EuroBSDCon 2022

No videos are available yet to provide much-needed context to presentations, but we'll keep you posted.

Day -2 - Arrival in Vienna

After being thoroughly delayed by Deutsche Bahn, I hopped off an InterCity Express train to check out the hotel room for people speaking at EuroBSDCon, which was An Experience in itself. There was a mural of a shirtless man with a sword covered in snakes next to my bed, what else do you need in life? Lots of coffee, obviously.

Begin the march to the conference to listen to Marshall Kirk McKusick lecture on schedulers.

Day -1 - NetBSD Developer Summit

Around 16 NetBSD developers gathered in a room for the first time in two years. I was a little bit distracted and late due to Marshall Kirk McKusick's very detailed lecture on filesystems melting my brain somewhat, but we had the opportunity to present various informal presentations, after we'd finished showing off suspend/resume support on our ThinkPad laptops.

Benny Siegert opened with a presentation on the state of the Go programming language on NetBSD (and whether it is "in trouble"), covering various problems with instability being detected inside the Go test suite. Go is particularly interesting (and maybe error-prone) because it mostly bypasses NetBSD libc, which is unusual for software running on NetBSD, instead preferring to implement its own wrappers around the kernel's system calls.

A few problems had been narrowed down to being (likely) AMD CPU bugs, others weren't reproducible in production (outside of the test suite) at all, and others may have been fixed in NetBSD 9.1 - the NetBSD machines running tests for Go do need to be updated. If you're from AMD, please get in touch.

We've got a very impressive test suite for NetBSD itself, but outside tests are always useful for identifying problems that we can't catch... that said, they do require a lot of work to maintain, and a lack of patience is understandable. We'd love any help we can get with this.

I pointed out that we get occasional failures bootstrapping Go in pkgsrc, and better debug output would be nice -- Benny was able to arrange this within the day, and we should get nice detailed bootstrapping logs for Go now.

Pierre Pronchery (khorben@) discussed cross-BSD collaboration on synchronizing our device driver code bases, including his recent NetBSD Foundation-supported work on the emuxki(4) sound card driver, where other BSDs have taken the same code base but improvements had not yet been universal. We all agreed that collaboration and keeping drivers in sync is important. We talked about the on-going project to synchronize NetBSD Wi-Fi drivers with FreeBSD.

Martin Kjellstrand then gave us a very nice demonstration of his NetBSD docker images, and how easy it is to spin up NetBSD on-demand to run a command (this also has wide potential for being useful for testing). In turn, I rambled a bit about my own experiments of dynamically creating NetBSD images. This would lead to a later discussion about whether we need to prioritize improving the resize_ffs(8) command's support for new filesystems.

The theme of creating NetBSD images "for the cloud" continued, with Benny Siegert presenting again about NetBSD on Google Compute Engine.

Stephen Borrill then stepped up to give us an incredibly detailed history of the British computer company Acorn Computers, complete with his personal experiences servicing Acorn machines in the early 90s. We discussed the history of the ARM CPU, and NetBSD/acorn32.

Nia Alarie (surprise) finished up with a very short unplanned demonstration of some of the projects she's been working on lately - using NetBSD as a professional digital audio workstation, improving the default graphical experience of NetBSD with dynamically generated menus, and (again) creating customized micro-images of NetBSD. We discussed support for MIDI devices (I'd later chat with some of the FreeBSD people about collaborating on JACK MIDI).

We then retired to Thomas Klausner (wiz@)'s favorite ramen restaurant and discussed, among other things, Studio Ghibli films, and trains. Trains would be a recurring theme.

Day 0 - start of talks

We began the day with two NetBSD presentations scheduled back-to-back. This mostly meant that I got to talk about some of NetBSD 10's upcoming features, and why it's taking so long to a small crowd of interested people who didn't have much prior experience with NetBSD, while in another room Taylor R. Campbell (riastradh@) discussed his very dedicated efforts to make suddenly disappearing devices more reliable and not crash the kernel (we're still waiting for a live demonstration).

Next, Pierre Pronchery (khorben@) discussed the power of pkgsrc for creating consistent environments across platforms for software developers, serving as a nice portable, classic Unix alternative to technologies like Docker and Nix.

The final presentation of the day was riastradh@ again, this time providing a live lecture (from Emacs!) about memory barriers in the kernel. We all learned to appreciate the nice abstractions technologies like mutexes provide to stop CPUs from re-ordering code on multi-processor machines in inexplicable ways.

Day 1 - final talks

The second day of EuroBSDCon presentations was mostly devoid of anything NetBSD-focused, so we had a nice opportunity for cross-pollination and to learn and collaborate with other BSD projects. I chatted a bit with an OpenBSD Ports developer about the challenge technologies like Rust pose to developing a cross-architecture packaging system, and with a FreeBSD person about the state of professional audio on our respective platforms. Michael Dexter finished the day of presentations with a very passionate speech about why we all need BSD in our lives, regardless of our preferred flavour.

More topics were discussed in the various break periods, including whether our newest update to the GPU drivers is stable enough to include in a release (verdict: works for me).

We then watched as various BSD t-shirts and boxes of chocolates were auctioned away to support a local refugee center. The organizing committee forgot to include the NetBSD Foundation on the list of sponsors, but we forgive them.

Other news from the Project

I've recently made sure the NetBSD 10 changelog is up to date with all the new goodness, so you should check that out.