01:24 <noocsharp> didn't they only publish the kernel modules?

01:25 <noocsharp> i think userspace is still proprietary

01:30 <illiliti> seems so

01:37 <illiliti> well, it looks pretty useless then

01:37 <illiliti> why do i need bike without wheels lol

01:40 <noocsharp> a step forward is a step forward

01:53 <noocsharp> in theory mesa can target it once the interface is stabilized

02:39 <acheam> testuser[m]: kavin.rocks is crimeflared btw

02:53 <acheam> phoebos: i'm honored to be the first bookmark on your website

02:53 <acheam> "unordered" be dammed

03:19 <noocsharp> sotd: https://www.youtube.com/watch?v=F8dKVbP1Nzo

06:33 <testuser[m]> illiliti: noocsharp it'll be forever useless cuz it uses hardware stuff introduced in the 30xx series

06:34 <testuser[m]> acheam: ik but it really doesn't matter tbh

06:34 <testuser[m]> Hi

06:39 <testuser[m]> https://github.com/NVIDIA/open-gpu-kernel-modules/issues/19#issuecomment-1124360887

06:45 <testuser[m]> And, to be clear, while we do share a lot of the code here with the proprietary driver, this is a significantly different driver architecture. There are currently no plans to either drop support, or to open source the proprietary driver which supports Maxwell and later

06:45 <testuser[m]> So bad

06:46 <testuser[m]> Anyway the userspace being opem source is more important cuz the proprietary kernel driver works on musl just fine

06:46 <illiliti> i wonder what nouveau devs think about this driver

06:47 <testuser[m]> U can use nvidia-smi with gcompat

06:47 <testuser[m]> I was able to run sway too with gcompat hack but it stopped working a few driver versions later

06:51 <illiliti> > Will the source for user-mode drivers such as CUDA be published?

06:51 <illiliti> > These changes are for the kernel modules, while the user-mode components are untouched. The user-mode remains closed source and is published with prebuilt binaries in the driver and the CUDA toolkit.

06:51 <illiliti> garbage

06:53 <illiliti> it's yet another gpl condom guys

07:12 <testuser[m]> https://nitter.kavin.rocks/marcan42/status/1524615058688724992

07:44 <testuser[m]> https://github.com/advisories/GHSA-h8c3-8522-vxc6

07:45 <testuser[m]> How does this work

07:45 <testuser[m]> U can reset terminal and do whatever u can with escape sequences but how arbitrary code execution

08:42 <illiliti> very old and buggy terminals allow injection commands into the terminal input buffer

08:43 <testuser[m]> lol

08:45 <illiliti> today terminals sanitize input, so in order to achieve rce user interaction is needed

08:46 <illiliti> unless attacker found a way to bypass sanitizer

08:48 <illiliti> you can do many things with escape sequences without needing to execute commands

08:48 <illiliti> for example, you can stole clipboard via OSC 52

08:52 <illiliti> not directly related to this but don't copy and paste from internet to terminal: https://vid.puffyan.us/watch?v=LFXZqQL4vTY

09:58 <phoebos> acheam: hehe

09:58 <phoebos> it's unordered, it's a <ul>!

13:01 <testuser[m]> illiliti: is there an alternative to pipefail for posix sh

13:09 <illiliti> yes, wrap commands in {} and check their status

13:10 <illiliti> also pipefail will be available in the next posix issue

13:11 <testuser[m]> illiliti: can u send example

13:11 <testuser[m]> wouldnt checking status discard output

13:25 <illiliti> what are you trying to achieve?

13:25 <illiliti> it depends

13:27 <testuser[m]> jq -r .[].file compile_commands.json | while read -r file; do echo $file; done

13:39 <illiliti> i don't see the need for pipefail here

13:42 <testuser[m]> illiliti: https://termbin.com/wvbu

13:42 <testuser[m]> i want the shell to exit

13:51 <illiliti> var=$(jq ..) || exit; for file ....

13:51 <illiliti> or heredoc

13:56 <illiliti> ^ this is much better than trying to emulate pipefail

13:57 <testuser[m]> how can u use a heredoc here

14:05 <illiliti> set -e; while ... done <<EOF\n$(...)\nEOF\n ....

14:51 <phoebos> testuser[m]: { jq -r .[].file "$1" || kill 0; } | while ....

14:53 <testuser[m]> o i tried that but was not sure how to kill the shell that the sub command ran in

14:53 <testuser[m]> thanks

14:54 <phoebos> yeah kill 0 is like kill $$ i think

14:54 <illiliti> hacky hack, not needed here

14:55 <phoebos> illiliti: how can you do it without a subshell

14:55 <testuser[m]> illiliti: btw var=$(...) makes an extra copy right

14:56 <testuser[m]> but piping directly feeds stdout into program

14:56 <phoebos> ^

14:58 <illiliti> | makes extra syscalls

14:59 <phoebos> pipe? rather than a new shell?

14:59 <phoebos> *sub

15:03 <phoebos> actually pipes create subshells too

15:03 <testuser[m]> ye

15:09 <phoebos> ig var=$(jq...) || exit 1; while ... done <<EOF\n$var\nEOF\n is best

15:09 <phoebos> or while ... done << EOF\n$(jq ... || kill 0)\nEOF\n