Is there a simple way (simple command) to turn a root on zfs created zpool from the installer to a mirror?

freebsd-src/usr.sbin/bsdinstall/scripts/zfsboot might be the best thing to follow.

skered, zpool attach pool old_device new_device

I'm not sure it's going to be that simple.

skered, it is

Need to parition and play with swap

although it's zpool attach pool device new_device

what disk name?

old and new

> If device is not currently part of a

mirrored configuration, device automatically transforms into a two-way

mirror of device and new_device.

The issue is to mimic the install as if zroot is a mirror with swap mirrored too.

And maybe other issues with bootcode and efi boot?

so you want to recreate the partition layout of the original disk too?

what disk name? old and new

VVD: nda0 (existing) nda1 (new).

er well nda0p4

gpart show nda0 nda1

Sorry. Had to setup gh-gist.

Title: gist:e67c825ee0057029c91287ed6973da20 · GitHub

why is perl hardcoded...

sigh

skered, gpart show nda0 nda1

seems like the new device isn't nda1 ("gpart: No such geom: nda1.")

Title: ZFS - [HOWTO] Convert Single disk ZFS-On-Root to Mirror | The FreeBSD Forums

nmz: a report of something similar suggested there was a problem with the reporters perl setup bugs.freebsd.org/bugzilla/show_bug.cgi?id=275549 so maybe try reinstalling perl?

Title: 275549 – devel/p5-Locale-libintl requires a full specific and larger version of perl5.

then the requires are f-ed?

nah, I'm just gonna ln -s perl to perl35

its functionally the same no?

hm, pkg install -f perl5 should just reinstall perl without affecting the rest of the system

(why is default version handling of perl like that?)

freebsd is so bizarre

there is no wrapper for tclsh either

hm? I have some tclsh8.6 installed by tcl86, do you search that?

lua installs as lua54 instead of lua5.4 as well

its fine, I just had to fix the hashbang

yep, it did nothing

nimaje: It is. Just means it isn't in use.

VVD: This forum post looks good.

Thanks

Why would you use geom-mirror for swap and not a zfs volume?

zfsscript in the install is using geom though.

zfs has some memory overhead you wouldn't want for swap

Seems like you can disable features such that isn't an issue however there are other features that might not work if it's a volume.

Oh well. I guess we can't be that fancy.

the waiting game for 2 past hour is painful

Is a "process group" all of the child processes under a parent process?

(ah that's what the "PG" stand for in PGID)

Okay, so the kernel manages each process into segments of immutable machine code text and mutable data and stack. Even though the latter two are different memory segments, doesn't the data exist in each layer of a stack or does the data get allocated in its own different area in memory?

Are "system pages" the way the kernel manages the dynamic allocation of process such as a heap? I guess that would explain why the heap is slower since pages are allocated else where from the memory segments.

voy4g3r2: what are you waiting for?

good people of #freebsd, hello ;/

after every boot I got from ntpd: "giving up resolving host pool.ntp.org: Address family for hostname not supported (1)"

any idea?

what does `host pool.ntp.org` say?

host shows me normally ip addresses

morena, Does "ntpq -pn" shows anything?

No association ID's returned

Thanks. (That clears my suspicion of ntpd starting before a network connection is available)

it shows in console pretty late, already while "login:" input is availavle few seconds

Is ntpd currently running?

yes

I have termbin.com/vuan in /etc/ntp.conf (output of "awk '/^[^#]+/' /etc/ntp.conf")

Oh! I am using "{0,2}.freebsd.pool.ntp.org", not "pool.ntp.org".

ye, I changed just freebsd to pool.ntp.org after that issue and bc. of high ping

Ah

and instead pool replaced with server

but nothing changede

changed

with pool instead of server

I get not giving up resolving host ...

but "error resolving host ....

if I replace server with pool infront url

can't be that related to ipv6 on a system without it?

but I tried ntpd flag -4

when I try like service ntpd restart I don't get any error

also I get ntpd starting later

part in /var/log/messages termbin.com/1v1b

I have just ntpd_enable="YES" in rc.conf

morena: Can you add: ntpd_flags="-4" in rc.conf, restart and check it again?

Also I'd suggest having: ntpd_sync_on_start="YES" and ntp_leapfile_fetch_verbose="YES" too.

I tired -4 flag before, nothing changed, but will will to add all you mentioned

tried

morena: Are you sure it was exactly: ntpd_flags="-4"

the same error now

with all above flags

Hey all! Bought a new server and have a question. If I move the old disk to the new hardware and do zfs import will it import the zroot for /usr/local/etc /home and such? Or do I need to backup everything with rsync?

when I disable ntpd and start it manualy after login, then there is no login

error

I have static ip, simple as in freebsd doc. dhcp and resolv not enabled

maybe I can try local-unbound

anyone can help with some dtrace?

is netif running?

here's the code that I need to dtrace

Title: online/net/Socket.cpp at master · CollaboraOnline/online · GitHub

specifically this line

rc = ::bind(getFD(), (const sockaddr *)&addrunix, sizeof(struct sockaddr_un));

with local-inbound default resursive (I assume) it does not show more that error

nmz netif is enabled at least it shows with 'service -e' but no more info about it

morena: local unbound would help

tercaL ye, I wanted to use it anyway

if did not touch any setting, just resolv.conf to 127.0.0.1 does it by default resursive into root servers?

no any forward nonsense?

with local_unbound I got also 'edns0' into resolv.conf

is that udp thing in dns?

what may be the issue before as pool send more ip which may be in udp not tcp?

something like that I remember with alpine musl as it did not use udp, so some server when sent many ip in udp it did not resolve at all

edns0 permits groups of udp packets; needed when a dns response exceeds a single udp mtu

you still have to go to tcp eventually

okay

I want to test a portupgrade in poudriere. Since I want to reduce runtime and only test my port, I would like to use packages for _all_ dependencies of the port during "poudriere testport". Is there any way to tell poudriere to not build the dependencies, but to install them as package?

ok, poudriere option "-b latest" seems to do what I want... Let's see how it finishes

skered, I'm not using geom-mirror for swap on my hosts. I'm using 2 plain swap partitions - "stripe".

how likely is it for me to get a pull-request accepted for a change in ppc_pci.c and uart_bus_pci.c.. Is it worth the hazzle?

drobban: procedurally, the github pull request system (assuming that's what you're referring to) seems to work fairly well, so i think the chances of it being affected would depend on the actual change

s/affected/accepted/

it would probably help to open a PR describing the issue and put the pull request url in there, which makes it more likely someone responsible for that code would see it

(if there's anyone left who is responsible for ppc, anyway :-)

unixwitch: I see, will give it a shot. Humm.. perhaps I should try to roll my own system in dev mode and test my change on that before I send my commit in a PR.

drobban: btw, when i said "open a PR" there, i mean a freebsd bug, not a github pull request... kind of confusing that the same acronym is used for both. basically open a bug, then open a pull request for the fix, then put the pull request url on the bug

PR stands for problem report here

It stands for Puerto Rico here.

WTF: bpa.st/PWUQ

Title: View paste PWUQ

network problems?

Nothing apparent - name resolution works can ping the host. . .

A sniff reveals I'm getting 404s.

Oh. . .

12.4-RELEASE-p4

That's probably why.

Is there a way to do a git clone of just the OS version you're tracking, vs. the head/worktree method people seem to be pushing?

you mean you want to directly clone releng/X.Y?

'Zactly.

CrtxReavr: "git clone -b releng/14.0 git.freebsd.org/src.git"? and add --single-branch if you want to avoid fetching the data for other branches you don't want

Title: src - FreeBSD source tree

PR, like many abbreviations, is context-sensitive and means a lot of things, so it's up to people to use clues, as there's no point in discussing which is "right".

And in other news, water is wet. ;)

unixwitch: ppc64 still gets packages built

debdrup: this is about ppc(4), the parallel port driver

Welp!

unixwitch, tell me again what was wrong with svnup?

in non-technical spheres, it can mean "public relations" ;)

ISA support is gone (which won't matter, since it's ISA and PCI is handled the same via newbus), but since it's part of the base system and not marked for deprecation in 15-CURRENT so far as I can spot (someone might check for Giant cf. locking(9)?), so a problem report seems like a good idea.

ZedHedTed: even in technical spheres, it can mean that - because you can't divorce technical from political.

Or, maybe you can, if there's only one person involved.. :P

heh, true. just look at RMS. :P

I'd rather not, if it's all the same to you ;)

CrtxReavr: i've been here since CVS and out of that, svn, and git, i find git the most useful... i think a more interesting question is which of git, darcs or hg is better :-)

RMS has often made the nerd news for his uncouth behavior. . . though I used to work with a lot of people who experienced it first hand.

unixwitch, yeah. . . used cvsup

CrtxReavr: well, svn is _eventually_ (as in, some time after all relevant branches that can use it, have been EOL'd) - so at some point it won't work anymore.

debdrup, the FreeBSD SVN servers have been taken offline.

i remember netbsd didn't want to move to git because some people there thought it was unsuitable for large repositories, but it seems to be working okay in freebsd. although this discussion about the commit with a wrong date is interesting...

CrtxReavr: svnweb is still up, as far as I can see.

unixwitch: back in the day, git absolutely wasn't suited. When FreeBSD was added to a git repo, it'd just cause git to be sad and hang any number of ways - that's why it took forever for FreeBSD to move to it.

NetBSD, I believe, are going with Mercurial?

hehe

Few months ago, I was in this quaint little tourist down about an hour from here.

i remember opensolaris used hg, and back then it didn't come with any built-in way to merge branches, which was... weird. that was probably >10 years ago though so i imagine it's improved

I see this old guy in a NetBSD shirt, walking with his wife.

As I passed them I said "someone has a bunch of old computers at home that no one else wants."

His wife started pointing at him vigorously and persiing her lips.

The guy laughed and shook his head. . . said "Well, if you're going to just call me out like that."

I said, well, you're the one wearing the shirt.

Seems fair.

It's a bunch of stuff, mostly retro, that _isn't_ in my rack.

i have a bunch of old sparc/parisc/alpha stuff lying around that i'd like to use, but nowadays it's hard to justify the electricity bill...

My rack is full, but the electricity cost certainly is also a reason to not run it..

I can't quite make myself throw any of it away, though.

I can _just_ about justify a HP Microserver Gen10+ for all the stuff I need available on the network (it hosts both nfs and iSCSI via ctld(8) because all my other regular machines only have NVMe SSDs that aren't _that_ big), but the backup server with +200TB storage and all the other stuff in the rack is turned off unless it's needed (which is handled by WoL or IPMI).

i don't have any computers that will boot netbsf.

s/sf/sd/

Is that even possible?

maybe a modern amd64 system with some weird hardware that makes the kernel panic?

depends on how you interpret that 'will', if you interpret it as 'could' then I'm sceptical too if that statement is true, if you interpret it as 'at some point in the future' then maybe

On 10-CURRENT?

nimaje: true.

if this wasn't #freebsd it could also be someone who only owns a smartphone, but that seems unlikely here :-)

I would think that wouuld be true for most of Libera..

unixwitch: irc with a smartphone is really painful. typing in general is painful on a smartphone if you ask me

Hello. I just upgraded from 13.2 to 14.0 using the release / install notes. After reboot, when I try to run "pkg" I get the following error: ld-elf.so.1: Shared object "libssl.so.111" not found, required by "pkg"

use pkg-static to upgrade pkg

thank you

Was there anything I did wrong in the upgrade to cause this? Trying to determine what had caused this

run freebsd-update 3rd time?

When I ran it the third time, it said everything was up to date

depending on which pkg it is, freebsd-update wouldn't help

RoyalYork: you're supposed to updates your packages after the second freebsd-update install and before the third run

presumably it's the ports pkg though, and you'd just have to know to upgrade it with `pkg-static` or `pkg bootstrap -f`

because the third run removes the shared libraries from the old release, which packages will require

pkg bootstrap -f had worked

thank you

unixwitch, I guess i missed that

I will read it the notes more thoroughly in future

maybe worth adding that you need to use 'pkg upgrade -f' here, to make sure it reinstalls all the packages for the new release even if the version numbers haven't changed

Im showing userland as 14.0-RELEASE-p3 and kernel as p4 - Is this all correct?

vice cersa

shuld be the other way aro-

yeah

great

should, even. -p4 was just an openssh change

it might not be a bad idea for freebsd-update to print a warning about this, since it's not an uncommon problem...

hm, I throught pkg should detect the change in ABI, so that -f isn't needed?

hm, maybe. it's been a while since i did that

Thanks for your help everyone

unixwitch: it's usually a peripheral like the laptop camera that it stops on. but even after i disconnected the camera, it stopped over a different part. so instead of taking the laptop apart further, i tried openbsd and that booted fine.

'it' meaning the boot process.

is anyone else getting an error building current

yacc

error code 134

wasn't sure if it was something to do with a previous installworld

[Gentoo]: incremental buildworld of 8e36732e6eb50976ecc139a9c499f1396d340d1a ("systm: Annotate copyin() and related functions with __result_use_check") is fine for me on amd64. haven't tried a full build for a while.

make + make install in usr.bin/yacc is also fine

is there anything in dmesg? (sometimes things that cause SIGABRT might log there)

i recently did a buildworld and installworld the other day so im very recent now it wont build

wonder if it could be something src.conf or related

do you have CPUTYPE or something like that set in src.conf?

Title: View paste WTUA

Title: View paste M2VA

^ my kernel config

the only thing that stands out to be there iw WITH_RETPOLINE, did you add that recently or did it work before?

s/iw/is

worked before

hm, not sure then :-/ someone on freebsd-current might know

ok, thanks

can anyone recommend an RDP (the Windows protocol) client which is in ports, works on Wayland, and handles HiDPI scaling properly (by doing it on the server)? Remmina doesn't do the last bit, FreeRDP claims to with /smart-sizing but it doesn't scale properly, it just scales the bitmap

ah... maybe freerdp does do it once you find the right combination of 3 different scaling options

yeah, freerdp should do that but it probably requires experimentation

i wish people who post on freebsd mailing lists would learn to trim quotes... so bored of pressing space three times just to get to the actual content of the mail. </rant>

Agreed!

unixwitch, try FreeRDP 3.1.0: reviews.freebsd.org/D43127

Title: ⚙ D43127 net/freerdp3: new port 3.1.0 + fix net/remmina to build with installed net/freerdp3

VVD: i got it working in 2.11.4 in the end, it turns out /scale-desktop:150 /f does what i want

now i just need to script something to prompt for the password

cron execution question here: I am working through (still can't get a snapshot) zfs-snapshot utility through packages, i have added an hourly job to the crontab and when it executes at 2 past hour.. all the output i get is: bsd.to/J0BA

Title: dpaste/J0BA (Plain Text)

and my periodic.conf just never seems to run.. but when i run the command OUTSIDE of both of these methods i can get it to run.. bsd.to/UgGU

Title: dpaste/UgGU (Plain Text)

i guess what i am asking for, is a second set of eyes.. is there something "missing"

Where did you locate the crontab file you set up for it?

I used /etc/cron.d/zfs-auto-snap-local as the file location here for example.

But /usr/local/etc/cron.d/foo would also be equivalent.

i just do a crontab -e as root

let me find exactly where that went

i assumed it was right as you can see the entry where it tries to execute periodic hourly

Hmm... I suggest that you use 'crontab -e' for personal non-root crontabs and use a whole file for system level root crontabs. It's easier by a lot in the long run.

Your root crontab -e will have gone in /var/cron/tabs/root but that is not important.

i use /etc/crontab for root's crontab, i find that easier than having them scattered around different files. (ymmv though)

yup, the file has an entry

Note that the syntax is different between a personal crontab and a system crontab by the extra field to specify the uid running the task.

voy4g3r2: did you forward root's mail somewhere in /etc/aliases? otherwise it'll go to /var/mail/root (you can run 'mail' as root to read that, or mutt or whatever)

i'm assuming the problem here is that command isn't working... if it does work i wouldn't expect any mail output

unixwitch, The paste shown seemed to not show any task being run from cron other than atrun and periodic hourly.

yeah i just noticed that

voy4g3r2: what exactly did you put in crontab -e?

(crontab -l should print the current crontab)

I prefer to avoid needing to merge files on upgrades so I always prefer to put my own tasks in a separate file in the /etc/cron.d/ directory.

Title: dpaste/xgRE (Plain Text)

this is whaat the package recommended to do

remove 'root', you don't specify the user in a personal crontab

that syntax is for system-wide crontab files like /etc/crontab or cron.d

also is 'periodic hourly' a thing? i guess this is something zfs-snapshot adds

from what i can read, from stepping through zfs-snapshot, it requries the periodic(8) subsystem and to pass the "frequency" hourly daily weekly monthly

voy4g3r2: fwiw, i would have expected that erroneous 'root' to generate an error mail, so i still recommend checking /etc/aliases to make sure root's mail is going somewhere

echo "my⊙ra" > /root/.forward

looks like everything is going to root but nothing in the root mailbox

ohh here we go just cat /var/mail/root

and there it is ....

.. /bin/sh: root: not found every 2 minutes past hour :)

reading mail with cat, that's very 31337 :-D

well i do mail and it defaults to /var/mail/chrisdavidson

i do not have email setup on the box.. i can barely get these damn backups to work :)

voy4g3r2: at least set 'root: chrisdavidson' in /etc/aliases so you can read it as your normal user

getting remote mail delivery working is a bit more effort...

yes, i am stuck with my l33t gmail account.. can't seem to get rid of it

full name gmail accounts are unheard of now a days

Didn't Google delete a bunch of inactive GMail accounts by Dec. 31st of last year?

TommyC: i think i did read that somewhere

voy4g3r2: if you left it alone, your l33t gmail account is gone now

oh no, i use it all the time

since like 2004

\p

CrtxReavr, Google Gmail won't be accepting email from a randomly configured machine probably sitting on a home dynamic-ip network. The days of being able to forward email like that are long past.

Meanwhile... JFTR but all of my machines are configured to send email. It's the only way.

Might, with the right spf record, and maybe some smart-relay'n.

unixwitch, 1337% Pi ~= 42

Yes if someone does all of the right things then they will have email working. As many of us do. But those people are not here asking the questions. :-)

really? ilythia ~> echo 'scale=10; 1337 % 3.14159' | bc

.000000000150523

rwp: and i never want to know how to do it :)

It used to be that everyone ran their own email server. Not an unusual thing at all. It was the normal thing.

But people are the problem. And so now we have spammers and scammers.

rwp: my machines all have working email delivery but that's because i host my own mail, put hours of effort into making this work and pay £££ for an Internet connection i can send mail from... a bit much to expect from random users :-D

unixwitch, literal percent, not mod.

rwp: oh yeah, i setup a mail server in university.. on slackware linux and somehow made an oopsy.. the system administrator for penn state shut me down.. they were getting my root emails to their mail system and couldn't figure out what was wrong

I host all of my own mail. Also host other domains mail too. But it's not something that I think everyone should do these days.

CrtxReavr: idgi :-(

Unless someone *wants* to host their own email. In which case I think they definitely should do so. It's not forbidden. It is okay for everyone that wants to do so to host their own email.

basically they blocked my ip address from their mail server.. buggers never did tell me how to fix it, i got a RTFM response

yeah.. because sendmail is SOOO easy to configure (mind you i was 20 years younger and a college kid)

sendmail is very easy to configure if you don't use sendmail :-D

unixwitch, echo "scale=10; 1337 / 100 * 3.1415926539" | bc

even freebsd replaced it with dma (finally)

postfix is probably a lower learning curve nowadays.

+1 for Postfix. FTW!

unixwitch: yes, the only way i fixed it.. waas moving to postfix

When I run an actual MTA, postfix is my goto.

all i wanted was my university emails to goto my linux box dang nab it!

So well-documented and easy to setup.

i sort of with we'd got postfix in base already, but that would have been annoying when you need ports postfix to get ldap or whatever, and now you have two copies of postfix...

Yes, the postfix docs are excellent, if not a bit circular.

Postfix just had a birthday last month. It is 25 years old now.

thumbs, I'm not sure I've ever needed more than the comments in the default config file.

CrtxReavr: Sure, most of the shipped configuration is a good start.

I was referring to the self-referential nature of some of the online pages.

Most of it is very obvious once you understand that it is all table driven decisions. Every decision maps through some type of table. Match the LHS do the RHS.

But some things like setting up a custom slow turtle transport to communicate @gmail.com mail through a rate limited protocol is not quite so trivial and hints and documentation are useful.

hrm.. there is something else going on with this zfs-snapshot, i would of expected the daily periodic jobs to run.

Google likes to reject a lot of valid email saying that they are being abused but much of my spam email incoming is outgoing from Google!

voy4g3r2: what time is it where you are?

335PM

it is telling me, in man, that is runs "early morning"

daily runs at 3AM

(see /etc/crontab)

hrm.. i would of epxected SOME output in /storage/.zfs/snapshot then or any of the other datasets

i am getting ahead of myself, lets see what happens at 402pm

back to this excel and powerpoint hell

thank you all for the help.. we shall see

do you see the hourly command running in /var/log/cron now?

i missed the 2 minute past hour.. i gotta check at 402

Title: dpaste/CTDn (Plain Text)

that is the latest from that file

mail.core is new :)

mail.core sounds like you found a bug

it waas around the time i added root: chrisdavidson and ran the newaliases command afterwords

voy4g3r2: if you happen to have gdb installed, could you run "gdb /usr/bin/mail /path/to/mail.core" and then type "bt" at the (gdb) prompt? (this may not produce any useful output without debug symbols though...)

i guess you could also do this with lldb, can't remember if that uses the same syntax

if it is not installed, i can install it :)

this new machine is soo much nicer than that raspberry pi

ah, "lldb /usr/bin/mail --core /path/to/mail.core" apparently

(lldb is in base)

Title: dpaste/qzie (Plain Text)

that is with gdb

let me do it with lldb real quick

that's very unhelpful, i wonder how to make it load the debuginfo

i did the lldb one

is there any special command (yes i have no clue how to use these tools)\

# lldb /usr/bin/mail --core /var/cron/tabs/mail.core

(lldb) target create "/usr/bin/mail" --core "/var/cron/tabs/mail.core"

Core file '/var/cron/tabs/mail.core' (x86_64) was loaded.

voy4g3r2: bt

Title: dpaste/vL29 (Plain Text)

looks just as unhelpful as gdb one

yeah... there must be a way to make it read the debug symbols in /usr/lib/debug

i know i did not install debug package.. that i can recall

figure, i am not writing C code.. why do i need it

do you have anything in /usr/lib/debug?

i got the boot directory, nothing else

voy4g3r2: are you running a binary release (like from an iso or freebsd-update)?

i am running an iso.. i have upgrade to p4

FreeBSD momas.home 14.0-RELEASE-p3 FreeBSD 14.0-RELEASE-p3 #0: Mon Dec 11 05:07:37 UTC 2023 root⊙adn:/usr/obj/usr/src/arm64.aarch64/sys/GENERIC arm64

ohh wrong box

hold up

debug symbols are in base-dbg

FreeBSD harley.home 14.0-RELEASE-p3 FreeBSD 14.0-RELEASE-p3 #0: Mon Dec 11 04:56:01 UTC 2023 root⊙adn:/usr/obj/usr/src/amd64.amd64/sys/GENERIC amd64

voy4g3r2: would you mind gzipping the mail.core and uploading it somewhere (or if it's not too big, mail it to lexi⊙lo)? i can grab the debug symbols for that binary

to save you having to install it

sure i can do that

it looks like a meg in size

note, it might contain private data like the contents of the mail you were reading

I wouldnt personnaly send a core dump of a privileged deamon, but...

it's not a daemon, but yeah

it is literally a file server, that is it :) there is no top secret stufff in here

unixwitch, did you "gi" yet?

oh crap, 11 megs.. lets see if gmail does not like this

zip to the rescue

CrtxReavr: i don't know what this means

unixwitch: should be in your inbox within a matter of seconds

voy4g3r2, worst case look at the command 'split'

woohoo hourly populated!

voy4g3r2: yup ty

you are welcome

Title: dpaste/abm4 (Plain Text)

just beautiful!

unixwitch, surely you know the answer to the Ultimate Question of Life, the Universe, and Everything is 42?

now time to get zfs send | zfs recv these snapshots

unixwitch: i have to step away for a little, if you requrie anything else from me.. could you message me here or the email works too

thank you for your help.. maybe i may even need to submit a bug? that would be fun.. my first one

well i got a more useful backtrace, but it's not really clear why it crashed here: bsd.to/XcwC

Title: dpaste/XcwC (Plain Code)

aha

# mail root

Subject: ^C

^CBus error

unixwitch, Well look at that! I also reproduce that here on 13.2R-p9 with /usr/bin/mail here.

No need to be root. Reproduces as rwp just fine.

yeah, something really weird is going on there... now rebuilding /usr/bin/mail with ASAN :-)

==39955==ERROR: AddressSanitizer: stack-overflow on address 0x7fffffffd3ac (pc 0x7fffffffd3ac bp 0x7fffffffbe70 sp 0x7fffffffbab8 T0)

same on 14-stable

definitely a sanitizer

error error error LOCAL SYSTEM: UNLOCKED

if someone wanna build libc with -O0 -g3 ...

i'm pretty sure the problem is that in tty.c:99 it sets up a setjmp which is called by the SIGINT handler (ttyint) that gets triggered from inside the libc stdio code, which i guess it doesn't like

this code is so nasty

i didn't think that would cause so much problems

proposed patch: bsd.to/Aucf -- the problem is that both grabh() and collect() want to set up their own SIGINT handler, which ends up calling grabh's longjmp() when the environment is no longer valid

Title: dpaste/Aucf (Plain Code)

ilythia /s/m/u/mail (main)> mail root

Subject: ^C

(Interrupt -- one more to kill letter)

^C⏎ ilythia /s/m/u/mail (main)>

taht quick you had a fix?

holy moly

voy4g3r2: bugs.freebsd.org/bugzilla/show_bug.cgi?id=276119 (fix in the URL field)

Title: 276119 – mail: typing ^C twice during compose causes SIGBUS

i'm still trying to understand the problem

awesome

it's the SIGINT handler pointing to garbage ?

so basically my inability to type found a bug, sweet

babz: as far as i can tell (although tbh i find this code extremely hard to understand) the problem is that first collect() installs a setjmp() handler, then grabh() installs its own setjmp() handler, so the first ^C aborts grabh which returns to collect, then the second ^C uses the now-invalid setjmp environment of grabh (which is invalid because the function returned)

thx that makes sense

unixwitch: that mail bug is wild

meena: it is, and i'm really not convinced this patch is actually the right fix...

meena: i think this is the right fix -- bsd.to/Pe0N ... optimiser fucking us over again

Title: dpaste/Pe0N (Plain Code)

unixwitch: might be a better idea to try remove the complete setjmp/longjmp from the signalhandlers

satanist: that would be a good cleanup but also a much larger job

problem with longjmp out of a signal handler is that you get realy fast ub

also the code is fragile as fuck

i'm pretty sure this code is already doing a bunch of UB in its signal handlers, like updating non-volatile globals

but fixing all that would require a fairly substantial amount of rewriting

problem is after you have called longjmp out of a signal handler you are still limited to async-safe-function

voy4g3r2: this was quite a fun bug, and i learned something, thanks for finding it :-)

unixwitch: no problem, glad you enjoyed.. all hail the mighty if statement, is what i got out of reading the patch

the actual correct fix was a bit simpler freebsd/freebsd-src #993/commits/f720c7c64e718bd566baf5a733a1a6d50939b8c7

Title: mail: add volatile in grabh() by lexiwinter · Pull Request #993 · freebsd/freebsd-src · GitHub

i may have another somewhere.. for the love of god.. netatalk just "stops" working and have to do a restart

but it is high school basketball time

unixwitch: just beautiful

unixwitch, Cool! Good work on tracing down that bug! And I'll note in passing that /usr/bin/mail is also /usr/bin/mailx and /usr/bin/Mail too.

So it seems that 12.x install for UEFI installs for mirror installs setup both /boot/efi for both disks EFI paritions. In 13 (and 14) this is only done for the first disk. It seems to do this for geli/mbr... Was this an oversite or is the expectation this is just copied/setup post install?

I guess it's this. freebsd.org/releases/13.0R/relnotes/#boot

Title: FreeBSD 13.0-RELEASE Release Notes | The FreeBSD Project

er well.. bugs.freebsd.org/bugzilla/show_bug.cgi?id=258987

Title: 258987 – 13.0-RELEASE installer broken redundancy with UEFI and ZFS

Hey

How to update ports tree without mucking things up? I usually use pkg to install stuff, but sometimes, i.e. Enlightenment desktop needed a patch, so I installed via ports

Is there a way to ensure that pkg and ports installed things are as close to similar as possible so as not to upset things?

AumShivaya: "git pull", but if you're mixing packages and self-built ports, make sure they're both on the same branch, either the quarterly branch or main/latest

ok, I have come across this term: "quarterly branch" but do not understand what this is

AumShivaya: you might want to look at synth which is designed to support this use-case (mixing ports and packages), although Remilia pointer out earlier that poudriere is also capable of this...

okay

AumShivaya: okay, so there are two versions of ports: what git calls "main", and what pkg calls "latest", is the absolute latest version of the ports. what pkg calls "quarterly", which git calls by a branch name (like 2023Q4) is a quarterly release which only gets important updates.

by default pkg uses quarterly (see /etc/pkg/FreeBSD.conf), so if you're building ports by hand, you should use the quarterly version of ports in git, e.g. git checkout -b 2023Q4 git.freebsd.org/ports.git. otherwise, you might find 'make install' tries to update a bunch of things you already have installed

Title: ports - FreeBSD ports tree

ok, where is it set?

for pkg, /etc/pkg/FreeBSD.conf. for git, specify the current branch when checking out the tree: git checkout -b 2023Q4 ...

I also did a pkg search enlightenment and noticed it is still 0.25.4_2.2

where freshports says the latest is 0.26.0.2

when a new quarterly comes out you have to update your git tree to that one at the same time the pkgs are updated

using portsnap or?

oh, I did pkg update, and it still does not reflect the change in e versions

no

what

is pkg slower to get updates than ports?

portsnap is dead. please don't try to revive it

oh :o

I must have been reading something old!

AumShivaya: if you look at the freshports page for enlightenment, it shows 0.26.0 is latest but the quarterly package version is 0.25.4

I have not really used FreeBSD fully since version 6 or 7

was six the bad one, or was that 5?

by default, pkg uses quarterly, which is why you have 0.25.4 in pkg

5 was the bad one, 6 was good

well, 5.0, 5.1, and 5.2 were bad, 5.3 was supposed to be okay

I had good experience with FreeBSD so I am back :P

welcome back

ty :)

we have ZFS now.

AumShivaya: the tl;dr here is that if you want to build anything from ports, it's really a lot easier to build everything from ports. but not everyone wants to do that, i understand...

unixwitch: why is it easier to build everything from ports?