I just checked Chromium on my Mac (I build it from the Git repo about once a week). There are two AI features currently: 1) History search and 2) Real-time suggestions (I've disabled both).

yikes!

theres some kind of bin file isnt there, the name of it escapes me

Schamschula: tech2geek.net/google-chrome-is-secr…-a-4gb-ai-file-heres-how-to-stop-it

Yep

89% 23h10m

f451: what is that a stat of?

building qt6-webengine

how many hours left to go

id guess 1-2 hrs

Godspeed

;)

it is huge

cpu is E5-2690 x2 @3ghz ncpu=20

E5-2690v2 *2

going to need a project to do on one of my systems, just not sure as to what...

They're just desktops, no need for server stuff

Although, I do plan on taking network+ in a bit, so I should probably set things up

Sounds like you should repurpose them as servers and start selfhosting services

Thinking that might be the way to do things

this is the way

been wanting to actually run a mock website to run python stuff

Guess that will be the week coming up

my office servers are ex-datacentre gear

all the pi1/2/3/4/5 are servers

actually to get an accurate estimate of cost id need to subtract avg wattage over time 247 from avg wattage over time qt6-webengine

for the latter its 296W 1hr avg

think it idles at 110

If you pay for heating, you can reduce the costs from there

yeah heating is turned off even in winter

in the office

we have a bit of a spring heatwave rn. got to 30degC indoors, 31 outdoord. rh is a bit better today, 45%

its 15degC rn

it's been fairly cool here, near freezing every morning about 5C

honestly haven't packed away the sweaters just yet

Thanks for the tips guys, the goal from network+ is to get into CCNA

best of luck with your build, take care

ty

after 1day 7 mins it failed. ah well

I'm trying to use ffmpeg's kmsgrab, which "Requires either DRM master or CAP_SYS_ADMIN to run". is there any way to manage this on FreeBSD without running ffmpeg as root?

FreeBSD doesn't seem to have setcap at all

my user is in the video group, which seems to have rw access on the /dev/drm nodes that /dev/dri points to

tm512 run it under ktrace and see what fails and why.

I've got ktrace.out, but now what? it seems to be a binary file

man kdump

ktrace makes a kernel trace, and kdump reads the dump details out for inspection

What you will see is a step by step list of the things ffmpeg asked the kernel for.

These will include return values and errors.

the only thing I'm seeing so far aside from logging output is an openat() call for /dev/dri/card0 which returns 8, and then a bunch of following ioctl calls that all return 0

I'd expect you will find ffmpeg tries to open() some device. If you find an open() with an error, EACCES (reference github.com/freebsd/freebsd-src/blob/main/sys/sys/errno.h) that's probably the blocker.

returning zero is conventionally success for most functions, but open() style calls return a file handle number. You'll see that 8 used as arguments for read() or similar afterwards.... fyi.

If you don't see any errors that mean you don't have permission then its unlikely permission is preventing your frame grabbing and you'll need to look further.

sorry for the double-negative.

I don't get this error if I run ffmpeg as root, so it seemingly is permissions-related

I may have assumed from your question that you had already established permission errors as a fact.

Did kdump reveal any errors?

not that I see

If that's the case, the kernel never denied any syscall.

It's possible ffmpeg is giving bogus error messages.

another common thing that can fail which shows up in kdump output is missing libraries showing up as dlopen() errors.

again, it succeeds to get a handle for the framebuffer as root. the kernel might not be giving something that appears as an error

handles[0] in the struct returned by drmModeGetFB2 is NULL when run as my unprivileged user, that's what's causing ffmpeg to throw an error

That's one step closer. Maybe the DRM device has some silent security fencing.

it seems like I just have to live with it. to be a bit on the safer side maybe I could run a second, unprivileged, instance of ffmpeg that actually handles RTMP streaming, so I don't have ffmpeg as root connected to a remote server

or I wonder if ffmpeg could be made setuid, setting privileges only for grabbing the DRM handle and then dropping them

it'd be nice if I could just override the security stuff with a sysctl or something. I'm on X11, the full screen's contents are already available through x11grab

why is there a jail_parallel_start but not a parallel stop?

Not sure but one jail can depend on another jail, and you usually don't want to stop them in parallel. If you have a httpd running in one jail, and database in another, normally you want to take httpd offline first, then the database, not both together or in a non-deterministic manner.

Hi guys, why is free bsd not so popular as linux, when obviously there is better order in things (like documentation) than it is on linux distros?

The more I read the better the FreeBSD sounds, but then I would expect to be more popular

JurassCZ In life you'll learn that the most popular option is not the best option.

JurassCZ: ibm

JurassCZ: Unix Wars of the 80s/90s and fear for lawsuits for using a Unix based system

is it possible to upgrade from 15.1-RC1 to 15.1-RELEASE ?

hernan604: it will be, yes. currently it isn't since 15.1-RELEASE isn't released yet

JurassCZ, Linus Torvald is actually on record for having said he never would have written his kernel had the BSD project(s) not been being sued by Bell Labs & Sun.

Everyone expected BSD to lose the suit. . . 'cause Bell Labs & Sun both had more money and lawyers than god.

But little old BSD prevailed with a pro bono lawyer.

we dodged a bullet there - imagine how freebsd would look if it had the sort of corporate control linux has

lolll

ivy: thanks for the suggestion with cloned_interfaces and /etc/start_if.<interface> but sadly that doesn't work for me, as if_wg doesn't support domain names as endpoints, so the wg setconf has to resolve the domain name, which doesn't work before network is setup and all hacks I tried to delay that setconf didn't work (defining a function that first sleeps and calling that with & and sleep 5

&& wg setconf … & )

any suggestions how to setup a wg interface in rc.conf with only base when the endpoint is a domain name?

Use rc.local? Use a cronjob which every minute tries to detect the connection and creates it if not found?

(Try rc.local first)

rc.local is. . . pretty sloppy.

nimaje, you can't use the IP?

no, the ip changes sometimes, that's why we have domain names

So. . . do a DNS lookup as part of your process?

that hack works

i mean you can do it elsewhere too, not just rc.local

when I try to configure the interface via /etc/start_if.<interface> that dns lookup is to early and my hacky tries to delay it didn't work, as I wrote a bit up

so do it later?

when dns is up

retry until it resolves

one thing here runs

out of crontab @reboot

a

while true; do if host "`hostname`"; then service postfix start; break; fi; sleep 1; done

could do getent hosts xxx here

Sometimes rc.local is what you need. Less elegant perhaps but it indeed works

Arguably the fault is with other systems being sloppy about waiting for proper networking

lts: well it's hard to wait sometimes

and what if proper networking never comes up?

locals now run via /etc/rc.d/local, you can make own script or use that

then you're crapped?

you wanted to make wg tunnel by ip from dns

so no dns no ip no wg

nimaje: I guess then you get "Job systemd-networkd-wait-online.service/start running (1min 58s / no limit)" :-)

lol

(Yeah I'm just happy using rc.local even if it's less elegant)

sleep "`jot -r 1 0 10`"

but the system should still come up, even if the ethernet cable isn't connected (or whatever is missing for proper networking), so ideally nothing should wait for proper networking

yeah don't do that

run daemon -f boo.sh from rc.local pr so

or

thanks, using daemon from /etc/start_if.<interface> helps, as well as not typoing the interface name in ifconfig_<interface> so that the ip address gets applied, no rc.local needed, wg setconf can just try to resolve the endpoint in the background until the network stack is setup and the dns server responds

hmm never tried wg

it loads config i see

setconf

and it resolves host there and then?

i would try to resolve it before but anyway

yes, for some reason if_wg doesn't support the endpoint being a domain name, so the wg tool resolves it on setconf and if it doesn't resolves it waits a bit and tries again to resolve it in a loop (probably with growing delays, but not sure)

hmm

well how would it use host anyway

do dns in kernel?

man -P 'less -p "WG_ENDPOINT_RESOLUTION_RETRIES"' wg

hmm

any mdo user?

building the stabilization week freebsd head right now, is the toolchain relatively "stable" compared to HEAD or can i expect to be rebuilding it often? (running buildworld now)

or is it even the toolchain? is it for the userland? Im used to building netbsd where you start with the toolchain