Posts
3711
Following
217
Followers
381
Kernel plumber at kernel.org. Uncloud plumber at parity.io. Entrepreneur at Siltakatu Solutions Oy.

OpenPGP: 3AB05486C7752FE1
@oleksandr It's a cool driver to work with. Has been in mind over the years but companies do not care about it ;-) Now is my chance given that I'm not employed to do kernel... Last three weeks have been gone to figuring how to test media from kernel perspective (both in general and specific) but have got side now together, and learned a lot in process.
1
0
0
@oleksandr I don't know I did not do research, only heard the claim. I guess you could do application level solution when e.g. browser would use some Pipewire interface to query cameras provided by it but I personally think that we need a more low-level and stack independent solution.

It's also bad for open source. I mean we should do open source stack overall in a way that someone can e.g. provide alternative to the whole Pipewire even tho at this point we love it.
1
0
0
@Foxboron ... which can be disabled from the settings. I don't want to use it either but it is not like the first feature ever that I dislike :-) So yeah I do agree that AI crap sucks but I don't see this VLC feature as such a big deal.

I still use Audacity too...
0
0
2

Jarkko Sakkinen

Some have argued that "we could do the same with Pipewire".

1. I love Pipewire. It's ****ing great. For the first time Linux has multimedia stack that surpasses Apple's.
2. With video we still might want to have option to cut down the stack depending on device. I.e. have a choice in product design whether to use or not use gstreamer and pipewire.

So from this premise IMHO a loopback video device does make sense.

#v4l2 #pipewire #gstreamer
1
1
3

Jarkko Sakkinen

This is quite welcome actually: https://fframes.studio/ (also rust-av)

Unless you enjoy libav interfaces or always layer it with gstreamer or libvlc...

A Rust rewrite that actually makes sense!

#rust #rustlang #ffmpeg
1
1
2
@kernellogger i need that kernel just keep my machine alive long enough to write a test program with a driver running using a new API :-) so as long as it boots and logs into GNOME, I'm good... So not looking for perfect result.
0
0
1

We're looking for the next Fedora Project Leader. Is it you, or someone you know?

https://redhat.wd5.myworkdayjobs.com/en-US/jobs/job/Boston/Fedora-Project-Leader_R-044932

3
12
1

Jarkko Sakkinen

Doing a distribution kernel for testing feature patch in Fedora is actually quite trivial now that I see the forest, and not only trees.

Up to the SRPM's it is exactly these steps:

1. git clone git@gitlab.com:cki-project/kernel-ark.git && cd kernel-ark
2. git -C ~/work/kernel.org/jarkko/linux-tpmdd diff master..v4l2-loopback > redhat/linux-kernel-test.patch
3. make dist-test-patch
4. make dist-srpm

Cool: three (!) weeks of improving my QA flows but now I can move on to the API breaking changes in the V4L2 loopback device :-) It is easier to develop the test code for anonymous inode version when you have the driver running in your desktop. Otherwise, it would be super painful.

@kernellogger thanks for getting me up on speed with compiling Fedora kernel, could not have mastered it w/o you :-)

I'd like to remark steps two and three because those I added by trial and failure and peeking at redhat/Makefile. It probably would make sense to put a remark here: https://cki-project.gitlab.io/kernel-ark/

#fedora #linux
2
4
7
@fedora I sincerely hope you get a good one as it is probably the best workstation operating system in the world, good luck!
0
0
1

The opening for the Fedora Project Leader position is live. If you know someone who would be a good fit for this role, please share this link with them!

Learn more: https://redhat.wd5.myworkdayjobs.com/en-US/jobs/job/Boston/Fedora-Project-Leader_R-044932

1
11
0

not a big fan of the new logo tbh

2
10
1

With Facebook currently in the news again, it's perhaps worth pointing out that I left the platform 11 years ago because they were already generating fake replies to fake posts, presumably in an attempt to increase engagement with ads. I got rid of my account and blocked facebook URLs and IP addresses on my router so that I cannot accidentally engage with any of their stuff at all.

That is of course not the only thing which was bad about Facebook. By 2016 they were working to undermine democracy.

So there's nothing actually knew about the harm they're causing now. It's been obvious for more than a decade which direction they were heading in. But I'm happy that there's more awareness now.

So, have you all deleted your facebook / instagram and whatsapp accounts already ? If not, why not ? Meta, just like Twitter, has been a nazi bar for some time now.

https://davidhembrow.blogspot.com/search/label/facebook/

0
2
1

Jarkko Sakkinen

Edited 14 days ago
The tech thing outside of crypto bro bullshit that got me interested in Polkadot is some "math inventions" in the distributed computing.

If you distribute block generation let's say by N in a blockchain you easily end up to an architecture where also network security also divides by N.

What Polkadot and its spec generalization JAM (https://graypaper.com/) do differently is that with the help of game theory parallel chains balance back in collaboration.

The basis of all this is described in this math paper: https://eprint.iacr.org/2024/961

With these shenanigans, the network sort of creates a scheme where you have a distributed computer where the relay chain is actually just part of the overall machinery tracking the history.

What I personally work on is a RISC-V relative bytecode VM called PVM, which has eight registers instead of 16 and immediate ECALL. This speeds up JIT over let's say WASM or eBPF given that register allocation to ARM and x86 becomes much more straight-forward process and generally it is more close proximity of those CPU architectures. I.e. what I hack is essentially a software-defined compute core and overall this network forms a sort of like "GPGPU" infrastructure.

The value part is more to optimized to book keeping than increase investor value so that you can make shared services where everyone gets their part when being a sub-provider of some service. E.g. there's no fixed number of tokens but there's instead rate-limited inflation.

So yeah, I'm not interested in blockchains or cryptocurrency per se but I'm very interested on distributed systems and providing alternatives to a data center. I think this make this work valuable.
0
1
2
@kernellogger Related to testing V4L2 loopback driver. It kind of grew out of BuildRoot (testing side, hard to get some packages etc.) and Yocto feels way too big hammer. Fedora Server in libvirt feels like the right scale for this work...
0
0
0
@kernellogger thanks, that page looks useful!
1
0
1

Jarkko Sakkinen

In kernel-ark:

1. "make dist-fedora-configs": Can I make only process x86 stuff and not all the archiectures?
2. Let's say I've applied patches containing a new kconfig flag. How to fix that up?

#fedora #linux #kernel
1
1
0

Jarkko Sakkinen

I feel that Rust has grown up. People use it like a pro for their business 😎 No extra fuss...

#rustlang
0
0
4
@kornel Yeah so I ASSUMED that it must be along the lines what was described here but I don't want to live in a belief system so had to ask (and I'm glad that I got so many great answers here).
0
0
0
Show older