-something-core.txt, if I remember right.

msiism: /var/crash ?

Let's see.

Seems like it. However, the file just tells me the system is unable to find a kernel debugger.

So, I should probably install gdb then.

msiism: also, dmesg or /var/log/messages might have the info you were denied

Title: Chapter 10. Kernel Debugging | FreeBSD Documentation Portal

Yeah, /var/log/messages has it.

"reboot after panic: vm_fault_lookup: fault on nofault entry, addr: 0xfffffe0097be7000"

So, I'm not sure what to make of this. Could this be because of broken RAM?

Good day,

Would someone know which define in C can be used to differentiate FreeBSD, for example #ifdef __FreeBSD__ ?

Thanks

Title: Which preprocessor symbol to check to detect FreeBSD - Stack Overflow

meena: Is there an already existing equivalent to `pkg which` (or `pkg provides`) for base packages?

meena: Currently, I'm using a separate tool that greps the archives for file names.

_xor: not that I'm aware of. I usually have a system somewhere lying around with the full installation and query that

less ideal with CURRENT where shit shifts on a weekly basis somee

Thank you

msiism: we don't use gdb anymore

meena: How are you querying it?

oh, with pkg-which after pkgbase packages are installed?

_xor: yes

but i just opened this bug, rosorio/pkg-provides #7

Title: How to provide a provides DB for a PkgBase repo · Issue #7 · rosorio/pkg-provides · GitHub

I've been meaning to self-host the pkg-provides db (which supports doing that, so no changes required there).

It would need to be generated though.

and to think all this time i was just grepping plists

rtprio: Yes, pkg-provides is quite useful :P

I do kind of wish though that pkg natively had a method to run query on remote repos.

...or at least include a pre-generated indexed plist db in the repo that can be optionally fetched to do local queries on not-yet-installed packages.

github.com/rosorio/pkg-provides/blob/master/pkg-provides.8 i don't see a way to generate the db

Title: pkg-provides/pkg-provides.8 at master · rosorio/pkg-provides · GitHub

Hang on, I have it somewhere I think.

Link to the source that does that.

Hmm, apparently not. I guess I was just downloading the generated db.

maybe poking at the files, we can reverse engineer it, _xor!

meena: It wouldn't be too difficult, given that it's loading it and running a PCRE against the db.

meena: Not sure which fmt it's in, need to look further at the source, but that's not too hard.

meena: I tend to think though that generating a SQLite db in /var/db/pkg/${REPO_NAME}-plist.sqlite is a better approach.

meena: Then FTS capabilities that are built into SQLite can be taken advantage of.

...also, this way makes it possible for it to be per-repo, since currently AFAIK pkg-provides always just looks for and loads /var/db/pkg/provides/provides.db, which makes it inflexible.

I just put it on my secondary TODO list for later to take a gander at whipping something up.

I just got done patching sysutils/podman since one of the dependencies that upstream is pulling isn't written to respect build tags and tries to use cgroups on FreeBSD to get container stats (which obviously fails).

About to test it now and see if it works.

Good lord did that require more patching than I wanted :|

_xor: meanwhile, bapt has revealed to be that it's a locate DB, produced using locate, but not yet how

_xor: we need to port *something like croups* to FreeBSD. I reckon Solaris contracts might be a good candidate, but maybe I'm also talking out of my ass

meena: why is the puppet-caddy module so sad

i want to bring it up to caddy2 and freebsd but it's so... linux-y

rtprio: i have looked at it once and i really don't know where to start. caddy is… weird

it is, but i'd take 2 line vhosts over 15-20 of boilerplate like nginx

hahh

i use the puppetlabs-apache module, with custom vhost only

codeberg.org/pkgbase/puppet-environ…nch/production/data/common.yaml#L21 so this is what my vhost definitions look like

Title: puppet-environment/common.yaml at production - puppet-environment - Codeberg.org

I need to switch to mod_tls

but i don't wanna compile it myself…

rtprio: Lots of projects are unnecessairly Linux-y, unfortunately :/

rtprio: I use templating to produce my configs, and nginx also supports includes natively, so maybe look at that for 2-line vhosts? (I used to do that and it worked fine)

meena: Ah, cool. Didn't know it was a locate db.

If FreeBSD were to get cgroup-like functionality, then I'd say just port cgroups to leverage the large body of existing projects out there that use cgroups, but...I'm not super familiar with cgroups and don't know what it provides over jail-specific namespacing.

_xor: that's difficult because GPL, and because the Linux Kernel is weeeeeeeeird

Freezing/restarting seems to be something cgroups do that jails don't.

Hmm, I was expecting cgroups to be more granular than jails, but so far it doesn't exactly seem like it.

Was just looking at the isolation capabilities offered by cgroups and as I was going down the list I was expecting to hit something that jails don't do, but nothing so far?

i used to, but they were still so verbose

and caddy handles letsencrypt for me

That was the one thing I didn't like about docker when I first looked at it years ago, was the lack of namespacing for UIDs/GIDs. Then later it was bolted by mapping from host to container (e.g. host 1001 -> container 1001001, etc).

I think jails got that right, isolating ownership and membership as well.

_xor: this vaporware lives in my head for years now, and describes what i want more from jails / FreeBSD: scratchpad.pkgbase.live/Tn-dtb8OT--CttN1Djn8lw#

Title: jrc: Principles - HedgeDoc

rtprio: Yeah that's fair, I've heard caddy does that.

I use mod_md for that

Holy shit it really has been two years

it's also weird that no linux seems to packag caddy, for reasons i have not entirely been able to find out

Available via Homebrew

meena, Are accounting & resouce control easier to apply to a jail than an account?

...s/account/normal user account/

Same mechanism, different subject.