<zorz>
musl is not slower actually is a bit faster 30:06.33 Finished `release` profile [optimized] target(s) in 29m 36s
zorz has quit [Quit: leaving]
_whitelogger has joined #opencrux
<remiliascarlet>
farkuhar: Oh, didn't know that it was supposed to be Bash from the get go.
<remiliascarlet>
SiFuh: I had one, but I replaced it with NetBSD a few days ago because Void kernel panicked on boot. I can still install it or Alpine in a virutal machine though.
<remiliascarlet>
s/virutal/virtual
ppetrov^ has joined #opencrux
_whitelogger has joined #opencrux
remiliascarlet has quit [Remote host closed the connection]
remiliascarlet has joined #opencrux
<ppetrov^>
hey guys, do you have opencrux somewhere, like github/lab, etc?
<farkuhar>
SiFuh reported that glibc needed python to build? It's not obvious from that Pkgfile, but then again, dash depending on perl was not obvious from the dash source tree.
<remiliascarlet>
farkuhar: "SiFuh reported that glibc needed python to build?" Skill issues.
<remiliascarlet>
If your libc requires Python to build, then you have skill issues.
<remiliascarlet>
Actually, as much as I'm not surprised considering it's GNU, at the same time I'm surprised they didn't require Lisp to build glibc.
<SiFuh>
libxcrypt requires perl to build, dash requires perl, glibc requires python It's a fucking joke
<SiFuh>
farkuhar: So if libxcrypt gets used (Which it probably will as much as I hate it), then perl comes in
<SiFuh>
farkuhar: # Depends on: db gdbm libnsl libtirpc libxcrypt <-- perl depends on libxcrypt and libxcrypt needs perl to compile?
<farkuhar>
SiFuh: You're right, it's a fucking joke.
<remiliascarlet>
"perl depends on libxcrypt and libxcrypt needs perl to compile?" The chicken and egg problem.
<remiliascarlet>
Also, if we use Bash, why do we need Dash?
<remiliascarlet>
From what I understand, Dash is an Ubuntu-specific shell?
<SiFuh>
Dash stands for Debian Almquist Shell. It is a POSIX-compliant implementation of Bourne Shell.
<SiFuh>
But this is not true
<SiFuh>
It's _mostly_ POSIX-compliant
<SiFuh>
OpenBSD's ksh is POSIX compliant. OpenBSD has extra modules that are not, but you can set set -o posix or export POSIXLY_CORRECT=1
<SiFuh>
I'm probably going to build dash, if I have to build perl for libxcrypt anyway.
<SiFuh>
farkuhar: remiliascarlet: dash also needs autoconf and aclocal and the URL is http and not https for gondor.apana.org.au and the kernel.org version needs you to run autoreconf
<SiFuh>
So for the time being. dash is in the trash again since I am strictly working on minimal core and not downloading 50 dependencies used by no other core port for one program
<remiliascarlet>
Why don't we just swap the Linux kernel for a kOpenBSD kernel at this point?
<SiFuh>
farkuhar: I guess the reason that certain core ports a using other ports is possibly due to the ISO bootstrapping that jaeger does.
<emmett1>
guys, how to setup pkg-get?
<emmett1>
i mean, i tried 'cd <package dir> && pkg-repgen' it generate zero package
<ppetrov^>
well, pkg-repgen extracts deps infor from prt-get's conf, right?=
<emmett1>
okay, but i think only pkgmk.conf point PKGMK_PACKAGE_DIR
<emmett1>
prt-get only point to port repo
<ppetrov^>
oh, right
<ppetrov^>
but you should list your repos in prt-get.conf, no?
<ppetrov^>
because dependencies info should be extracted from somewhere
<emmett1>
yes ofcourse, its done
<ppetrov^>
actually, better wait for farkuhar to chime in
<emmett1>
owh okay then, wait for the CRUX master
<ppetrov^>
yey, having fun writing man pages for my small crux-related tools
<ppetrov^>
weird syntax, troff
<farkuhar>
ppetrov^: in the revised core collection that SiFuh is working on, there are some conversion tools that generate troff from more modern markup languages. I'd consider looking into scdoc. https://dpaste.com/FARJ64B7H
<ppetrov^>
farkuhar, thanks a lot
<ppetrov^>
i installed a colouring plugin for vscodium
<ppetrov^>
i don't think it's relevant anymore, as the issues I had, you already fixed
<farkuhar>
Not that issue in particular, but the fact that emmett1 is getting an index page with zero packages. Maybe the file glob is too rigid to match anything in his package dir.
<ppetrov^>
It's some form of Elvish... I can't read it...
<ppetrov^>
ye, got it. thanks
<farkuhar>
Without the --no-pkgmk-conf flag, pkg-get will try to parse /etc/pkgmk.conf to determine the compression mode, and then use a more restrictive file glob. It might happen that someone builds a bunch of packages with PKGMK_COMPRESSION_MODE="xz" and then changes this value to "lz", after which pkg-get will not be looking for any of the xz-compressed packages when building its index.
<farkuhar>
The wireless networking guide in https://crux.nu/Main/Handbook3-8#ntoc23 might need some adjustments. I have one laptop whose SERVICES array only points to /etc/rc.d/net, and somehow the appropriate wpa_supplicant process still gets spawned, even without copying the example file /usr/share/dhcpcd/hooks/10-wpa_supplicant into /lib/dhcpcd/dhcpcd-hooks/
<farkuhar>
Another laptop has /etc/rc.d/wlan in the SERVICES array, which is more transparent about the exact sequence for starting or stopping the wpa_supplicant process.
<ppetrov^>
farkuhar, I also think the wifi guide in the handbook should be fixed
<ppetrov^>
then tell users to turn on /etc/rc.d/wlan
<ppetrov^>
and that's it
<ppetrov^>
also, it's funny we have wlan0 as an example here and DEV=wlp3s0 in /etc/rc.d/wpa_supplicant
<farkuhar>
ppetrov^: So basically just replace "net" with "wlan" in the SERVICES array that ships with the default rc package?
<ppetrov^>
that's what I do
<ppetrov^>
also my recent installs on my side computers, I turn on the wifi immediatelly after setup-chroot, among the first tthings
zorz has joined #opencrux
<ppetrov^>
then i set up pkg-get and have a wiorking system in 30 min, from my own packages
<farkuhar>
What bothers me about /etc/rc.d/wlan (and why I hesitated to recommend it when revising the Handbook for 3.7) is that it's basically just a wrapper around two other service scripts (wpa_supplicant and dhcpcd). Why not just add those two scripts to the SERVICES array directly?
<ppetrov^>
I don't know. Why not?
<ppetrov^>
since wlan is there and will not be removed, just use it as an exmaple
<farkuhar>
But in fact it seems you can get away with the default SERVICES array (which points to /etc/rc.d/net), and not even bother with copying the example file /usr/share/dhcpcd/hooks/10-wpa_supplicant to /lib/dhcpcd/dhcpcd-hooks/. So why not tell users to run wpa_passphrase and then change the DEV line in /etc/rc.d/net (no need to change the SERVICES array at all)?
<zorz>
farkuhar: this is what i do
<farkuhar>
"wlan is there and will not be removed" <- Heh, remember what channel you're in. OpenCRUX is free to change anything, including the scripts that get shipped with an opt port.
<zorz>
in runit... i run dhcpcd service with hooks. in sysvinit net
<emmett1>
farkuhar: thanks for the link, it seems I already did that on pkg-repgen part
<zorz>
where modeforker SiFuh_ ?
<emmett1>
and i'm only use xz COMPRESSION
<emmett1>
how actually pkg-repgen works?
<emmett1>
I cant read perl
<zorz>
hi emmett1 excellent work your iso. minor issues fstab creted for me same prtition / & home. major issue your python keeps calling for glibc.
<emmett1>
hi zorz, i noticed your issue with my iSO
<zorz>
basically your iso with farkuhars python should be okay.
<emmett1>
Will fix those in next release
<zorz>
no rush... just informing.
<farkuhar>
emmett1 can read bash, though. The glob that pkg-get is using to populate its index page is basically the bashism "${1-$PWD}/*.pkg.tar.{gz,lz,xz,bz2,zst}" if you call it with --no-pkgmk-conf. Otherwise it reads /etc/pkgmk.conf and uses the value of PKGMK_COMPRESSION_MODE to perform a more restrictive search in the current directory (or in $1 if nonempty).
<emmett1>
python is from CRUX, i guess same issue with CRUX-glibc/
<emmett1>
zorz: for glibc issue, add alias to /etc/prt-get.aliases so glibc wont get pulled
<zorz>
python from your iso, when goes for python3-{mako,markupsafe} calls glibc
<ppetrov^>
"remember what channel you're in" I beg your pardon. I'm discussing CRUX, not the fork and I am only here, because this became the most active channel. If you'd like me to address CRUX-related issues, I can do it in #crux.
erilun06-mobile has quit [Read error: Connection reset by peer]
erilun06-mobile has joined #opencrux
<zorz>
emmett1 i went back to glibc.... its faster for me.... but i need to do something like your LLVM because crux LLVM gives host for everything, Aarch64 riscv etc etc same with gcc just trim packages
<ppetrov^>
for the record, net also have stuff like remember what channel you're in
<ppetrov^>
eeeh
<ppetrov^>
stupid paste
<ppetrov^>
net also has stuff like DEV=enp11s0
<farkuhar>
Heh, #crux these days exists for cruxbot spam and for old-timers to report signature mismatches. It's not even very active for troubleshooting beginners' problems.
<zorz>
ppetrov^: you either run net or dhcpcd service with hooks to wpa_supplicant.
<zorz>
in crux i prefer net.
zorz has quit [Quit: leaving]
zorz has joined #opencrux
<emmett1>
sorry guys, got caught in something
<emmett1>
farkuhar: so what should i do. I dont have weird config in pkgmk.conf
<emmett1>
i dont use --no-pkgmk-conf.
<emmett1>
just 'cd <package dir> && pkg-repgen'
<emmett1>
zorz: those glibc pulling issue caused by ports in non-core defined glibc as dependency. Thats why you need to alias it
<emmett1>
wait. seems like its not. i wonder which port keeps pulling glibc
<emmett1>
looks like some ports depends on libxcrypt, which is depends on glibc
<emmett1>
btw thanks zorz for report the issue with my ISO
<farkuhar>
emmett1: yes, 'cd <package dir> && pkg-repgen' is how it usually works.
<emmett1>
farkuhar: hmm..but its not work now..its still detect zero package
<emmett1>
wait, currently i'm on VPS in crux's chroot
<farkuhar>
zorz: I'd be curious to know how long the dhcpcd service has supported auto-spawning a wpa_supplicant process. It used to be that the example file /usr/share/dhcpcd/hooks/10-wpa_supplicant needed to be copied by hand into the /lib/dhcpcd/dhcpcd-hooks/ directory, otherwise you'd have to start wpa_supplicant manually.
<emmett1>
because it just works fine on my laptop
zorz has quit [Quit: leaving]
zorz has joined #opencrux
<zorz>
yeah farkuhar I learned this trick from you/handbook [dhcpcd hooks]
zorz has quit [Quit: leaving]
<SiFuh>
Perl id in
<ppetrov^>
if you guys include xfce in your fork, i'm ready to jump ship :P
<SiFuh>
....
<emmett1>
after tinkering for a while, re-setup crux rootfs, somehow my pkg-repgen is working now, not sure whats the issue
<SiFuh>
farkuhar: I did dash too but I didn't set it up as a core port for now. It's in opt. It doesn't have to stay there though.
<zorz>
bravo sifuh
<zorz>
dash is great.
zorz has quit [Quit: leaving]
<SiFuh>
ukky: I will try to bootstrap pkgsrc on CRUX
<SiFuh>
as a user
<ukky>
SiFuh: good luck. I build packages as a user. The build user is owner of /usr/pkgsrc, where Makefiles are located. System should ask for root password only when you do 'make install'.
<SiFuh>
ukky: I think I need it. Keeps erroring out here pkgsrc/bootstrap/work/wrk/pkgtools/libnbcompat/work/.packages/libnbcompat-20250303.tgz
<SiFuh>
Ahh I think I know what it is doing.
<ukky>
You should not need root permissions when building.
<SiFuh>
I am doing --unprivledged for the bootstrap
<SiFuh>
ukky: bootstrap succeeded
<ukky>
I think --unprivledged would require installing everything in home directory. But I might be wrong.
<ukky>
Also, when building packages, root permission will be required when installing build and runtime dependencies, and then one more time when doing 'make install'.
zorz has joined #opencrux
<SiFuh>
ukky: It wanted a folder under /opt/ for the db