I liked Clipious too. But most, if not all invidious instances are now crippled, all thanks to Scroogle

yewtu.be still works

wondering how do i configure udp-encap esp when using setkey ?

is there any general rule of thumb when it comes to sharing files between jails?

drobban: ro or rw?

one jail need to be able to write, the other only needs read

nullfs ?

babz: perhaps. isnt that ro?

nullfs suports also rw

you also can normaly rw mount the dir in the jail which writes and nullfs mount it to the jail which needs to read from it

Any news on getting wifi working on RPi4? how do I go about debugging?

Success: freshbsd.org/freebsd/src/branch/main?q=BCM43438 doesn't look like it

Title: FreeBSD / src - FreshBSD

ahh, thx for the clarification satanist - will have a look at the nullfs manual

is upgrading from 12.4 to 14.0 once its out directly a bad idea?

sopparus: maybe upgrade to 13 until 14 is out?

i could but it would be extra work, hence the question :)

I'm sure someone has documented that somewhere

should work

ill just try, it has not much on it

just a drive by "thank you freebsd" for the calm and low maintenance life i enjoy

poudriere keeps rebuilding qutebrowser when i add some pkgs to my repo's list and relaunch bulk, is there some way to check what triggers those kinds of things (i assume qutebrowser needs to be reconfigured and thats why ?)

cedb: poudriere testport has a lot of debugging output. especially when you set DEVELOPER=yes, IIRC

meena: you just launch it like literally $(poudriere testport) ?

cedb: you launch it the same as `bulk`

right im asknig cause i just put a pkglist in my /usr/local/etc/poudriere.d/ as all the guides ive seen were saying but

there doesnt seem to be an obvious assoc between the previously built stuff and the current build (apart from the jail obv i mean in terms of seeing the delta between two builds)

poudriere testport takes a single port parameter

cedb: was the package for qutebrowser successfully created?

ahhh sorry didnt get that, yes it did get created properly!

its just been three times that i add a pkg to say foopkglist and do "poudriere bulk -j <jail> -p head -f foopkglist" and

cedb: so if poudriere decides to rebuild it, you should see a line in the output saying it deleted the pkg and why

hm i should rtfm and investigate a big more i thought this maybe would be one of thsoe things thats super obv for people ith a little experience (bsd's newbie here, mostly used binary pkgs and ports "by hand" so far)

RhodiumToad: gotcha ill look into the logs

which btw is there some "best practice" or conventino to manage the lgos( verbositiy, rotation)? a mostly self hosted desktop "daily driver" generates pretty impressive amounts

did you look at poudriere logclean?

yes ive run it on previous isntalls the emphasis is on "conventions" like just nuking everything after a successful build seems a bit harsh yk

RhodiumToad: im running the testport subcommand for qutebrowser is there some way to narrow down "what im looking for", my wild guess would be that an old failed build is polluting the jail and the successful build doesnt cancel that

the other thing would be ccache ive enabled according to the obvious defaults online but i dont know much about these tools ill check that after

testport will always build, it's no use for finding out why something builds when you didn't expect

you want a normal bulk run and at the start there should be a line saying why it deletes a package

the question is what message do you get when doing a bulk build

also make sure the bulk build isn't using -c or -C

"Deleting qutebrowser-3.0.0.pkg: stale package: unwanted origin www/qutebrowser@qt6"

what exact command did you use?

ah, report a bug, because the first entry in FLAVORS isn't the default flavour, but not sure if against www/qutebrowser or against poudriere (probably www/qutebrowser)

yeah, that's a bug in the port

cedb: did you put just www/qutebrowser in your list of ports to build?

cedb: putting www/qutebrowser@qt6 instead might be a workaround

(assuming the qt6 flavor is the one you want)

ya the whole qt6 vs qt5 thing is often a bit messy

nimaje: hell youre right!

the bug here is that the qutebrowser port has FLAVOR?=${FLAVORS:[2]} which is not allowed - the default flavor is always ${FLAVORS[1]}

ya no i didnt specify a flavor i didnt even know this was the proper syntax

interesting oh well well

i guess setting up tls for the nginx fronting the build wasnt worth it (hence the delay in answering) but in any case: poudriere.pynchon.me

nah, just someone didn't read properly that the first flavour has to be default one and having the ports themself having the FLAVOR?=${FLAVORS:[1]} line causes duplication and people thinking they can choose which one is the default, that logic should be moved into the framework and FLAVOR?= in the individual port makefiles disallowed

ya just surprised cause qutebrowser is pretty popular usually so yk, id expect some rude person to have already yelled at the poor maintainer

i still dont get why itd retrigger a build if its the same flavor, unless it says "stale-deleting->building" but really its just doing some basic sanity/smoke testing?

nimaje: that makes it unnecessarily hard to use ${FLAVOR} within the port makefile itself

cedb: it retriggers the build because there is disagreement over which flavor is the default

cedb: poudriere thinks it's supposed to build www/qutebrowser (default), but the package came from www/qutebrowser@qt6

sure but is there anything in that statement that says default=/qt6?

ah, yes if and stuff need the variables defined at parse time and can't do lazy eval

cedb: but when it builds it, the port builds the @qt6 flavor because of the incorrect FLAVOR?=

i guess to ask my question more pointedly: can you get this "im gonna decide this is incosistnt" because a port's "spec" is more specifically or explicitly asked for eventhough its the same in the end

damn ok my sentences suck sorry

nimaje: and you'd get into order-of-operations problems with things like OPTIONS_SLAVE or OPTIONS_EXCLUDE which need to be set before including the framework includes but which might need to be flavor-dependent

can poudriere .. bulk numbers@B trigger a rebuild even if default is A <: B ?

i guess im just linenoising at that point thanks RhodiumToad nimaje i would have pulled my hair on this one

cedb: flavors aren't versions, they don't compare

i meant that as a subtype relation

as in they could /contain/

cedb: there is one default flavor which is the one you get if you don't specify @flavorname

there are no subtypes for flavours only equality

ya okay im overthinking this its a classic freebsd thing where i think this could all go horribly wrong but people are actually sensible so its fine

cedb: otherwise you can list flavors explicitly using cat/port@flavor (multiple times if needed)

right ya im considering starting this whole multiple jails multiple zfs dataset userlands thing

and ${FLAVORS:[1]} (what should be the default) != ${FLAVORS:[2]} (what is the default because of a faulty FLAVOR?= line)

yall have a favourite way to manage "ports list", i just have this textfile i edit in vim and ehhhh

I don't have a ports list, I create local metaports

nimaje: the throwing off part is that the logs say "deleting stale" which seems a bit scarier than it is (this qtwebkit stuff...)

RhodiumToad: i guess sameish question but for metaports?

there's only one problem case i've run into, which is that while poudriere can build misc/zoneinfo when it's listed as a port, it can't build it as a dependency of some other port

(because it installs in places that poudriere doesn't mount writably)

hmm that sounds a bit too far for me idk why youd want to include that in a build

?

the reason to use zoneinfo from ports is that it's easier to update

sure just, does it need to be parametrized

ah, maybe I should start creating a overlay with local metaports as I already started to sort related stuff in my packages.list file and it makes it easier to install stuff

oh god, freshports last update: "Briefly: Egypt now uses DST again, from April through October."

damn people will /not let go/ of making calendars hellish on computers eh

if you've never looked at tzdata updates before, you would not believe how messy it is

like countries trying to change their DST rules on less than 1 week's notice

hmmm

every update is a mess, either a technological or a polite one that inspired it

funny part is my job i just quit was at a shop that does a pretty popular software for optimizing aircrew flight schedules

RhodiumToad: ok what the actual ... " Mexico will no longer observe DST except near the US border."

post.lurk.org/@acousticmirror/111008144295294599/embed <-- recently posted on the tz mailing list

fing lol

id ask who the hell willingly signs up for this but I tried to fix the moth ball of perl that we used for building at $work so i guess sometimes we just try to fix soemthing and get caught up

I subbed to the tz list to try and get some stuff deprecated

like W-SU, GB-Eire, etc

don't you want old timezone info present so you can retain the ability to deal with data expressed in those zones?

those are old alias names, not actual data

W-SU for example is just a link to Europe/Moscow

surely can't be as bad as this.... mpv-player/mpv 1e70e82

Title: stream_libarchive: workaround various types of locale braindeath · mpv-player/mpv@1e70e82 · GitHub

much worse.

there are historical timezones which contain non-contiguous days

i.e. the current date goes forward, then backward, then forward again

what is the mailing list you referered to?

the iana tz list. but it's actually quite boring

OK

it's messy because every country in the world thinks it can just tweak its own timezone/DST rules and everyone will immediately follow

which is why there are 3 to 10 updates per year to a data file which ought to be predictable years in advance

ah

for an example of the kinds of breakage, see catless.ncl.ac.uk/Risks/32/16#subj9

Title: The RISKS Digest Volume 32 Issue 16

heh

> CrtxReavr feels like cut(1) should have a way to select fields from the right.

so it was the fact that the birthdates were being defaulted to midnight?

I usually use awk when cut won't CUT (heh) it heh.

RhodiumToad: i suppose that since it was such a long time ago, there was not necessarily a record of the actual birth time, so for those old birthdates, they just default to midnight

so i guess that's the reason for the idea "don't do timezone changes at midnight" since it's a natural point where a lot of default things might happen

jmnbtslsQE: the combination of representing a birthdate as a timestamp with a time of midnight, plus the fact that midnight didn't exist on that day due to DST mess

yeah

phew, need to make sure i don't have any daily zfs snapshots in Saskatoon on april 13

most places have learned not to do DST at midnight. I believe there are still some that do

e.g. Chile

_xor: does awk allow that?

Sure, negative column index.

$(NF-2)

I appreciate awk and have been using it more recently, but of the tools I use it's probably the one where the differences between GNU and BSD surface most frequently.

Hmm, actually, it's between awk and make for me. Close call, though now that I think about it, I'm leaning more towards make.

how is C.UTF-8 not a thing?

?

C doesn't need UTF8.

i'm confused because C.UTF-8 is definitely a thing

it's even the default since... freebsd 13?

where? it is the default on freebsd and I think on debian to, on arch the butchered the name to C.utf8

$ locale -a | head -n3

C

C.UTF-8

POSIX

weird

it's just a way to get sane sorting and multibyte recognition at the same time

TIL

kevans: according to mpv-player/mpv 1e70e82

Title: stream_libarchive: workaround various types of locale braindeath · mpv-player/mpv@1e70e82 · GitHub

meena: not a fan of this person already, and I'm only two paragraphs in

i don't think i'm interested in reading the rest of this

there are some good points in it

however, it is from 2017, and so it's out of date on some key points

the fact that functions like sprintf are locale-sensitive by default and don't have a locale_t parameter variant is a real problem

This is gonna sound dumb. say if I do `for p in "/bin/*"; do echo $p; done` I get `/bin/[ /bin/cat /bin/chflags /bin/chio ...`, but I'd like to get `[ cat chflags chio`.

how about cd /bin; for p in "*"; do …; done; cd - ?

nimaje that would work, indeed, excpet I have like 10+ dirs other than /bin :D

okay I have an idea

using $p like that probably isn't safe

hm, using % or #? I can't remember which one was for prefixes and which for postfixes

that's also not safe

why do you want the list to begin with?

id generate the paths outside the loop and feed it ?

like while read f; do foobar; done <(find *pretend i know how to use find properly*) ?

looks like `for prog in $(ls /bin /sbin)` works and is safe

however, ls(1) keeps printing the directory name as well

I guess I can just grep that out

nooo use bash expansion for nerd points

sorry aehm POSIX** expansion

antranigv: what do you need the list for?

nerd points :) lol

RhodiumToad to drop a page to the Wiki :))

RhodiumToad try this: cd /bin && man * | awk '/^N.NA.AM.ME.E/{f=1;next};/^S.SY.YN.NO.O/{f=0};/^D.DE.ES.SC.CR.R/{f=0};/^D.DE.EP.PR.R/{f=0};/^Q.QU.UE.ER.RY.Y/{f=0};f;' |less

antranigv: ${foopath##${footpath%/*}} or something i always mess it up

${foopath##*/}

wait no thats the opposite of what u ant

nono i was right

your version is wrong if the path contains any metacharacters

i knew there would be some shell edge case weirdos KNEW IT

its like googling the best way to parse cli args: the answer is just dont use bash

if i had infinite free time, I'd build a system where /etc was renamed to "/System Configuration" and see just how much of the system explodes as a result

if i had infinite time id write an anti-debasher in some fancy lang for formal verification that *proves* you cant escape the madness in posix with discipline and you should just abandon all hope and use fish

if I had infinite time I'd just create milions of web sites, and do SEO where FreeBSD would be mentioned more than Linux

beat them in their own game

don't you know FreeBSD is dead

netcraft said so

dkeav: lol netcraft

LOL

do we have a tool that allows us to get the src path of a program in base?

something like `srcwhich /bin/ls` that will print `/usr/src/bin/ls` ?

crest , you might know the answer to that ^

not really, but for any specific tools it should be easy to find out

which ls # prints /bin/ls

cd /usr/src/bin/ls

replace the inner / in /usr/bin/... and /usr/sbin/... with a .

there are a few exceptions

i mean, sure, but zpool is not in sbin/zpool, it's in cddl/sbin/zpool

need a proper way to find them

i'm not aware of a good tool to map back from an executable to all the sources it's compiled from

maybe we need to generate a DB from the src

you may be able to extract it from the .meta files after building the source with meta mode enabled

e.g. parse the sources for /bin/ls from /usr/obj/usr/src/amd64.amd64/bin/ls/ls.meta

but do keep in mind that the .meta files reveal the ugly truth about which files compiler, binutils, etc. (may) have accessed to build a file

unless you want to learn that every executeable depends on $(CC) and all its sources

just computing the full transitive hull isn't useful unless you're the build system

oh no plz no more hulls folks i dropped out for a reason

cedb what's hulls?

cedb: ?

I'm not sure exactly what we're talking about, but I think it's the minimal set of dependencies *effectively necessary* for something

(as opposed to the dependencies explicitly set by the programmers)

RhodiumToad you're pretty active on the wiki. do you think this is worth publishing? antranigv.am/misc/wiki_baseutils0.png antranigv.am/misc/wiki_baseutils1.png

well, I did publish it. any thoughts would be nice wiki.freebsd.org/BaseUtils

Title: BaseUtils - FreeBSD Wiki

LxGHTNxNG: the first time i heard the term was in optimization for convex hulls

seems it was used sort of the same way when mentionning "hull of dependencies", rougly its just given a bunch of data find the shape of it

antranigv: while I'm not RhodiumToad, I think it's great! For me, this helps with the "discovery" of them.

TIL about `cu` command, hah, awesome