@rwp Run from terminal 'firefox --safe-mode' and check again 'lsof -i ...'

If GeoBarRod comes back I'll say that disabling all extensions is a good test. And having done that things are basically the same. bsd.to/ft1a/raw

Title: ft1a

I noticed this a long time ago

basically, if the connection times out, the socket stays in the CLOSED state until the client calls recv()/send()/close() ect

It doesn't immediately break anything but I think there is a definite resource leak there. I don't see it on a Debian version of firefox. So this seems specific to the FreeBSD port and/or system.

think firefox keeps a pool of already used sockets with http2

and dont check them all the time

Those are good ideas as to the situation.

I swear I'm not a maschoist but I've installed FreeBSD 13.2 on my gaming computer. Let's see how soon I have a stroke trying to get games to work haha.

Just upgraded from 12 to 13, and freebsd-install told me I have to reinstall all my ports to complete things. I realize I want current tools, but is this strictly necessary? How do I know if I have older ports that are affecting the load of the kernel?

strictly... maybe. highly recommened... yes. unsupported? Most likely.

ghoti, The binary packages are compiled using shared libraries from base. But when base is upgraded some of those libraries are upgraded, old ones removed, new ones added. Various ports using libs that have been removed can't work anymore. That's the reason for the need to resinstall.

ghoti: alternatively you can install the compat* package for the previous release if it exists

ghoti: freshports.org/misc/compat12x -- but also, note that packages likely won't change dramatically on upgrade anyways. all releases' quarterly and latest are build from the same branches

Title: FreshPorts -- misc/compat12x: Convenience package to install the compat12x libraries

hi, trying to find what package has qmelt? It's not in shotcut

In the last upgrade cycle I accidentally locked myself out by having switched my shell to bash, a port, and bash would core dump due to missing libs from base. I rescued it with a previous Boot Environment and fixed things that way. But bash from 12 definitely did not work on 13. I didn't try the compat12x package.

the lesson is never upgrade

rwp: yeah, compat12x would likely solve that. there's always a chance there's some latency before the package is updated as it needs to be, but 12.x is now two branches back-- any new library bump would put the latest version in compat13x

Yes. And that's why people become hesitant to upgrade. Once burned is twice shy. But instead I suggest learning how to upgrade successfully.

I installed shotcut but no qmelt. tried some commands but I can't see which package provides it? also have some source code here but I don't know the incantation to compile it

(which is why you actually neeed each compat*x package that exists from the version you upgraded from to one before the version you upgraded to)

last time my system was unbootable was from upgrading zfs. somewhere in the line of boot magic something did not grok new zfs. I wrote installer to usb and someone here gave me the incantations to fix things up. next install I will probably do ufs or whatever on / for simplicity sake

kevans, In my case I booted the previous Boot Environment, then fixed my shell in the DEFAULT one, then booted default, then finished the upgrade by pkg upgrade -f which worked trivially easy and my problem was solved.

*nod*

it's always good to have a recovery plan in place

even if that plan is "revert as best as we can and figure it out"T

It was my own fault. I got out of sequence in the upgrade. I rebooted too early. Oops. But that was the experience that taught me how much I appreciate Boot Environments. So it was well worth it.

And also that I had removed all options to login from a shell that was in base. That was actually my first mistake.

Now I always enable the toor user and use it. I leave root defaulting to csh. I use toor routinely which keeps it in my routine use memory (otherwise I had forgotten about it).

I got a little further in compiling : Project ERROR: mlt-framework development package not found

is there any terminal (tty or xterm) setting/software you can get some of the extended characters you used to see in DOS, etc.? For example in the games Rogue, Moria, NetHack, etc. (dungeon games) you used to see walls as a solid white block. All I've seen on various UNIX is they substitute a '#' for the block and it looks much worse

SDL isn't a good substitute because you can't copy & paste

darwin: foot shows nethack characters just fine; which terminals have you tried

just defaults

but it has been a few years; lately I use FreeBSD (and other *BSD UNIX) servers but haven't played games on it or other desktop stuff lately

so I used pure tty, plain xterm, and KDE konsole... maybe I was just thinking of a problem I saw on similar 'clone' (not 'genetic UNIX') OS that only have this problem

foot, or gnome-terminal should work

it could be your TERM setting, also

ok; thanks

does foot come with a certain X window manager (WM) or desktop environment (DE)?

darwin: foot is the default terminal for sway (wayland)

nethack also shows fine with alacritty,

darwin: it doesn't, it's just a terminal emulator for wayland compositors.

darwin: that being said, some wayland compositors do ship with menu entries or keyboard shortcuts that will spawn foot, but that's simply because it's a sane choice.

thanks

after a decade of running FreeBSD, it's time I learned how to type the euro symbol € on my US international keyboard...

I only need it in X, what's the best way to do this?

I might add 3 or 4 more symbols once I learn the secret incantations

I can do ctrl-shit-U and then type the unicode character code 20ac

which works

at least in my terminal, but not in browsers where I normally need it

lol control-shift-U, don't try that at home please

oh perfect, `setxkbmap -option eurosign:e -option lv3:ralt_switch us` works!

can't believe I waited a decade to do this and copy-pasted every damn time

is there a way to do this generically? I use FR & DE enough to make that useful

if you have a WM or DE sometimes they let you use a keyboard mode so you get the international keys. I suggest USA international keyboard with dead keys but you might prefer another

then it has an easier keystroke to get it... like alt something, not having to type in the unicode

dch: compose key?

I created my own custom XkbLayout

I have an alt-gr key but evidently I do not use it yet

also dead keys is a word I do not yet understand

Compose = e ⇒ €

whats with the = ? does one type that too?

I write most of my funny letters with compose, and only switch layouts when I need a full-on different alphabet, like Cyrillic.

ő

So Compose ' c gives me ć

ok so how does one discover (for example) e-acute or c-cedilla, or etc? is there a secret table?

é

ok

è

ä

I'm almost done

â

ç

dead keys are a concept from typewriters, you press down a key and it doesn't move back up until you press a 2nd one. it'll then modify the 2nd key

ah, that makes sense.

bien, ça suffit maintenant, j'ai rien d'autre à découvrir

zwr: that is a useful piece of lore, thanks!

in a computer keyboard, "dead keys" are typically accentuation characters, you press them and no text is inserted into the program. It's only when you press a letter that you get the accented letter

so alt-gr indicates "dead key mode", then one of ,`'= and a few others, then the required key

I can probably almost remember this

no

in the US international keyboard, ` becomes a dead key

so these are different things?

dch: cgit.freedesktop.org/xorg/lib/libX1…1/plain/nls/en_US.UTF-8/Compose.pre

as you can guess from that, you can make your add your own compose key sequences

in the US international keyboard with no dead keys, all the keys except AltGr behave exactly the same

when you press AltGr in the US-intl-altgr layout, it becomes a new shift level

meena: thanks, now I even have a local reference /usr/local/lib/X11/locale/en_US.UTF-8/Compose was installed by package libX11-1.8.6_1,1

TLDR the word I didn't know to search for is "compose" and now these things are easy to find.

I have a table to look things up in, too

this wikipedia table is prettier: en.wikipedia.org/wiki/Compose_key#Common_compose_combinations but the one you found is more complete.

Title: Compose key - Wikipedia

PąŔŧy time

the new shift level in the us-intl-altgr keyboard is how you type accentuation in that character. I think there's an euro somewhere too

altgr-5 in the us-intl-altgr

alt-gr = e works too

it's common for programmers who speak languages with that fancy accentuation stuff to use the US International keyboard with no dead keys because you can type faster if ^, ~, etc aren't dead keys

Also, getting a compose key on macOS took me almost 4 months…

you have to buy them ?

setxkbmap us -variant altgr-intl -option compose:menu

babz: probably, I don't remember any more… it involved a perl script, that much i do remember

seriously

you need karabiner-elements.pqrs.org and then you need to transform cgit.freedesktop.org/xorg/lib/libX1…1/plain/nls/en_US.UTF-8/Compose.pre into a format that it'll understand

Title: Karabiner-Elements

perl script form here bob.cakebox.net/osxcompose.php

Title: Bob Kåres homepage

might have to touch it up

Title: Chapter 26. Updating and Upgrading FreeBSD | FreeBSD Documentation Portal

the part about running `# freebsd-update install` is in a "Note"

but I am almost certain that before rebooting you need to actually run freebsd-update install

The "Note" makes it look like a side-note which is just elaborating something which was just explained but you never actually get asked to run `freebsd-update install` after `freebsd-update ... upgrade` and before rebooting if you just ignore it.

Am I correct in assuming this is not _just_ a "Note" and is in fact a required step?

tk: yes, that should be either a WARNING or be an actual element

Thanks

also it appears not to mention *anything* about boot loader updates (bios boot blocks or efi partition) which is probably not good

Ah, sounds ominous, if it did mention anything, what would it mention if you don't mind me asking?

I don't think I've ever had any trouble following these steps in the past (running freebsd-update ... upgrade then install then reboot then install)

But I have a rather vanilla configuration

iirc there were a breaking change when switching to zfs 2.0 that required upgrading the bootloader

But I am assuming the switch to ZFS 2.0 is not happening between 13.1 and 13.2?

nope

okay, looks like docs.freebsd.org/en/books/handbook/zfs covers the bootloader stuff

Title: Chapter 22. The Z File System (ZFS) | FreeBSD Documentation Portal

Anyway, thanks for the help.

Latest portmaster run pulled in OpenSSL 3.0. Now I' I cannot build ports anymore, several program complaining about "libcrypto.so.11 not found". Could anybody please advice me how to rebuild all ports depending on security/openssl with portmaster?

> several program complaining about "libcrypto.so.11 not found"

which ones ?

you have a statically linked 'fetch' in /rescue

e.g. PHP82

There is no file /rescure/fetch on my machine.

wht ports needs php82 to build ?

O.o

"libcrypto.so.11 not found" appears both with some running programs like e.g. php and when compiling ports like e.g. mariadb-server 10.11.

Hey all, I'm a new to FreeBSD (had used it a loooooooong time ago, so now back to noob) and am struggling with the installation of FreeBSD on a server. Got some help from the mailing list, but if someone here can help further will appreciate.

The issue: the server is quite a stupid one Xyratex/Kontron with SAS drives connected to a SAS controller from which is NOT possible to boot. For that reason there is a flash card inside to boot from.

The question: How to install FreeBSD?

Background: I've being able to install Linux on that machine and boot successfully with the SAS drives in RAID6 and the /boot located on the flash card.

So anyone willing to waste some time on that? :)

From the mailing list was suggested to use the mini-memstick to flash the flash card and i've done that. Now I'm able to boot through it, but I'm in the installation menu and back to how to partition the drives.

Since you can only boot from the flash card then it's Hobson's Choice, Take that or none. Which simplifies things by removing your choices.

Install to the flash. I personally would use Root on ZFS for it. And then after getting the system installed then I would install ZFS on the SAS drives.

but the flash card is quite small, it's only 4G and I'm pretty sure that it will not be enough for the installation.

Is it possible to have only the /boot part of the installation on the flash card and the rest on the SAS drives with ZFS?

as it is done in Linux (I've some XP there and was able to install it without problems in this way). I suppose this is pretty trivial in BSD, but ...... am lacking all the knowledge how to do that on terminal and the "auto isntaller" doesn't provide such choices.

I am right there with you in lacking the experience handling too small of a boot part. What you are asking about is perfectly reasonable. I am simply lacking in the answer.

got it. Despite that, thank you for your willingness to help. :)

Also unfortunately I have a work meeting starting in one minute. But hang out here a little longer. There are a lot of knowledgeable people in the channel.

Also if you have not yet read it then read through Chapter 15. The FreeBSD Booting Process docs.freebsd.org/en/books/handbook/boot in the handbook. It might already cover what you need.

Title: Chapter 15. The FreeBSD Booting Process | FreeBSD Documentation Portal

I've read that, but cannot figure out what exactly to configure

got the 3 stages of the boot

I think 4GB should be enought for the base system (but not sure about that) and then you could put /usr/local and /usr/home on another zpool (and /usr/ports if you want to do something with ports)

I've tried to make manual config on the flash card, as suggested in the mailing list with:

gpart create -s GPT da0

gpart bootcode -b /boot/pmbr da0

gpart add -b 40 -s 512 -t freebsd-boot da0

gpart bootcode -p /boot/gptzfsboot -i 1 da0

and then to install the system on the SAS drives, but after reboot it dies with:

>> FreeBSD/i386 BOOT

Default: 0:ad(0,a)/boot/loader

boot:

which according to Chapter 15 is dying on stage 2

Title: boot.config(5)

nimaje: that way if I lose the card I'll need to reinstall the entire system and its configs. I'm trying to keep only the bare minimum to the flash and eventually to have it on the SAS drives, so in case of dead flash I can just reimage it from one of the drives

Thanks VimDiesel and babz. Reading over this and boot(8)

Hopefully will understand something :)

when you see that line

boot:

press any key to interrupt the default boot sequence

you should get a prompt

linux just gets better and better. lunduke.locals.com/post/4740497/gno…al-shaman-as-new-executive-director

Title: Shared post - GNOME Foundation hires "Professional Shaman" as new Executive Director

as it gets better and better I prefer freebsd more and more.

further off topic, but the Silicon Valley show is hilarious, and that made me think of the CEO of Hoogle

I see a host with 120 connections to the mysql server. sockstat shows me it's all ruby27 - it is easy to track that down to a given PID? ie. this connection is from this PID?

sockstat is showing the pid, surely?

Hello! I would have a question how to manage packages and ports respectively. I heard that one should not mix packages and ports, because it could lead to problems. But how am I supposed to not mix them if some programs are only available as ports but not as packages? What am I supposed to do if that happens?

rwp, kevans: last night I asked about package upgrades being required when I upgrade from 12 to 13. You both responded, but I fell asleep. Thank you, and sorry for disappearing. :) I *have* compat12x installed, but I guess I should reinstall packages anyway in case there are ABI changes that are important.

lockna, I think you should be fine provided you stick to the default options for what you can't install as packages. But I'm not an expert.

lockna: sometimes, newer packages are just not available in the quarterly package set. If something is not available as a package at all, I'd just advise that you do what you need to do, and if a package turns up in the future, use it if you can.

Many of the warnings about mixing have to do with versions of dependencies. I.e. you upgrade a library that a port uses, and suddenly a bunch of packages don't have that library anymore.

presumably you can rebuild the port and it'll be back and ready

Ahh, okay. And how do I go about updating ports? since, some again said I should avoid portmaster, but portmaster and portupgrade were the only thing mentioned in the handbook. So, I'm not sure if I should use them now or not

ghoti, Sleep is good! Much restorative. Highly recommended. :-)

ghoti, kevans is of course the expert far above myself. But I see the compat* packages as a way to be able to run binary executables verbatim that were compiled for the previous OS releases for which one might not have the source code for compile immediately available.

lockna: `cd /usr/ports/category/portname; make install clean` is the usual advice. You can also use tools like ports-mgmt/portmaster.

That's definitely a good thing to be able to do. One can upgrade the OS and still run some legacy binary executable via installing a compat* package.

However if one has the source then having a newer compilation of that source to link with the current shared libraries is a preferred solution.

Okay, okay, one last question. Is it normal when I use portmaster -a to update all my ports and it does update quite a lot, where I more or less does build things like chromium from scratch? I found that kinda odd

synth is pretty cool it builds all your ports in a clean environment and creates it's own repo to install them once all the builds are complete

lockna: problems come mostly from stuff going out of sync (which happens easily as building packages takes a while especially if you build the whole ports tree, so when the official repos are updated the commit it was build from is likely some days old) and changing options and install some port but install packages that depend on it from the official repos, which are build against the dependency

with default options, so as long as you only build leafes you should only occasionally need to rebuild because of changed libs

lockna: therein lies the rub. If you are upgrading dependencies with portmaster, the alread-install binary packages will either have to get updated as ports (on which packages are based) or break.

lockna: if you have multiple machines or jails, you might consider poudriere, too.

rwp: thanks, yes, I've been using compat* to handle old binaries for years, but this was the first time I had seen that warning from `pkg`.

Ahh, okay. Yeah, it tried to update my packages as ports. Some things even broke. I updated via pkg afterwards where firefox was about to be deleted. So I had to reinstall it. VSCode was also removed, but I'm not able to reinstall, just via ports and build. When I do that, it wants to build electron25 again too. Much to figure out I still have

I have read about poudriere already, but was taking a step back, since I'm only using a single machine

lockna: I think you're seeing first-hand why it's not recommended to mix packages and ports. :)

What's stopping you from just using ports?

The compile time, since, well, takes a long time. Download takes some time, cause I only got about 8 Mbit and then compiling.

Well... Source downloads I would expect are smaller than binary package downloads. Maybe the compile time is less than the repair time.

Certainly it's less hands-on.

fair enough

ghoti, Tidbit: The pkg binary *itself* might be broken by an OS upgrade. That's why there is a pkg-static shipped in base. pkg-static can be used to bootstrap pkg in the case that a release upgrade breaks pkg which is itself a /usr/local/bin/pkg binary package.

rwp: yes, I have been known to `pkg-static bootstrap -f` in the past. Very handy to have static versions of things.

The software layers and permutations and combinations possible can be confusing as noted above by the discussion of mixing ports and pkgs but it enables the robust handling of possible problems as a general statement about the sum total of the parts.

And, oops, did I say pkg-static was shipped in base? That's wrong. /usr/local/bin/pkg-static is part of the pkg package. My bad!

And as with everything, if you get things to a state where stuff works, stop poking at it. :-)

That is time honored advice! :-)

rwp: Hmm, ya, no /rescue/pkg, but there is /usr/sbin/pkg in base which can bootstrap, presumably using whatever libraries you installed into base at the same time.

man pkg(7)

Yep. I mis-spoke there when I said base. "Missed it by _that_ much."

@isley I love synth.

Any information on quicksync for melt / ffmpeg on freebsd? not finding much

ok I think I need drm-kmod -- but that installs a bunch of packages. not sure which of the intel packages I need. there's lots of lakes ... but this is for haswell

I grabbed nbdkit from packages and running it spits out a sad `ld-elf.so.1: /usr/local/lib/libtasn1.so.6: Undefined symbol "strverscmp@FBSD_1.7"` which has me confused at how this could happen

:)

hmm I have dri devices now but not sure if ffmpeg includes qsv support

ffmodprobe spits out all the ffmpeg options it supports i think

ffprobe

that checks the format of a file

ffkldload on freebsd /s

yes, and it spits out all the things that are enabled or disabled

actually, i think ffmpeg does that too

ffmpeg -encoders 2>&1 | grep qsv

scoobybejesus: ffmpeg -codecs lists codecs. missing everything qsv. braindead.

comparing to alpine linux ffmpeg -codecs output, lots of *qsv in output

ah okay, didn't realize. very handy tip

i think it's based on a closed source SDK released only for windows and linux

(and now deprecated)

lots of the drm bits seem to be closed firmware

deprecated? how so? it's widely used

plex emby jellyfin ffmepg da vinci resolve shotcut etc etc

lol

babz: oneVPL version 2.x is an API that is the successor to Intel® Media SDK. The Intel Media SDK project is no longer active and will not be updated.

is this what you are referring to?

yup

do you know how far back the new stuff supports? after I build am I going to be encoding with *qsv or is it called something else?

try the VPL option while building the port

I downloaded the ffmpeg source directly. reading someone saying building it from ports is going to try to build all dependencies too

ok now I need to get libvpl then ... seems --enable-qsv is no longer an option in ffmpeg configure

k there's a package for onevpl ... configure is OK make is pissed lots of "make: "/tmp/src/ffmpeg/ffbuild/config.mak" line 2: Invalid line type"

gmake is making things happen now

probably missing a lot of shit I want but let's see if this basic build will go

./ffmpeg -codecs |grep qsv has the good bits yay

hopefully it doesn't shit itself when I drop it in

pkg lock to prevent upgrades overwriting the newly build ffmpeg

still can't encode yet ... Generic error in an external library

there's some samples apps installed with onevpl, can't get those to work either

computers are dumb

[ERROR], sts=MFX_ERR_NOT_FOUND(-9) which is similar to the error ffmpeg is throwing

drmn0: [drm] Cannot find any crtc or sizes ... I wonder if it is a requirement for an actual display to be running? I vaguely recall I had to set up some dummy monitor thing on winodws 7 before to get intel quick sync encoding to work for OBS game streaming

hey kids if anyone else is trying this bullshit, DO NOT try to kldunload i915kms !

Title: Reproducible panic on "kldunload i915kms" as of r236822

running out of patience to try to make this work. I guess I will remote mount and do the encode from windows or linux.

this is almost enough to make one consider moving fileserver to linux

bah!