Huh. Just updated an old VM from 12.1-Rp13 to 13.2-Rp4 and there is no longer a /bootpool/boot

That's nice.

Hrm. Looks like the zfs import for /bootpool was needed. Interesting.

poudriere is building llvm13 and 15 at the same time intituitively wouldnt doing one after the other be faster?

poudriere is designed to build lots of things in parallel

there's an option to do priority boosts for specific packages, which some use to ensure that toolchains et al are built first, and you can also set certain packages to be built using threads which helps when dealing with said toolchains

ek your bootpool is separate than the OS?

or you can use the seeded package features, which grabs dependencies from the package servers, so you don't have to build toolchains

all of the above is documented in poudriere.conf or the manual pages

anyone familiar with the current state of mono.. trying to port openra's latest update, but hitting into some weird problems with mono trying to build the project.

i have no idea how to fix sound now

any idea>

?

hmm, it'd be nice if you could have pkg install just the UID/GID entries for a package

like if you want to run a webserver in a jail, you'll still want the host to have www:www even if you don't want the package installed there

Freaky: why?

didn't I just explain that in the second line

save manual faffing about with pw or installing and removing the package in the host

I don't understand why you want to have the user on the host

numeric user ids suck and I never want to see or use them?

quick, what user is 143? 144? which one is the 'null' user for that service?

Freaky the right solution is to make libc understand jail, and get that info from a jail instead

for user 1000 on host is Freaky , but 1000 in Jail Foo is antranigv , while 1000 in Jail Bar is meena

it's hard to solve this properly.

I would solve that with a great deal of fire

for example, I have a PoC of ps that jexecs into a jail, gets the name and passes it to the host

however, I hated that implementation because 1) threads and forks are expensive 2) if you run DTrace on the jail, you will see some exec-like operations

I'd quite like jails to share the user namespace, I suppose LDAP is made for that sort of thing

Freaky yes, that's what I do

as a matter of fact I am migrating our last NIS infrastructure to LDAP right now :))

antranigv: how bad is NIS?

meena better than everything else

it easier to debug, it simpler to manager, it nicer than AD, it's more documented than LDAP

then why are you switching to LDAP?

because we're expanding our usage of directories beyond user/group/machine, which is something that NIS was not made for

meena also, I totally forgot, one thing I personally didn't like about NIS is the lack of caching. somehow these days we have faster but less reliable networks.

antranigv: what about nscd?

is there a way to install more minimal binary distribution of mc than mc-nox11? perl and python surely aren't mandatory for this simple file manager to run

the port says it needs them, if it doesn't then that's a problem with the port

RhodiumToad, there is a problem probably ;) this tiny tool should not have over 200MB dependencies (should be 30-50 if not less)

so, remove perl5 and python from the USES list in the port and verify that it still builds

if it does, then report it as a bug

there are also options you can use if it needs them to build but not at runtime

meena actually, we've never used nscd, that might be a good idea for caching!

RhodiumToad, I'm trying to avoid building because of extremely slow hardware (this is why I'm installing everything possible in binary form)

by extremely slow I mean oldest version of Raspberry Pi

I have plenty of disk space (32GB, fastest Sandisk card I was able to get my hands on) but I'm severely limited by other resources (painfully slow CPU and 256MB RAM)

if you're going to claim that something isn't required, then you have to demonstrate that it works

I've built ports on an RPI2B, it's not that painful

but cross-building for RPI2 using qemu-user-static and native-xtools is pretty easy

RhodiumToad, ok I'll demonstrate: mc does not require neither Perl nor Python on Debian GNU/Linux and Solaris

does it have features that are only available with perl or python?

RhodiumToad, it needs Perl at build time (but not at runtime)

it comes with a few "external filesystem" scripts (like for S3 storage) that are written in python or perl

then there should be an option to turn those off

grawity, oh ok :( so I guess compiling by hand is only way if I'd like to save disk space (and it's not worth the effort afaics)

thanks for help RhodiumToad and grawity

Archlinux marks several dependencies as optional: archlinux.org/packages/extra/x86_64/mc

Title: Arch Linux - mc 4.8.30-1 (x86_64)

Title: PKGBUILD · main · Arch Linux / Packaging / Packages / mc · GitLab

hi kids

should git die with "fatal: index-pack failed" error when trying to sync ports?

it always fails after downloading around 180MB

this is how I run git: git clone git.FreeBSD.org/ports.git -b 2023Q4 /usr/ports

Title: ports - FreeBSD ports tree

hey there, I see OpenZFS 2.2.0 came out yesterday, seems great

how would someone go about testing it on FreeBSD 13.2?

veg: freshports.org/search.php?query=openzfs

don't look very upto date tho

anyone know how to see which PCI devices are in which IOMMU groups? lspci appears to only support this on Linux

remexre: the FreeBSD equivalent of lspci is pciconf

meena: I'm not seeing an option in the manpage to show IOMMU groups; can it do this?

remexre, Try with the -lev options: pciconf -lev

I don't know anything about IOMMU groups though. Just how to list out the PCI bus with the above.

rwp: yeah, I'm not seeing anything about them there either

there's also devinfo…

i have a service running in a virtualbox vm on a host-only adapter (vboxnet0) with the ip 10.127.1.11. how can i tell pf to redirect incoming traffic on port 8080 to the vm?

i thought it should look like this, but it's not working: rdr pass on $home_if proto tcp from any to $home_if port 8080 -> $vbox_vm_ip port 8080

meena: not up to date indeed, hence my question

veg: you can open up a bug report to update the port

meena: thanks for the suggestion, that makes sense

or I could wait for FreeBSD 14? ;)

that being said, is there a way to know which OpenZFS version is planned to be shipped with 14.0?

veg: if you want the latest and hottest ZFS, then you need to stick with CURRENT :P

veg: there's no update happening before the ongoing release. We're at BETA5! Unless someone finds huge gaping holes

My file system is one of the things that I specifically do not chase the bleeding edge about. Is there a specific feature you are looking to have available in the new version?

chasing the bleeding edge in your file system is one of those things that is called playing stupid games to win stupid prizes :)

quite a few, rwp

- Block cloning (#13392)

- Corrective "zfs receive" (#9372)

- Prefetch improvements (#13452 is the most interesting to me)

and the rest seems very interesting too

I really don't feel OpenZFS is hellbent on pushing new features without rigorous testing

Title: FreeBSD / src - FreshBSD

look at the raidz expansion patch which has been delayed for 2 years

vs freshbsd.org/freebsd/src/branch/releng/14.0

Title: FreeBSD / src - FreshBSD

zfs: merge openzfs/zfs at 95785196f (zfs-2.2-release) into stable/14 <- thanks \o/

so we'll see that in the next BETA / first RC

meena: AFAIK first RC is out

never mind, we're already at RC1, so we'll see that at RC2.

sounds like 2.2.0 will be part of 14.0-RELEASE which is good enough for me :)