pkubaj: bsd.network/@lattera/111316217550807247 <- this might be of interest to you

Title: Shawn Webb: "Made a lot of progress on supporting Cross-DSO CF…" - BSD Network

I'm hoping to enable Cross-DSO CFI in HardenedBSD 15-CURRENT in January 2024. this will cause us to also enable LTO in ports. ample opportunity for you to punish us guinea pigs with experimental patches ;-)

(I mean "punish" in a good way here)

pkubaj: I'll also report issues we come across, and potential fixes/workarounds, if you'd like

Is anyone running 14.0-RC3?

this is not going to max depth 5. Can mistake I am making? du -h -depth=5 /path/home-backups/ | sort -rh

Not I

Not I?

Beladona, That was to BraveheartBSD query of "Is anyone running 14.0-RC3?"

ok

Beladona: did you try simply '-d 5' ?

Beladona, Option name spelling is "-d" not "-depth"; from "du(1)" manual page: "-d _depth_ Display an entry for all files and directories _depth_ directories deep." As in: du -h -d 5

there any cli spellchecker? so i can like spchk myfile.txt and have it spit out possible misspellings

"aspell" may do that (use it only via "vim")

oh aspell might work. know how i can make it output misspelled words? -c puts me into interactive checker mode

ah -a

ty!!!

If I get no return from dmesg | grep pcm after loading snd_driver am I screwed?

bradd: ok trying -d

bradd: ok trying -d=5

no equal sign

ok

hi, i just setup a host-to-host vpn with strongswan, it said IKE_SA conn is established, but now how do i reach services internal to the remote host?

there doesn't seem to be any tunnel interfaces on either site

oh, i do see failed to establish CHILD_SA, keeping IKE_SA

traffic selectors unacceptable

dang I got disconnected, did anyone answer my question

no

do i need to enable gateway if i just want to do a port redirect?

i'm trying to port mirror a remote host's port locally

rdr pass on ena0 proto tcp from any to any port 2222 -> 192.168.1.23 port 22

what's the best tcp proxy?

i put it in /etc/pf.conf, and did pfctl -f /etc/pf.conf, is that all i need?

pf, i would think..

polyex: haproxy seems to be the standard bearer in enterprise software

ty!

:)

are you echelon on hn?

no

remember when echelon was a conspiracy theory? now it's just like, accepted by everyone that the US federal gov illegally spies

yeah, like when it was revealed that at&t had pipes that directly funneled traffic to the nsa

ya, room what again? don't remember the number

was a big scandal during bush admin

641A

they probably do far worse things now

now they arrest you for whistleblowing

ya

lattera: interesting, thanks for letting me know

i would be happy to keep updated if possible

do you use LLVM_BINUTILS?

pkubaj: for base? yeah, we use a full(er?) llvm toolchain, including using WITH_LLVM_BINUTILS

Running a R59 Radeon (and a integrated intel gpu), amdgpu from drm-kmod gets loaded but startx returns no screens found

neofetch only reports the integrated intel gpu

polaris12_sdma1.bin gets loaded

loading i915kms with amdgpu crashes the system btw

i have a question on fifos... when i open a fifo with O_NONBLOCK, i can remove the O_NONBLOCK with fcntl, and then a plain read will block until something opens the pipe for writing and writes something

but this seems to violate pubs.opengroup.org/onlinepubs/9699919799/functions/read.html

Title: read

ascreen, What about it seems to be the violation? Do you have a small test program you could post that shows this?

"If some process has the pipe open for writing and O_NONBLOCK is clear, read() shall block the calling thread until some data is written or the pipe is closed by all processes that had the pipe open for writing."

pkubaj: sourceforge.net/p/elftoolchain is also used

Title: ELF Tool Chain download | SourceForge.net

it's a collection of software that started in the base system, then got spun off into its own project; not unlike libarchive

rwp: isnt a fifo with no writer and one reader an empty fifo?

rwp: mkfifo myfifo, then run ix.io/4KgF

curiously on darwin it gets eof on both reads

ascreen, Thank you for the small test program. You have convinced me. I agree. It's not operating as per the standard.

The FreeBSD and Linux kernels are disagreeing on this behavior too.

i think it's due to github.com/freebsd/freebsd-src/blob…ain/sys/fs/fifofs/fifo_vnops.c#L167 but I don't know the code well

Title: freebsd-src/sys/fs/fifofs/fifo_vnops.c at main · freebsd/freebsd-src · GitHub

but arguably the freebsd behavior is nicer in practice...

also on Linux/MacOS there's an inconsistency with select: A descriptor shall be considered ready for reading when a call to an input function with O_NONBLOCK clear would not block, whether or not the function would transfer data successfully.

but select waits for actual data (or a reader opening), while read immediately returns 0

ascreen, Hmm... The standard says "If some process has the pipe open for writing and..." That's the critical part.

no, i mean the first clause

Running that very nice small test program just by itself does not have any process with the pipe open for writing.

The first clause... Looking... "If no process has the pipe open for writing, read() shall return 0 to indicate end-of-file."

that is the case here imo

the fifo is empty and has no writer

Right. Sorry. I was focusing on the wrong part. Okay. I am in sync now. If no process has the fifo open for writing.

Then the read non-blocking should return 0 (as it does on Linux) but on FreeBSD it blocks. Gotcha.

same on openbsd fwiw, dont have anything else around to check

but tbh it requires an unusual dance to trigger, if you open blocking it wont happen of course

Subtle differences between systems like this cause problems while porting software so it is better if they don't exist.

just another autoconf check /s

IMNHO if code tickles these types of issues then probably it indicates that the code is doing something off the normal path and would benefit from getting back on the path.

Some years ago I was porting expect to HP-UX which is based upon the BSD4.2 kernel rather than the more modern BSD4.3 kernel. There is a similar subtle difference in select() handling of non-blocking I/O there too.

On modern kernels select() returns an indication if data is present to be read. On HP-UX based upon 4.2 it returns an indication of if it would block. Which are different things in the case of non-blocking I/O.

On HP-UX select() with non-blocking I/O it always indicated that it would not block. But everywhere else it would indicate if there was data to be read.

I think this is a FreeBSD non-conformance issue and at the least deserves a PR for it. And it is subtle. Not very often to be hit. Which is good. But when it is hit that's not good.

i think select should not depend on state of O_NONBLOCK

posix says "ready for reading when a call to an input function with O_NONBLOCK clear would not block"

it leaves the O_NONBLOCK-set case undefined

yes, but you can toggle that after

no, read again :)

me? I did.

Title: pselect

3 mentions of O_NONBLOCK

but if a blocking read doesnt block, why would a O_NONBLOCK read not read?

I'm not suggesting it would not

so it does the same for nonblock and block, which is what i argued

you're speaking to the read behaviour; I'm talking about select itself

I fine point, but since we're language-lawyering

let's move on and use poll instead of select, where it clearly says: The poll() function shall not be affected by the O_NONBLOCK flag.

finding systems where select and poll does different stuff is also fun...

select's worst issue is the FD_SET size limit. Ran into that, on FreeBSD as it happens. Converted to poll purely because of it

yep

poll api is nicer in general

(I had an fd leak, oops)

hm, is ppoll in posix-next?

3.35x compression ratio on a NetBSD VM's zvol with zstd-1. Nice.

neat

Hi freebsd community, does anyone here work with freebsd on a raspberry pi? I am having performance issues, as in dev.cpu.0.freq is only registering 600 out of a possible 1500 and looking for any tips/tricks someone may have. Thank you in advance.

ugh i need znc

xscreensaver keeps crashing my computer x___x

does it trigger a suspend power state

yeah i had power settings on, i turned them off after the last time, i’ll see if that fixes it or not. might also be the video card driver. i need to learn how to debug crashes

hey

is there some other way to set ports flavours, or is there only dialog4ports?

but normally the xscreensaver power settings are only for the display, not the system itself, so, meh

because somehow my dialog4ports does not show properly, has to do with own termcap whatever

...haha, oh no, it also crashes everything without power savings enabled.

Question: When I open neovim, try to copy something with shift+ ctrl+ c or Shift+"+ `+`, it does not gets copy. However when I cat something on terminal. I can copy with Shift+ctrl+c. How can I fix this?