kevans: no no, why would you have 13,14,15 kernels with the same rootfs

rtprio: oh, well, good reasons for that sometimes, too. :-) admittedly, the last time I had to it was rerally just bisecting for a kernel bug over a few releases

much easier to just pin userland down low

there is something really weird with jails and IPv6, few weeks ago I spent all night debugging it to find out that when I reloaded the firewall it just worked randomly, now with the same firewall and haven't touched it since, IPv6 now randomly stops workign

ndp is working, LLA are working, firewall is passing quick on icmpv6 as the first rule

the *only* thing I can think it could be is the fact I chopped up a /64 into smaller blocks which is not technically standard

but has never caused me issues in the past on other platforms

only SLAAC should stop working if you make a smaller subnet then a /64, if you assign addresses in another way, then any subnet size should be fine

yeah thats what I thought

and no SLAAC isn't used, just chopping up /64 into smaller blocks for lots of jails :)

but something is clearly broken, I think it might be OpenBSD which is the issue (running on the router)

its not responding to NDP for one of the other devices, I can see there is NDP being broadcasting asking who is holding the LLA of the router, and never getting a reply

maybe its address conflict...

but I doubt it

another OpenBSD host keeps spamming the broadcast with NDP, despite the fact that the OpenBSD router can actually ping it, so I am going to take out the other OpenBSD host on the vlan and maybe that might be the problem

this is what happens when you try to use an entirely BSD stack I guess :p

or maybe im just too stupid to do it properly who knows :p

what the fuck: bpa.st/Y23Q

this was literally working a few weeks ago

oh wait, the host ndp cant see the address

I figured out the problem :p

iirc to route packets you cant have the same subnet on different interfaces

wait no

it's anycast

what I assumed the problem was is that I am routing packets over LLA but LLA doesnt care about anything else

the inet6 2001:8b0:57a:2391::1:4 prefixlen 128

routed over LLA

afaik this *should* work

but the host is unable to find the address

LLA - like link-local address ?

although its also weird that the host is retrasmitting the packet 50 times until TTL is exceeded and *then* dropping the packet

mzar: yes

LLA communication works fine :)

do you have any ip6 firewall ?

in fact the packet is routed both on the host, and the router as you get the response from google which is what im pinging to test, but the response is then dropped on the host as it cant see the address

mzar: I am passing quick icmpv6

do all icmp6 pass ?

pl

also this is not a firewall issue

ok

as I said

ndp cant see the address

which means there needs to be a route

have an efficient troubleshooting polarian, good luck !

bbl

I think I just rubberducked this by spamming the channel :p

I bet there is no route and thats why

2001:8b0:57a:2391::1:4 link#6 UHS lo0

oh there is a route, but its loopback, and that explains why the packet is being looped 50 times until dropped due to ttl-

oh shit no thats on the container whoops mixed up the terminals :p

okay this is a routing table issue.

Question I asked a while back though, how do I specify multiple ifconfig commands in rc.conf, I cant seem to assign an IPv6 address AND set a static local link to say fe80::2

fe80::2/64

netif complains its an invalid format

if I use the same ifconfig_<iface> string, and you cant set two of them, it will just execute the last one and ignore the one before it

any recommendations for kvm-over-ip hardware (either dongle or appliance) with solid edid and usb emulation, bios control, and virtual media?

also 2560x1440 rez

kvm-over-ip with 2560x1440 rez ?

promising setup

what do you mean by promising?

i found GL.iNet Comet Pro

has that feature and pretty decent price

mzar

ok

cool great contribution

i guess your job around here is to just dump on ideas then get proven wrong immediately after?

I am not against new ideas

still lagged behind, though

ok

tryin to update jails with bastille.

yet after the steps the jail is still the old release in `bastille list`

and bastille upgrade says the jail is already running the new release

and also packages within the jail are still at 15.0, not 15.1

its just text, you can update what it says in myJail/jail.conf or via bastille edit myJail

osrelease = 15.1-RELEASE; it alrady seems to have the correct value

what it says when you exec in the jail and run freebsd-version is a better indicator

-r is 15.1 (correct) -u is still 15.0

-k is unable to locate kernel (don't know what that's supposed to read)

did i botch the release?

Attempting to update release: 15.1-RELEASE

pkg: Setting ABI requires setting OSVERSION, guessing the OSVERSION as: 1500000

-r asks the running kernel for its version and as a jail doesn't have its own kernel, it probably results in reporting the version of the host, -k tries to read the version from the kernel on the filesystem, but as a jail doesn't have its own kernel, you probably don't have a kernel installed inside the jail

this sounds weird, that should be 1501 something

i don't care about the kernel as much?

but the "bastille upgrade TARGET 15.2-RELEASE"

start the jail

in the instructions didn't actually do the thing

it still mounts to the old thin release

oh maybe it's because etcupdate doesn't work

There's a bug in bastille. I ran into it as well. It's on their tracker.

so if i update the link in /usr/local/bastille/jails/myjail/fstab will that fix it

Aye, that's all you need.

You might also need to edit the jail.conf as well, if it doesn't show the new release. I needed to do that for one or two jails.

oh, they all were showing the new release

Yeah, it was the same for me.

Just the mount in the fstab was wrong for most of them.

yay, that was it

You're welcome.

i'm finding the bastille commandline rather tedious

i get the concepts but he seemed to overload operators in a nonintuitive way

boru / sbr: should pkg be able to 'see' the freebsd-base repo / try to install packages within the jail

or are those hidden as they're part of the jail

hello people

I'm trying to install freebsd on a laptop using the mini-memstick image, I can get to the menu but when I boot, no matter which option I choose, the menu is still there, and the booting/installation screen is appearing very tiny at the top of the screen, repeated 20 times (so I assume it's something like 80x45 pixels). It's so tiny it's unreadable and I can't use it. Has anyone exprienced

something similar ? Thinkpad T520 using HD graphics

I've ended up in charge of a small FreeBSD installation due to the person in charge entering permanent retirement with bad timing.

I presently suspect most executables were being handled via `freebsd-update` while ports were being done via `portmaster` and `gitup`.

what is the output of `history` as root

Fairly long and several invocations of `portmaster -BDg <pkg>`.

(you don't actually have to send it, you just have to have understood it, because root's history will tell you what the previous admin was doing if they didn't turn it off)

Which is why I saved a copy of .history which has had minimal modification.

good

If git.freebsd.org supports filtering, I suspect `git clone --filter=tree:0` may be preferable to `git clone --depth 1` in section 4.5.1 of the manual.

(alas this was someone older who disliked Git, so I'm worried about whether any packages were modified though `gitup` may handle that)

ah...

I understand the dislike of rename detection, but real world experience suggests it works quite well.

freebsd-update is used for updating the base system (up to the 15 series I believe)

I *think* it was used that way and no customization there, though a customized kernel was built.

ah

The irony is I've been playing with an uncustomized installs and using prebuilt packages, but not doing everything from source.

I've mostly got a plan and I think I know next steps, but I'm nervous.

If you know how FreeBSD generally works, you probably won't break anything too major.

I don't claim that, I've got a decent feel for how most OS setups go, but that means I could severely break things.

what are you trying to do, geometry

Get a package security update done, then I'm thinking about getting things move onto later release of FreeBSD (since this installation is at EoL).

I /think/ the package security update is minor for this installation, but I don't like those being even the slightest bit out of date.

gotcha. I would mainly just be concerned about the kernel. if it has any custom stuff going for hardware or whatever, then make sure to take a note of that

I wouldnt worry much about the pkg upgrades until I did the system upgrade

Looked like a slightly cut-down GENERIC.

I'll strip out Hyper-V since that driver is a pig.

freebsd-update will quite easily update you to the current 15.x branch

then you can handle the packages

Well, the fact they were using portmaster and not pkg makes the assumption that they also needed custom builds for their software. So, you might wanna take a look a the build options for the ports they were installing.

I'd think they would opt for packages if otherwise.

oh yah thats true

It's an easy fix, though. I'd focus on the base/kernel updates first, and the software afterwards.

my own inclination would be to wipe all local packages and upgrade the system. then reset whatever local packages I want and reinstall them

Some of it might also be history, I've got the impression portmaster predates pkg by several years.

You can use "pkg" to upgrade everything and just see what does and does not work. Then rebuild the ports with the custom options (if needed).

I'd also suggest migrating to Poudriere for the custom port builds.

geometry: Yes. Portmaster has been around for a VERY long time.

<3 portmaster

I think moving to prebuilts for most packages would be best, likely there is fairly little customization just the person in charge figured out portmaster when that was the only tool for the job.

at my installation i have dovecot customized

Ah ha, finally found the patches!

:O

geometry: Excellent!

That should make it easier.

Indeed.

nice

ac_laptop: that's very blurry. what is on the top line of the screen, with the blue background

rtprio: I assume it's the installation screen, repeated 20 times.

but it's too small for me to read

but the installer is text, and that still looks like you're in text mode;

if you move the arrows / press enter does it change? what about ^l ?

or alt-f2 to alt-f4 to see if another console comes up

rtprio: Alt-F3 and Alt-F4 indeed show what appears to be an other console

but back to alt-f1 and does it redraw iwth the installer or no

rtprio: yes, Alt-F1 redraws it with the installer

I've been given the impression shallow clones are now sort-of discouraged since filtering works better in most cases.

geometry, Do you have any references for that assertion that git filter is better than shallow clones? Because I don't agree.

Those two things do different things entirely. One is the horizontal. One is the vertical. By analogy anyway.

I prefer shallow.

Well, depending on what I'm doing with that particular ports tree.

... or src tree.

If nothing and just building? Definitely shallow.

Filtering keeps you synchronized with history so you know where you are when you do updates.

Shallow-clone only downloads the data for a single-commit and loses history and therefore ends up with incorrect commit hashes.

Filtering downloads the history, but none of the files/data attached to those commits.

Isn't vertical versus horizontal, more filtering gets you the entire backbone whereas shallow-clone only gets you a single vertebrae.

Yep.

If I'm only pulling/cloning to build, a single, latest commit gets it done just as I need. I don't need a ton of old data.

Filtering or depth'ing works fine for both.

If I have any reason to have previous commit info, I'll pull the full depth or filter for commit info without the full data.

The strength is more being able to cleanly update to a later revision.

Sure. I just generally need commit hashes for a single thing. So, commit hash history is important on dev systems. On non-dev systems, I just need the base --depth=1.

Like, I'll pull an old commit for a single port. I very rarely need to pull an entire, old ports tree.

But, it has happened!

If someone is just cloning to build something then it should be a shallow clone to avoid the long history and bandwidth consumed by it. If someone is filtering then they will get all of that and will not have the files needed to do a build.

In the end it depends upon what you need and what task you are doing. But I would not say that filtering is preferred over shallow clones. Or people will set up CI/CD services using filtering and that will wasteful of bandwidth and server resources.

^ Is this about Git?

(now with mandatory depency on Rust)

the corrosion is spreading??????

kill me now

whatever happened to rust-in-base in regards to fbsd?

Anytime anyone wants to move to fossil then I am supportive. Just say'n! :-)

... isn't that written by christian fundamentalists?

Did we somehow skip to talking about the King James bible when I wasn't looking?

SQLite has some fanciful comments in the code base that has caused this strange meme but let's say it is true. Does it matter?

No it doesn't matter.

Christians are good peeps

Nearly every religious text is good moral principles.

Then everyone spends centuries or millenia arguing and killing each other about what *exactly* was meant.

I don't judge software based on the beliefs of its creator(s). I judge it on its merits and design. I don't care about the politics of it. I might disagree with them. I might dislike them. But I'll still use it.

I have heard this before that because there are fanciful comments in the code "the blessed artifacts" and other such stylized language that this means something other than bored programmers trying to have fun with it. I mean the language is called "fossil" for goodness sakes! That by itself should lead people to think it is the other direction.

i have verses of the quran in my test suites.

this conversation reminds me of the movie "Pi" ..

I saw a report years ago that Lewis Carroll's books about Alice were the most quoted in code comments. This gives me hope when I read an AI generated something with quotes from Holy Grail.

Then he would have verses of the talmud if it was Pi

Great movie btw

mason, I never figured out what caused freebsd-update to not add the new file sshd-auth (maybe because I had modified the sshd_config?) so in the end I created a new empty Boot Environment, installed a fresh pristine 15.1-RELEASE there, installed the configuration, activated it, booted it, and voila it is upgraded. I might just do that for upgrades from here on forward. It was simple.

The use and coding of computers are esoteric... Look at old games like World Of Warcraft, there was biblical and kabbalistic stuff woven through the stories

Loads of symbolism everywhere

Why do you think there's a bite out of the apple on the Macintosh?

/\

/ o \

/ \

Freemason btw, tubal cain apprentices

because apples are tasty

Keeps doctors away?

Adam and Eve

Forbidden fruit

objectivist: Based on Jobs' biography, I don't think that is accurate.

But we may also be leaning a bit #freebsd-social at this point.

objectivist: that kinda thinking is not unique or interesting, go to any in-patient mental health clinic and thats all you'll hear, or variations of it, maybe your version is a bit more coherent. its not a healthy way to think. i cant avoid thinking this way, but wouldnt recommend it.