Title: Unofficial FreeBSD pkgbase repository

SamuelMarks: looking at string.h, you need __BSD_VISIBLE; but it should be already set in default compilation environment

yuripv: Define it to what? - This gives an error: /usr/include/sys/limits.h:73:19: error: invalid token at start of a preprocessor expression

#if __BSD_VISIBLE || __ISO_C_VISIBLE >= 1999

ah, to something non-zero, the check is #if __BSD_VISIBLE, not #ifdef, so it wants some integer

ok `1` should do the trick

Hmm I'm still getting `ld: error: undefined symbol: strnstr`

compiling on FreeBSD? :)

yuripv - Yep, here I just pushed it up cirrus-ci.com/task/6010050590801920

Title: Cirrus CI

OH wait that actually worked, weird, OpenBSD it didn't

SamuelMarks: I don't need to define anything to use strnstr()

I'm on 13.2-RELEASE

It should be visible by default if you're not messing with things

I'm having fun: #if defined(__FreeBSD__) || defined(__NetBSD__) || defined(__OpenBSD__) || \ defined(__bsdi__) || defined(__DragonFly__) || defined(BSD)

Hello, is the ue network driver able to do 2.5gb speed?

I would guess no due to being USB && RealTek being involved

s/being//

Wait, there is one another manufacturer with name starting with "A"

parv: AX88179 adapter?

kenrap, Right, that one. Name is "Asix".

It seems that many 2.5gb usb ethernet adapters available are rebranded realteks, so that's disappointing.

asix.com.tw/en/product/USBEthernet/Super-Speed_USB_Ethernet/AX88179 "NOT FOR NEW DESIGN"

Title: AX88179 USB 3.0 to Gigabit Ethernet | ASIX

parv: I appreciate the help and insight. I'll look for a pcie adapter instead now.

I guess the safest to look for would be an intel nic

Note that I could only guess. Do not have any experience with the hardware you are looking into

Mailling list archive is producing: "Error 503 Backend fetch failed". Is that due to ongoing maintanance or something?

parv: try running a rediculously overclocked RTL8139 /s

wrt 2.5gbps

whoosh

fonks, ^^No idea about your context

eck, I may have pinged the wrong person

joke is ruined :P

kenrap is who I meant to

For shame, fonks!

😞

There's a thought experiment though, could an 8139 run at gigabit speeds

hello everyone, after compiling emacs from ports with changed compile-time options, `pkg upgrade` is trying to upgrade it (28.2_3,3 -> 28.2_4,3). what will happen next? will i lose the compile options and stay with the binary one? or it will compile again to the new version?

crzjp: cant mix and match ports and pkg's

pkgs are compiled with the default compile time options if you need custom pkg's either go all ports opr use poudriere and build your own repo that overrules the default

cpet: thanks for responding, search for more info found i can `pkg lock` the package so `upgrade` will not touch it. is this a good approach?

personally I would do the poudriere approach

but I wouldnt lock it

ok, ill take a look

or send in a PR asking for yourtions to be the default

can be anther approach but it's emacs really doubt you actually need to do that

dunno, the compile option i altered is "--with-nativecomp" (or something like) that permits compiling .el files using libgccjit, there must be a good reason to be off by default

we use clang by default ?

cpet: sorry but I dont know, I'm new to freebsd system and its userland packages

yeah the base compiler is clang we dont use gcc unless required for whatever reason

hmm good to know

emacs with nativecomp is exactly reason for gcc, due to JIT. it is just no one upstream tried clang for that, hence the deal

nice to know too, thank all to the explanation

ill take a look on the mentioned tools and read some docs

good night for everyone!

Demosthenex: i mean nix/guix for defining reproducible systems. more like nixos or guix system i guess

dcunit3d_: ah, yeah, i suggest things like puppet

so i'm troubleshooting a *weird* wifi issue, wireshark shows lots of retransmits and duplicate acks, and ssh behaves *very* poorly, often hanging or disconnecting. then i dropped my mtu in half (1500 -> 800) and though a bit janky, suddenly it's responsive. i drop mtu to 400, and another increase in responsiveness. wtf

yeah, the sweet spot appears about 80

800 :P

Demosthenex: which wifi driver?

meena: iwm ?

its weird, i can ping ok, with some variance in RTT, but if i ssh, it stops and hangs up constantly

wireshark shows dupcliate acks, tcp retransmits, etc

wild

hello folks, i'd like to ask if someone here has experience with a wacom tablet on fbsd. its doing weird things for me like droping inputs rather randomly and i dont know how to approach fixing it. I suspect dbus or other input busses fighting for priority of some sort

meena: so i dropped the mtu to 800, and... it felt more responsive, further decreases didn't help

Hello, on FreeBSD, are the permissions of socket files irrelevant when a user wants to read/write data through the socket?

Is it the permissions of the socket's directory that matters for controlling reading/writing to the socket?

ns12, named pipe? en.wikipedia.org/wiki/Named_pipe

Title: Named pipe - Wikipedia

VVD: No, I am talking about unix domain sockets.

what is "socket's directory"?

VVD: e.g. in /path/to/the.socket, the "socket's directory" is /path/to/.

dirname "${socket}"

permissions work same as on files on FS

isn't it?

Demosthenex: wanna check bugzilla for existing bugs and report your findings?

meena: i'm not sure it's freebsd. the wifi is laggy, i'm trying to find out why the mtu impacted it

bad side of the house if you will

hi there, after upgrading 13.1->13.2 i got this message: pkg: Repository MyPudriere has a wrong packagesite, need to re-create database

nevermind... mistake in url ...

Is it possible to explicitly block a hardware, specifically GPU, from recieving power? or to run?

sozuba: yes, in the bios

my nios doesn't have such option, I'll check again though

re-checked, no such option, guess I amjust unlucky

Silliest question ever - I want to get into mailing list for current - if I click on that - just a popup comes and nothing else - cant find application to run ---- any advise ?

freebsd-current⊙Fo

is there a web access to read the discussions

Title: FreeBSD Mailing lists

As for opening an application when you click an email; the key word you're looking for is "URI handler" (probably combined with your choice of desktop experience or whatever)

hi :)

i use xsetkb to set keyboard language but I noticed after some time (sleep of monitor) keyboard language resets to defaults

how can I set it permamently or set it automaticaly after sleep?

debdrup: thanks a bunch --- I was wondering why it is such a gap of communication about the FreeBSD 14 status ---- I think people who take care of freebsd.org could link this lists.freebsd.org/archives/freebsd-current

Title: freebsd-current⊙Fo

cross-posting from Matrix: how would I go about debugging an OpenCL lockup? I cannot seem to run `clpeak` without it locking up my card (AMD Vega 11). Using `clover` and `drm-kmod`. Now if I swap `clover` for `pocl` (i.e. doing the work on the CPU) it works fine. This is on 13.2-RELEASE.

acu: mailing lists have been available on lists.freebsd.org for at least 20 years and is linked from several places (both on the main site, and in addition it's also mentioned in the handbook in the section that specifically talks about updating from source for -current (and -stable)) - so I think maybe it's on you if you somehow missed it.

ShinyCyril: what does "lockup" in this context mean?

ShinyCyril: first step would be to grab the symbols from the kernel-dbg and base-dbg distribution sets and extract them in the root filesystem, then rebuild the relevant ports as described on wiki.freebsd.org/DebuggingPorts

Title: DebuggingPorts - FreeBSD Wiki

Even if you could get it to crash rather than lock up (which I take it to mean, it doesn't generate a beep when you hold down more than 6 keys on a keyboard (which is an old trick to generate an NMI which results in beeps), you wouldn't be able to tell what's wrong without debug information.

deebdrup: for sure it is me, thanks for the pointer :) I just humbly suggested that this webpage would link to the lists not start your own desktop email application - freebsd.org/releases/14.0R/schedule

Title: FreeBSD 14.0 Release Process | The FreeBSD Project

meena: the X session basically locks up. If run from a virtual console then the process becomes totally unresponsive and can't be interrupted.

what does dtrace say?

there's an option for control-T to have it print the kernel stack of the process, that might be worth a try

acu: that's the release engineering schedule, which doesn't have anything t o do with the -CURRENT branch; by the time something is being handled by the release engineer, it's already well on its way to a -STABLE branch and will become a -RELEASE in time.

also procstat can print the kernel stack of another process

it's getting stuck on a _umtx_op: pastebin.com/ntS14r8v

Title: openat(AT_FDCWD,"/home/mike/.cache/pytools/pdict-v4-pyopencl-invoker-cache-v41-p - Pastebin.com

the above paste is from a different OpenCL application than clpeak, but symptoms are the same.

_umtx_op is the back end of the pthread mutex/condvar functions

acu: also, for what it's worth, those links are there because you're meant to be subscribed to the mailing list (because you can't send to it without being subscribed)

acu: agreed, i'd expect a list info page link, not mailto link

Really? I wouldn't.

debdrup: thanks, sometimes you want just to read :) I am not using 14 current - though it seem I will ....

If I am in 10 mailing lists - my box is flooded - and if I do not have things to say... anyhow thanks again

another question ---- is there a graphic depiction of FREEBSD architecture --- something like this developer.ibm.com/articles/l-linux-kernel ?

Title: IBM Developer

I like the figures - the depictions give you such a nice gestalt -

RhodiumToad: Control-t output seems to line up with dtrace capture from earlier. I will do some digging with the suggestions provided and see if I can get the callstack. I suspect I'm going to be in the guts of the AMD driver though - not sure what hope there is for me.

Only someone with deep knowledge and years of presence in FreeBSD can build such an useful graphic depiction of freebsd architecture - which is so fundamental for new comers and aspirational FreeBSdears :)

There's nullrouted.org/uploads/freebsdobservabilitytools.svg - which is more meant to give you an overview of what tools can be used to get information out of certain subsystems (though argualy that's more useful).

Brendan Gregg made that almost a decade ago, I just converted it to a svg at some point.

that's a great diagram! Brendan Gregg is a fountain of knowledge.

very nice - thanks debdrup and brendan - it is a good start - but I like that it mentions tools you can use to dig in :) nice ...

I have to print it an pin it on the wall :)

That's why it's an svg ;)

that is awesome diagram, thx

I think the IBM article about linux kernel architecture can be used to make one about FreeBSD - I know all the info is there I assume --- docs.freebsd.org/en/books/arch-handbook/book/- it just needs someone who knows to put it together ----

This article is so well written - even I can understand it :) developer.ibm.com/articles/l-linux-kernel

Title: IBM Developer

It was writen in 2007 - I wonder how valid is today

debian released version 12 with kernel 16 I think - debian.org/News/2023/20230610

Title: Debian -- News -- Debian 12 "bookworm" released

It will be interesting how far the split is between Linux and Unix (FreeBSD) ---- comparing Debian 12 with FreeBSD 14 would be an amazing article :)

Why does a USB device change from /dev/ttyU0 to /dev/ttyU1 (and back) w/o being unplugged? I have a daemon that hangs if the USB device changes.

Slowly, applications written for Linux will be more and more difficult to port to FreeBSD - I assume --- I just see that a new gimmick was included - systemd-resolved... other neurons required :)

makes sense why Linus Torvalds does not like distributions :)

Hi, got a quick question about the GELI/root-on-zfs combo in the installer: where are the GELI keys if you want to change your user key? /boot after the system is up does not contain them, and the boot pool's mount point is empty despite the pool showing as mounted

I believe the installer only sets a password

right, I'm pretty new and ignorant and thought there was no such thing as "only a password" from poking through the manpage. So I can just change it without specifying any keyfile?

if it's already attached, you can change keyfiles and/or passwords freely

hrm

but if you set a keyfile, then you have to be careful about whether the boot sequence can get at it early enough

are you using uefi or legacy boot?

I'm not at all interested in doing so if I can go without, so that's fine. The manpage just made it sound like a master key was a mandatory part of the setup

UEFI

the master key is mandatory, that's stored (encrypted) on the provider and nowhere else

there are two slots for storing the encrypted master key, each of which can be encrypted by a different keyfile/password combination

oh, there is also usually a backup of the provider metadata in /var/backup, which contains the encrypted master key under the user key(s) in effect at the time the backup was made

but for a root filesystem that's not much use unless you copy them elsewhere :-)

thank you

so just specifying the provider e.g. /dev/nvXpY is enough?

to change a user key? yes

yep, just want to nuke the old one

alright

I guess I also need to destroy the old copy of the master key once this is confirmed to work

when i look at the associated /dev/blahblah.eli in geli list, is UsedKey the keyslot?

or uh, can I otherwise somehow confirm what keyslots I use on a provider?

hm no

UsedKey seems to be which slot was used to attach the provider with

ok thanks, looks like all is well now

huh. looks like geli setkey on a detached provider is broken

huh ok, I found a bug report on that... like so? bugs.freebsd.org/bugzilla/show_bug.cgi?id=254966

Title: 254966 – geli setkey not working with detached provider

wonderful, another two-year-old bug not handled

with a patch, too!

did you just assign it to geom@ or was that already done?

In an hour or eight

RhodiumToad: i just set it from new to open (and probably added myself to cc)

makes me wonder who geom@ is

it's a mailing list

I was wondering why it didn't appear on the twice-monthly list of open bugs, but apparently that only gets bugs marked Open/In progress and not ones marked New

so thanks for setting it to Open, at least that should make it show up :-)

🤦🏻‍♀️

RhodiumToad: does that Patch make sense?

I absolutely loathe seeing #if 0 in code

the patch looks plausible? I haven't looked at the surrounding code in detail

I'm trying to, but my focus is kinda hosed

loads of new bugs that aren't open / properly assigned

there also seems to be something wrong with geli setkey's choice of number of rounds of key distribution

i don't know posix well enough to know if this is a bug bugs.freebsd.org/bugzilla/show_bug.cgi?id=271817

Title: 271817 – sed: range change with N results in no output

why can i assign these ppp crasher bugs to? bugs.freebsd.org/bugzilla/show_bug.cgi?id=271843

Title: 271843 – ppp can crash due to wrapping subtract in FsmRecvEchoReq()

that sed one is tricky, I don't think either the actual result or the reporter's "expected" result are correct

anyone ever wanted/needed a threadsafe way for a process running as root to open a file using the creds of a non-root user?

there's a bunch of cases in base where programs do that, but only by using seteuid to temporarily switch the whole process's creds

RhodiumToad: feels like something you'd be able to do with mac(4)

I don't think so

the main case where it comes up is where root needs to read files from a user's $HOME. Due to the vagaries of NFS, it can't assume that it can do that with root's creds

re: sed, nowadays it's usually "what gnu sed does" (that matches the "expected" output)

gnu sed is clearly wrong

here's how the spec interprets it:

first we read the line "a" into the pattern space and start the first cycle

the start address of the c command matches, so we then delete the pattern space

we are not addressing the last line, so the N command appends a newline and the line "b" to the pattern space

the spec isn't 100% clear what happens next, but on reaching the end of the script, it should write the pattern space, delete it, and start the next cycle.

Wonder how the SysV Unices behave in reality.

Since the standard's usually written to match them.

I'm not sure why this doesn't write "\nb\n" to the output, unless the pattern space is somehow still being considered as "deleted" thanks to the c command

anyway, the next thing is that we read the line "c" into the pattern space, it doesn't match /b/ so it gets deleted,

then we append a newline and "d" and the same happens as at the end of the first cycle

and again for "e" and "f", but notice that we just read the last line using N

and at that point we terminate without writing "x" (because the c command never ran on the last line of the input)

so the only thing I know is wrong is the "expected" output in that report, there is no way to justify that based on the spec

RhodiumToad: do you wanna update the bug with your analysis?

I literally just hit send

I'm just wondering, what's this delay about with FreeBSD 14-RELEASE?

it seems to have been on track since 25 April but then?

Title: FreeBSD 14.0 Release Process | The FreeBSD Project

wildeboskat: openssl 3.x

Ah ok

Compile problems?

Word on the st... what RhodiumToad said.

No, updating OpenSSL 1.1.x to 3.0.

There's a thread on current@ about it.

Ah what's current@?

I'd like to follow it

Good that they update OpenSSL tho

Title: FreeBSD Mailing lists

everything that uses openssl, in base and in important ports, has to be updated to allow it to use the openssl 3.x API

Thanks debdrup

It's a pretty significant amount of work, yeah.

Yeah that sounds like a lot

but with openssl 1.x going EOL so soon, it's not an optional task

I'm surprised they only decided this at the last moment but ok, it's a good thing to take on board

wildeboskat: it's a lot of work, and we've only now gotten a dedicated person on the task

And there's no hurry, 13 works fine :3

I was just wondering what happened

Also llvm 15

usually emaste would have worked on OpenSSL, but he was also very busy with LLVM integration work

what's in llvm15?

It's only 14-CURRENT that gets OpenSSL 3.x.

(13-stable has llvm15 too, fwiw)

Err, it's only what'll become 14.0-RELEASE that'll get OpenSSL 3.0*

RhodiumToad: i dunno, but loads of ports were failing to build with llvm 15

Conflicting types for dma_buf_stat ?

looks like it's stricter about warning about integer conversions

On console (in vt, 13/stable), why do I not get color in manual page rendered with "env MANCOLOR=1 man man"?

Bold & underline or italic rendered text is also misisng

is it missing in "man man" without messing with the environment?

Yes

I don't even get color in xterm, trying it now.

FreeBSD VM runs in VirutalBox if that would matter (though I would doubt)

TERM is set to "xterm" after log-in; I do not set that anywhere (set "xterm-color" for X11 session). I see root ~/.profile sets that

does CLICOLOR=1 ls -l get you colorized output?

Yes

what's the output of tput Co ?

No success with env TERM=xterm-color MANCOLOR=1 man man

RhodiumToad, That prints 8

and tput AX; echo $?

"1"

looking at man (it's shell script), all MANCOLOR does is modifying the MANPAGER to be `less -sR` (-R to allow ansi color codes)

something has to output those colors, and mandoc doesn't?

my $LESS is '-SsiXR'

I see expected colors and/or text with bold (depending on xterm configuration), etc attributes in X11 session on xterm

when doing what?

RhodiumToad, Asking me? Seeing a manual page

can you give an example of what manual page and what specific text you see in color?

Did "man man". "Name" header & name "man" are in green (for bold text); "Synposis", "man" & option names are in green, option values are in yellow (for underline or italic). Screentshot coming up

what freebsd version?

FreeBSD gaia 13.2-STABLE FreeBSD 13.2-STABLE zfs-arc-unit-B-n254945-b6310a2f8eaa INVM-efi amd64

if it's your xterm configuration, why do you expect vt to follow it?

vt uses xterm control sequences and termcap entry

but it does not use xterm configuration

parv: what do you have in the environment?

yuripv, I never did

Does "man man" supposed to display a manual page without any bold/italic or colors on vt?

I have so far failed to display any manpage in color on any terminal

Do you get bold or italic or underlined text?

yes, but note that it's less that renders that, not man

Yes, I understand. Here I am getting nothing

what do you have in the environment

env (on vt in zsh): termbin.com/1syh

Screenshot promised of xterm: files.mastodon.social/media_attachm…0/294/original/3135eb1a91f3d47b.png

How interesting.

parv: what is your TERM set to under vt?

meena, "xterm"

your colorized prompt works in vt, or not?

appres XTerm?

also what's the env when running in xterm

RhodiumToad, env in X11 session: termbin.com/geqp

RhodiumToad, Yes colorized prompt produces color on vt

do you still get colored output from man on xterm if you do env -U TERMCAP man man ?

er

do you still get colored output from man on xterm if you do env -u TERMCAP man man ?

Ok here is the thing: I need to restart VM in order to see text on vt after switching from Xorg. Will star X11 to show the resources for xterm

don't get uploaded into a pdp-11

oh right, i've been meaning to ask another thing

i have a host where the anope package segfaults for me on start

yuripv, I do not have "appres" installed, so Xresources: termbin.com/6cca

RhodiumToad, Re env -u TERMCAP, yes still same colors are produced in xterm

well, i should probably do some more troubleshooting on this first, so disregard

parv: what's the output of type mandoc

RhodiumToad, "mandoc is /usr/bin/mandoc"

and the output of man -dd man

man -dd: termbin.com/wwyb

that's in the vt?

No that is in xterm

where'd that get \less from?

That is from $PAGER

man -dd output looks similar enough on vt

I don't see a PAGER in the env you posted for xterm?

After "unset PAGER", output of "man -dd" changes to "zcat ... | mandoc | less -s"

and do you get color in that case if you omit the -dd ?

No.

note that PAGER is ignored if MANCOLOR is set

with MANCOLOR set, what does man -dd man output on xterm?

you have the colors set for bold, underline in those resources

aha.

In vt. Seems like I could differentiate between bold & not-bold text with "Terminus BSD Console, size 32" font. After unsetting PAGER (env termbin.com/ody0), "man man >x " produced escape codes that I see in the editor but cannot differentiate in pager

s/could differentiate/could NOT differentiate/

I have 1 for $MANCOLOR but I never managed to get colour. An ELI5 guide would be nice.

you won't get colors unless you modify mandoc to output them

redirecting output of man may change the behavior

Thank you RhodiumToad, yuripv, meena

(Too much for me to follow here.)

grahamperrin: as far as I can tell manpages can't actually produce color unless they're using groff, which base system ones don't

parv: i see bold in vt in vmware vm

grahamperrin: parv was using an xterm configuration that changes bold/underline to colors

Oh right :-(

and obviously that doesn't work in vt

I wonder what happened to the project to get better manual pages online.

no underline though (as well as no italics)

grahamperrin: you mean the web ones?

iirc, the web thing was doing something differently to what mandoc's html output does

(possibly it existed before that mandoc option did)

I need to find something that would make those (bold, underline) attributes in a manual page more obvious ...

... on console

With "MANCOLOR" listed in "man(1)" I thought I should be able to see such. But as RhodiumToad mentioned earlier, that is not to be

.oO( Man(!), what a goose chase .. )

parv: how does it look like for you in vt?

RhodiumToad: it certainly looked very different, more importantly the anchoring was great. I hope it happens.

yuripv, All text look the same; cannot identify if some important note easily without a search (found term is shown in reverse-video)

vbox right?

Yes

also uefi or bios?

yuripv, Screenshot of "man man" on vt: files.mastodon.social/media_attachm…2/583/original/0a1f36bfbe2e75c5.png

yeah, i have the bold text shown as bright

same terminus font

yuripv, UEFI, well "EFI" in VirtualBox settings. /boot/efi is empty on FreeBSD side

i mean vga, or efifb, or kms driver for vt

so far for me vmware/efifb/terminus shows bold as bright, no underline/italics

The only relevant "vt" or "fb" I see in "dmesg" output is "VT(efifb)"

the backend for vt shouldn't have much effect on its behavior, afaik

it's basically manipulating a framebuffer

Changed "kern.vt.color.(7|15).rgb" in "/boot/loader.conf" to #f0f0f0 (near bright white) & #f0c0c0 (pinkish) to no difference other than text is now pinkish

you said you had color in vt with ls (with CLICOLOR=1), so color escapes are obviously working, no?

RhodiumToad, Yes, color in ls output & shell prompt are visible

Now I am just testing if I might have decreased intensity of "bold" text

Checked that disabling "allscreens_flags='-f terminus-b32.fnt'" does not produce text of different intensities|colors in "man" output

Next is to test disabling "blanktime" ...

... nope

I give up

meena: I finished with that sed bug report.