after i have changed a *.c file under /usr/src, how can I just rebuild and update/install the needed stuff? is there a generic solution to this? a buildworld would rebuild everything...

deever: depends on which file you changed. if it's something in usr.bin/ for example you can just run 'make install' in the binary's directory. if it's a library or kernel change, you probably want to buildworld/buildkernel. if you're editing the source often, consider using make's meta mode to improve build times

(meta mode enables incremental buildworld without needing NO_CLEAN, which causes problems)

@#$%*!@#%@#&$!@$!$!#$!#$%*(&!($

libc++ in base is... FML

jbo: having fun? :-)

lw, I think I suck hard

lw, coliru.stacked-crooked.com/a/5e6ce9f166a01d44

lw, does that compile with base clang++ for you?

(I'm on stable/14, so llvm 17.0.6)

jbo: works fine with "FreeBSD clang version 16.0.6" (releng/14.0) and "FreeBSD clang version 17.0.6" (main), as long as -std=c++20 is provided

I must be heavily retarded then

I encountered another (real) libc++ bug ealier (no kidding, can show later)

lw, I need some serious help.

jbo: fwiw you can just write 'return m_payload' because span has an implicit constructor from a contiguous container

lw, yeah but I don't want the full range

ah right

lw, wanna have some look at my "real" code?

sure

lw, help

lw, coliru.stacked-crooked.com/a/e929df837ae1df51

L200

jbo: make return type std::span<std::uint8_t const> because m_payload is const in that function

omfg

afk jumping out the window

lw: thank you!

lw...

fucking hell, I have been staring at this for almost an hour

thank you... (obviously)

lw, I do have another one for you, if you feel like it...

anyone here using pkgbase for keeping a STABLE branch system updated? I saw one caveat mentioned about how it can break the system if a package gets split

jbo: i am always up for questions about strange C++ issues

I'm a bit confused about why a metapackage isn't being used that specifies all other base system packages as a dependency

so that if a package gets split the new package will automatically get picked up

tm512: probably because then you couldn't remove the individual packages, which is part of the reason to use pkgbase. but this is a problem, yes

lw, it's not as much of a strange issue as it is lack of understanding (and skill) on my end. I am trying to work with static std::span and I am failing hard.

lw, coliru.stacked-crooked.com/a/274c085370315ea2

tm512: probably pkg needs a way to understand that if a new package is split out of an existing package, it should offer to auto install the new package. this would help ports, too (with subpackages)

lw, as per my understaind of std::span::subspan, it shouldn't use std::dynamic_extend if I specify the Count in my call to subspan()

lw: is there a specific mailing list to track that would give advance warning about this? I guess I could also get in the habit of snapshotting the system and using it as a fallback boot environment, should things go awry. good idea regardless of whether I use pkgbase or not

lw, don't even dare to bring up L52:L53 - it's an illustrative example :p

jbo: you're using the wrong overload of subspan() in make() which is returning a span with std::dynamic_extent, which is (std::size_t)-1, which trips your assert. do this instead: return make(msg_id, data.template subspan<sizeof(msg_id), (NumBytes - sizeof(msg_id))>());

jbo: en.cppreference.com/w/cpp/container/span/subspan

Title: std::span<T,Extent>::subspan - cppreference.com

wtf

how did you figure that one out so quickly? do I truly suck so hard at reading the reference manual?

the clue is in the error message: test.cc:32:27: error: static assertion failed due to requirement '18446744073709551615UL <= message::max_payload': payload does not fit into message

and i happen to work with std::span a lot and know that std::dynamic_extent is (std::size_t)-1 which led me to the cause fairly quickly

the sad things is that I know that too and that I realized that that is what is happening

I just failed at figuring out why :(

tm512: i would assume that freebsd-pkgbase@ has announcements about this, but i only recently subscribed, so i can't say for sure

tm512: this might be worth bringing up on that list in case there's a better solution than on the wiki, the wiki can be out of date sometimes

jbo: btw, you might consider adding something like 'requires (PayloadSize != std::dynamic_extent)' in appropriate places to catch issues like this earlier :-)

lw, that is a very good proposal - I like it!

Psssst, that list has archives. lists.freebsd.org/archives/freebsd-pkgbase.

Title: freebsd-pkgbase⊙Fo

lw, now that I look at this after you pointed it out it is so painfully obvious.

jbo: in return will you quickly review my new port before i submit it? :-) no need to build it, just have a quick look at the makefile

lw, gladly. I certainly owe you at least a commit now :p

jbo: le-fay.org/tmp/30d/0001-NEW-PORT-da…gres_exporter-PostgreSQL-metr.patch ... ty!

lw, I am quite new to the ports committer game as you know so... take everything I say with a big grain of salt.

lw, I wonder whether PORTNAME=prometheus-postgres-exported is the more fitting name (after a quick look at existing ports)

hmm, looks like where that is the case upstream was decent enough to name it as such tho

hmm, there's net-mgmt/prometheus-collectd-exporter and dns/prometheus-dnssec-exporter but most of the rest are just called foo_exporter... but i actually prefer the prometheus-foo-exporter name. i wonder if there's policy on this

there's also nfs-exporter and nginx-prometheus-exporter so apparently not :-)

(also there's both php-fpm_exporter and phpfpm_exporter? apparently these are actually two different ports that do the same thing...)

given that upstream is prometheus-community I wonder whether naming it prometheus- would indeed be "reasonable"

I'm a sucker for naming stuff. but I don't know whether we have any "real" guidelines on this that would apply without argument here

yeah... that is not consistent with existing prometheus-community ports but i might try submitting it under that name and see if anyone objects

lw, did you check whether upstream has some "special" requirements regarding go:modules,<version> ?

how would i know? the upstream go.mod says "go 1.19", i guess that means 1.19 or later? github.com/prometheus-community/postgres_exporter/blob/master/go.mod

Title: postgres_exporter/go.mod at master · prometheus-community/postgres_exporter · GitHub

(which seems to be the oldest version in ports...)

that seems reasonable

lw, that port seems quite high-quality from what I can tell. nice work!

i don't know anything about go, this is my first time packaging a go project so i mostly just copied net-mgmt/blackbox_exporter

lw, can't speak much for GO_TARGET tho. also, I'm far from an expert. but this looks quite "polished"

jbo: ty! will open a bug and see what happens then

lw, I'll pick it up. drop the PR.

(if you don't mind the mentor review time...)

are you sure? i think you might have regretted offering to do that last time :-D

not at all! happy to help where I can. I just have to follow the usual proccedure. Also, this time around I assume (expect) you have done poudriere-testport on 13.2-RELEASE and 14.0-RELEASE :D

jbo: bugs.freebsd.org/bugzilla/show_bug.cgi?id=276996

Title: 276996 – [NEW PORT] databases/postgres_exporter: PostgreSQL metric exporter for Prometheus

oh, i didn't change the name in the end. but maybe i'll mention that in a comment

I'll check with somebody more authoritive than me before I submit the review.

lw, so the libc++ bug I ran into earlier is this: template<std::pair<int, int> P> auto foo() { return P.first + P.second; }

and "shockingly": gcc.gnu.org/bugzilla/show_bug.cgi?id=97930

Title: 97930 – pair is not a structural type

wait, libc++ or libstdc++?

which ever is the clang one - and the gcc one had the same issue :p

(according to that PR at least)

ah right

libc++ is llvm AFAIK

yeah

but my brain is totally fried after this stuff you just helped me out with

no, i was just confused because you linked to a gcc bug, but you mean gcc had the same issue... i guess they both implemented it in the same way

although actually no, libc++'s span doesn't have a base class

yeah, the "shockingly" comment was not very telling sorry

oh wait this is pair not span

yep

lw, I put wondering whether I should open a base PR regarding this on the ToDo list for tomorrow

yeah... struct _LIBCPP_TEMPLATE_VIS pair : private __non_trivially_copyable_base<_T1, _T2>

jbo: if it's libc++, better to open an upstream issue then open a src PR asking to merge the fix

that is reasonable

lw, do you feel like answering a basic C++ question that I should probably know?

always

i often don't know what i'm talking about though, perhaps don't trust my answers

lw, how come the compiler cannot deduct N automagically in this case? coliru.stacked-crooked.com/a/8577611f2fbb4012

good question: template< class U, std::size_t N > constexpr span( std::array<U, N>& arr ) noexcept;

foo(std::span(a)) does work as i guess you already noticed.. i'm not sure why it can't find the appropriate ctor automatically. #c++ might know

I was playing with the thought of asking SO but I am not yet mentally ready for the abusive comments :D

jbo: if you're on discord, there's a discord server called "#include<C++>" which has several people very familiar with the technical details of C++, including a few WG21 members... i'd ask there, they are quite friendly

lw, I probably will, thanks!

Title: #include ＜C++＞ - #include ＜C++＞

lw, anyway, I was wondering if I have to write some deduction guide or whatever. but first wanted to understand why it doesn't "just work"

> #include <C++> is a global, inclusive, and diverse community [...]

jesus christ...

i don't think CTAD is allowed for a function?

why on earth would it NOT be inclusive and diverse by default?

jbo: have you used the internet before?

yes?

what's going on here pkg: manifest parsing error: error while parsing <unknown>: line: 1, column: 0 - 'key must begin with a letter', character: '0x00'

uh, this is probably not right "/builds/obj/builds/src/freebsd/src/lf/amd64.amd64/worldstage/libcasper-dev.ucl" may be a binary file. See it anyway?

this is what happens when you use UFS (??)

dunno

only used UFS once

actually all of this is on zfs so that doesn't explain it. hrm.

lw, I really can't believe this...

lw, I was sitting all night at the office trying to figure these two problems out.

next time I'll just bug you immediately :p

lw, have you done coroutine shenanigans yet?

jbo: a little bit, yes. i wrote a coro-based kqueue wrapper for netd so you can do things like co_await read(...)

lw, I have to write a communication manager for CAN bus stuff. it's a request/response based protocol over CAN so I feel like I should take coroutines for a spin.

jbo: that seems like a reasonable idea

jbo: if it's any help: github.com/llfw/netd/blob/main/src/netd.async/netd.async-kq.ccm

Title: netd/src/netd.async/netd.async-kq.ccm at main · llfw/netd · GitHub

std::expected <3

where it gets more complicated (and i haven't addressed this here yet) is if you need to wait for multiple tasks, or make tasks cancellable

lw, yep. I need to be able to have timeouts. I did look into that briefly yesterday but didn't find a neat solution yet.

Mem: 532K Active, 352M Inact, 50M Laundry, 30G Wired, 113K Buf, 787M Free

532K active? that seems... low

if I opt to upgrade STABLE the old-fashioned way, I'm wondering how much time a from-scratch build would take on this laptop, has a 2-core/4-thread i3-10110U

I seem to recall building STABLE from source on my old i3-3225 was tolerable back in the 10/11 days

tm512: a long, long time, most of which will be spent building llvm. as a wild uninformed guess, 6-12 hours

with ccache and/or not running make clean before each build, am I still going to be waiting for LLVM to build most of the time?

ccache and meta mode helps a lot (especially meta mode), but at least on main it does want to rebuild llvm quite often. maybe less so on stable

I guess LLVM has gotten incredibly bloated since the last time I was using FreeBSD. sounds substantially worse than even GCC

I've got Dragonfly on my home server which requires building from source for any upgrade (even between stable releases), and even on my home server with its Atom D525, the build only takes like 8hr I think

like if I start it at night it should be done by morning

reminds me of building netbsd on m68k. i guess not much has changed since then :-)

it would be nice if you could use an out-of-tree llvm for building src, so you could just use a package

maybe there is a way to do this

huh, ran into this weird issue, wifibox seems to be blocking incoming ssh connections

also can't seem to use wpa_cli on the host even though wifibox appears to be forwarding the wpa_supplicant control sockets. instead, socat just seems to enter some kind of busy loop where I have to restart the wifibox guest in order to get it to stop

great, getting the "fault errors on pipe A" spam in my dmesg again

ls

went back to drm-515-kmod to check whether the bug with VT switching breaking acceleration was also identical there, and it seemed like the fault errors weren't happening anymore, but I guess that didn't last

back to 510 :/

wish I could try out the 6.1 drivers, but it sounds like some LinuxKPI stuff has to be merged back from main into stable/14 before those will build

I guess I could set up a 15-CURRENT boot environment, give things a spin, hrm

"Normally, the kdc.conf file is found in the KDC state directory, /usr/local/var/krb5kdc." - /usr/local/var? really?

Are there any main Drawbacks to running FreeBSD as your Desktop OS when compared with Linux?

SponiX: worse hardware support, worse application support

although on the second point, linux emulation can help, really depends on the application

SponiX: if you run into stuff you have to compile yourself, expect some annoyances there on occasion

gaming in particular seems to be one area where Linux excels especially now with proton being integrated into steam, which from what I read, seems like that's still a bit rough under the Linuxulator. all of the results I saw were talking about Linux-native games iirc

with this most recent FreeBSD setup I've made, I just don't plan on doing much beyond some emulators and maybe classic Doom which is about all I expect this laptop's iGPU to be able to handle anyway, so it's less of a factor

i've been using freebsd on my desktop for about 3 months, but most of what i use it for is... developing freebsd, administrating other machines on the network which also run freebsd, and listening to music... so... that's probably not a very representitive sample. i have been meaning to try some games when i have spare minute though

I went Linux-free between like spring 2015 and late 2017, with FreeBSD on my desktop and OpenBSD on my laptop, and tbh I probably won't be doing that again

Linux is just about as convenient as computing gets without resorting to proprietary OSes

still love me some BSD though, and I keep coming back to it when I have an excuse to use it, lol

I've been doing FreeBSD on the desktop since 2000, but with the caveat that 1) I don't do a whole lot of gaming, and 2) when I do, I treat Windows like an appliance OS.

Last time I gamed was with OpenRA. It was not a great experience, but I'm not blaming that on the game nor the OS

I do pretty much all of my gaming on Linux. haven't run Windows, even as an "appliance OS", since 2010. though I don't play a whole lot of PC games, even though that's more than viable on Linux nowadays

I play too many games that'd either need an extra layer of emulation of some sort, and while I'm sure wine works great for some games, there's a _lot_ it doesn't work well for.

If developers en-mass eventually target something other than Windows, I might consider another appliance OS - but Linux would just become it, if it's Linux they're targeting. I can't be arsed to have to admin either Linux or Windows more than absolutely necessary.

Linux would just become another appliance OS to me*

I've had too many times in production resulting in me running Linux in anger, to ever consider it anything other than minimal-effort.

in the last few years Wine has progressed significantly. Valve's Steam Deck is based on Linux, and they obviously want as many games as possible to run on it, so they've put a lot of development into Wine

seems like more than half of the titles on Steam are either officially verified or reported as playable by users

already wasn't inclined on installing Windows even just exclusively to play games but now there's even less reason to consider it

also generally not a fan of dual-booting, so Linux is just the obvious choice for desktop use for me. at the end of the day, the differences between Linux and the BSDs are way more subtle than the differences between Windows or Mac vs *nix, my workflows mostly translate over

doesn't make a huge difference what's running under the hood

windows for the occasional use of Word or MS Visual Studio

Linux for software testing but VirtualBox and access to the GCC server farm also covers a lot

mac for listening to music, some youtube  and web browsing

rpi playing around for the moment

and FreeBSD for most development stuff at home

and z/OS for bookkeeping?

9front for pipes and vpn ;)

Andoid for gaming ;(

not quite

ArcaOS?

do i need to compile my own kernel if i want to add my own boot splash these days?

i know i got a reply to look at splash(4) :p but it doesn't work

By boot splash, what do you refer to specifically?

Title: splash(4)

an image to hider the entire boot process

basically when i try to change to the old terminal/tty boot just hangs

Title: FreeBSD Desktop – Part 1 – Simplified Boot | 𝚟𝚎𝚛𝚖𝚊𝚍𝚎𝚗

Nixkernal: that's.. not at all helpful.

splash(4) is for screensavers at runtime, whereas the loader has its own thing going on (with files in /boot/), and the FreeBSD orb that's presented when the kernel initialization messages are muted requires kernel recompilation if memory serves.

I believe the default is /boot/logo-orbbw.4th for the loader, and I'm trying to remember which source is responsible for the muted logo.

Anyway, you can try adding boot_mute="YES" to loader.conf(5) to see if that's what you want, then we can figure out which file it is after that.

If splash(4) doesn't work as in it doesn't present a screensaver when configuured properly, that's a separate issue that should be reported on bugs.freebsd.org

Title: FreeBSD Bugzilla Main Page

The boot process on x86-like systems consist of a whole bunch of things, and there's only so much that FreeBSD has control over.

hi, when i want to open FreeBSD-14.0-RELEASE-riscv-riscv64.qcow2 with qemu, it doesn't resume like that ^

Title: View paste UYEQ

i run with this parameters

any idea, any wrong with my parameters or env?

aaah yes, the good old: "The complexity of an attempted match against a regular expression exceeded a pre-set level."

Since you guy having problems RTFM -> splash -> In order to load the splash screen or the screen saver,

debdrup: That's the bootscreen your talking about, a splashscreen is like the spinning wheel on windows or the apple on mac's

You can see it in vermadens post

Ooo neat, christos@ is doing some work on mixer(8) and snd_uaudio(4).

Mounting from ufs:/dev/vtbd0p2 failed with error 2: unknown file system.

does that mean my kernel doesn't support ufs?

rtprio: or that you need to load ufs.ko

ah nuts

at i think that's right, i got this error yesterday as it happens but i don't recall the exact message

ufs_load="YES" was just the thing

yeah, clear sign you're missing `options FFS` (and probably related goodies) and haven't loaded the kmod

switching to MINIMAL kernel; 8mb vs 28mb

trick is i just have to find all i need in loader.conf first

fwiw, on bhyve i load virtio, virtio_pci, if_vtnet (these are in MINIMAL anyway, but i remove them from mine), efirt, nvme, ufs, and sysvmsg/sysvsem/sysvshm if running postgres

yeah, i just need to add that to my template when switching to the minimal kernel

has anyone had an issue with freebsd's libgss where it ignores $KRB5CCNAME when loading the kerberos ticket?

During daily periodic... lockf: /var/run/periodic.daily.lock: already locked

It tried to run twice?

Or the lock file was never cleaned up?