Hello, all.

Any suggestions on a glob (or regex) pattern that can be used to match an exact package name, WITHOUT the version (so that it matches all versions), while also excluding similarly named packages but different packages? For example...

cups-2.4.2_1.pkg cups-filters-1.28.16_4.pkg cups-pdf-3.0.1_2.pkg

Glob pattern that matches cups-*.pkg but NOT cups-filters-*.pkg, etc.

I guess I could figure out a regex to match only characters that will appear in the version (e.g. :digit:|\.|_)

But not sure what exactly is valid and what will never appear in the version part of the full name.

going to finally install freebsd tomorrow, prob as a dualboot with windows 10 for ai stuff and any modern games i might want to get into

_xor: version can include more than that

e.g. make -v PORTVERSION -C /usr/ports/misc/zoneinfo/

Yeah, I didn't want to assume :/

what is the problem you are trying to solve?

Not really a major/immediate problem, but basically I'm making my poudriere environment as portable as possible. It's mostly there. The reason for the pkg glob is...

can't you get the package name from the package itself?

In pkgbuild hook for poudriere, when it gets a success event, I'm uploading the built package (*.pkg) to my central file server which holds my internal pkg repos.

should i go with my freebsed/w10 dual boot idea or not? im really interested on freebsd since ive been noticing some inconvenients with linux, but cant fully move my workload away (even with virtual machines) because i do ai stuff

Part of that means that I need to delete any previous versions of the package, otherwise pkg on the clients complains about duplicate dependencies.

technically i would be better off using linux as the other os, but setting up xformers in devuan has turned out to be a pain for me

pkg query --file /var/cache/pkg/zsh-5.9_1.pkg %n --> zsh

Yeah, that would be ideal, but from what I was in the docs, it only passes in the package prefix, not the full name sans version.

*saw

oh, hmm. good point.

wait no

Are there other packages that would match zsh*?

Problem is not being able to accurately restrict it to just the package name, where a prefix match can include more.

why a prefix? compare both name and origin

Oh that's going to be kind of annoying heh. That means that I'll have run remote ssh commands to read the origins.

The hook does get both the origin and the package name, so that's at least something. I can do an initial prefix match to reduce the candidate packages down to a handful, and then iterate over those candidates to match the origin and exclude any from the last that don't match. Then delete the matches.

Wait, can pkg-query take multiple files at once?

there's also pkg version --test-version to properly compare versions

Oh nice, didn't know about that one.

it would be nice if pkg clean was extended to do what you want

actually, try setting PKG_CONFIG_CACHEDIR to your repo?

Couldn't join this channel from Matrix (getting err_needreggednick). Is that intentional?

cydox: yes, the channel is +r

cydox: you'll have to figure out how to identify to libera's NickServ from matrix first.

Ok, thanks!

Got it working.

yuripv: Does that just allow you to change the default from /var/cache/pkg to something else?

yuripv: Does that just allow you to change the default from /var/cache/pkg to something else?

The the heck. That was weird.

For gpart bootcode -p, can it take the full device name or does it require -i?

hi, in linux fstab there is "nofail" option. It means that at boot time this mount is allowed to fail, if it fails the boot continues as normally. Is there an equivalent in freebsd?

I think that is already the default in FreeBSD (and most other Unix-like systems). It's only systemd that fails.

rwp: I tried, if the mount fails the boot is supended and it enters single user mode (it asks the shell you want to use)

gustft, rwp: man.freebsd.org/fstab(5) "failok"

Title: fstab(5)

thank you meena

gustft: depending on why it's falling, you can add "late" or just "noauto" to fstab

Is there any sense in disabling fsck check for EFI partition (ESP)? Also, how would|should I proceed in case of need to fsck the ESP?

meena, Thank you for that reference to failok.

a useful trick I discovered recently is that you can use "noauto" in fstab, and then run automountd with a "noauto" map.

RhodiumToad: neat!

only snag is that it doesn't play nice with also using a -media map and automount -c from devfs.

er, devd.conf

(automount -c tries to flush every autofs mount, but that's not the right thing to do for direct maps like noauto, since it ends up clearing the "automounted" flag (and other flags too) from the real mount)

RhodiumToad: no need to remember the difference between devd and devfs much longer…

?

RhodiumToad: never mind, i got overly excited, and… wrong.

it's: merge devmatch and devd (lib-ification)

hrm… i need to pass some "kernel options" on boot… and it's not working

how did you try?

escape to boot prompt, and do: boot kernel -ccloud-init=disabled (-c is supposedly ignored: but somehow it then fails to find the root partition properly

either way, once it boots, kernel_options is empty

just boot -cwhatever?

"Any flags or arguments are passed to the kernel, but they must precede the kernel name, if a kernel name is provided."

so sayeth the manual, if you can believe that

help boot shows: boot [<kernelname>] [-<arg> ...]

and the help text doesn't say anything that would contradict that

meena: set kernel_options="foo" from loader prompt?

oooh neat idea

the code seems pretty clear that if you pass more than one arg to boot, and the first arg doesn't start with -, then no options are processed

yuripv: success!

now, back to the acutal task

now back to trying to remember what the actual task was

also, "-c this flag is currently a no-op."

(from boot(8))

yuripv: so i thought i could smuggle stuff in there, but it doesn't work with boot, which actually tries to parse the options

yep, options for boot are documented in boot(8), i don't think they are passed through to kernel

yuripv: according to "help boot" they are

but I'm not sure i believe that any more

boot -v and boot -s certainly still work

¨/10

doh

it would be nice to know if service foo stop failed because it failed to stop the service, or because the service isn't running

check the service state after?

meena: what's the exit code of it?

debdrup: 1

yeah, that's not 0

wow, what an inane statement..

anyway, it is telling you that it failed by the exit code being 1

i know that bit, but I don't know why it failed

that's different from what you said first :)

it's also something that's not entirely easy to do without the solaris fault management framework or an equivalent of it

i dunno, service foo stop could return 2 or 0 if the service is already stopped

meena: thanks, those unsupported features don't matter to me, basically as long as ethernet and usb works I'm golden.

You know. . . unicode support in your IRC client is a double-edged sword.

CrtxReavr: ⚔️

Hi

evenin'