Abstract: There is a rising demand for embedded support of Blockchain and Distributed Ledger applications. In this talk I will present my recent work of integrating IOTA DLT into the OpenEmbedded space. I will present what has already been integrated into meta-iota, as well as the next steps. The meta-iota layer will soon be supported by IOTA Foundation’s Ecosystem Development Fund.

Speakers: Bernardo A. Rodrigues, Philipp Blum, IOTA Foundation

Bernardo Rodrigues works with Yocto/OpenEmbedded since 2017. He has BitBaked images for Automotive Road Traffic Control in Brazil and recently he used Yocto on an RCar H3 based platform for Advanced Driver-Assistance Systems (ADAS).

His work on the meta-iota OpenEmbedded layer received a warm reception from the IOTA Community, and he will receive incentive from the IOTA Foundation’s Ecosystem Development Fund to keep developing meta-iota.

Philipp Blum represents the IOTA Foundation as Developer Advocate of the Ecosystem team. He focuses on growing and developing an IoT developer community. Philipp lives in Berlin, Germany and has been in the technology sector for over 6 years. He has experience as a marketing developer for startups and wrote software to optimize Web- and TV-Advertising.

Abstract:

From-scratch builds, even using server grade machines (with 40+ cores) will take just under an hour to complete. Additionally this estimate is just for minimal, stripped down images;  Bigger images that bring up more than just core functionality and support things like web browsers/multimedia would take much longer (on the order of several hours).

Use of the sstate cache drastically cuts down on build times, especially for fresh projects. Xilinx makes full use of the sstate cache to speed up builds for its customers by hosting a comprehensive sstate cache (for all packages for different types of architectures) and allowing users to point their builds to this prebuilt and maintained sstate cache.

There are different ways of distributing the sstate. When building an esdk (An extensible software development kit), the sstate of all non-native components is packaged so that any build using the esdk will happen in the blink of an eye. However, when building an sdk from within another sdk, the sstate for the native components were missing , hence making the sdk build disproportionately long compared to regular builds. We introduced a patch into core that allows users to toggle the inclusion of nativesdk packages into the esdk by correctly handling sstate cache artifacts themselves as well as the corresponding signatures that are used to reference if anything has changed. With this change, a bigger esdk will be built, when required, that will skip rebuilding native components.

Speaker: : Jaewon Lee, Xilinx

Jaewon is a member of the Yocto Project team at Xilinx. He studied electrical engineering at Georgia Tech and went from focusing on hardware design to an embedded software position at Xilinx! He has been with Xilinx for two years. He is actively maintaining and contributing to Xilinx layers, and is eager to add more functionality to open source Yocto project.

Abstract:

The usage of Tegra, the CUDA-enabled system on a chip (SoC) series developed by NVIDIA, is increasing in embedded Linux devices for various industries requiring deep learning and artificial intelligence solutions. Often the Ubuntu-derived image provided by NVIDIA L4T and JetPack SDK isn’t flexible enough. Reliable high-quality industrial devices require a custom embedded Linux distribution for the exact needs of their complex operations. The Yocto Project, OpenEmbedded and meta-tegra BSP provide an excellent solution.

In this presentation, Leon will share his experience in customizing Poky, the reference distribution of the Yocto Project, for embedded devices with NVIDIA Tegra SoCs using OpenEmbedded build system and the BSP meta layer meta-tegra. Due to the limitations of the supported GCC version by latest CUDA versions, it is a challenge to make it work with the latest release of the Yocto Project. This presentation will reveal 3 ways for solving the problem by providing an appropriate GCC version through an external toolchain, porting GCC recipes to latest Yocto release or moving to an older Yocto release. Practical examples and testing scenarios based on NVIDIA Jetson development boards will be provided.

The presentation is appropriate for anyone with basic knowledge about the Yocto Project and OpenEmbedded. The provided information will help other software developers in the community to overcame faster and easier similar technical difficulties while using CUDA-enabled hardware.

Speaker: Leon Anavi, Konsulko

Leon Anavi is an open source enthusiast and a senior software engineer at Konsulko Group. He is an active contributor to various Yocto/OpenEmbedded meta layers, Automotive Grade Linux (AGL), Tizen any many other open source projects. His professional experience includes web and mobile application development for various platforms as well as porting and maintaining embedded Linux distributions to Raspberry Pi and devices with i.MX6, Rockchip and Allwinner (aka sunxi) SoC. Leon holds a masters in Information Technology from the Technical University Sofia. He is the author of the Tizen Cookbook printed by Packt Publishing. His previous speaking experience includes talks about open source software and hardware during events in San Francisco, Portland (OR), Hong Kong, Shanghai, Shenzhen, Brussels, Berlin, Bratislava, Edinburgh, Prague, Sofia and his hometown Plovdiv.

Abstract:

The Yocto Project is using Buildbot for continuous integration services.

AutoBuilder is a project that automates build tests and Quality Assurance (QA) upon a Buildbot configuration for the Yocto Project through metadata.

Buildbot is a software development continuous integration tool which automates the compile or test cycle required to validate changes to the project code base.

During this talk will be presented some case-study using CI techniques to strenghten the deployment of Linux embedded components using the Yocto Project.

Buildbot supports not just continuous-integration testing, but automation of complex build systems, application deployment, and management of sophisticated software-release processes.

When software development processes are automated, they are repeatable, reliable and can be run as frequently as available computing resources allow.

Automating the build and test process gives developers immediate feedback on their work. Tests can run on multiple platforms, ensuring that code changes made on one platform do not cause failures on other platforms.

Once a project is ready for use by users, it is either deployed (for hosted applications, such as web sites) or released (for packaged software such as desktop applications).

Automating deployment makes the process predictable and lowers the risk involved with each push. Changes can be deployed to a staging environment first, then deployed to production using exactly the same procedure, eliminating failures due to human error. Deployments can occur many times every day, with only small changes between each deployment.

Releasing packaged software, too, benefits from automation. The process can involve compiling and packaging on multiple platforms, signing builds, localizing strings, quality-assurance checks, and so on. When automated with a tool like Buildbot, all of this occurs repeatably and efficiently.

Speaker: Marco Cavallini, Koan Software

Open Source and Linux embedded evangelist since 1999 with the first StrongArm boards. Marco Cavallini is an OpenEmbedded member since 2009 and Yocto Advocate since 2012. He founded KOAN in 1996, an embedded software engineering company based in Italy, specialized in kernel development and training services for Linux embedded systems. He is a C/C++ programmer since the mid-80s. When not using computers, Marco is usually interested in mixing Physics with Philosophy.

Abstract:

In the 3.0 Zeus release, we have added ‘resulttool’ and ‘yocto-testresults’. This new age of enlightenment gives us much more data about how the builds are proceeding and a measure of the health of each of the components. What is resulttool? How do we interpret the results? What can we do with those results?

Embrace the power of the Force and realize that we can now perform Anomaly Detection on the results to catch regressions that a Human needs to pay attention to. Do humans need to look at every TestRun and every TestResult? NO. One thousand times NO. With some automation, we can ingest the testresults and perform Anomaly Detection on the results. We can then flag egregious offenders to the historic results and highlight the differences. This allows us to repurpose high-value QA resources to investigate regressions and unexpected results, rather than manually testing or manually investigating test results. In this presentation, we discuss some pre-release approaches to ingesting the test cases and test results. We also discuss some approaches to algorithmically detect significant changes (regressions) in the results over time.

Speaker: Tim Orling, Sr. Linux Software Engineer, Intel Corporation

Tim Orling is a software engineer at the Intel Open Source Technology Center. Tim joined Intel in early 2016 after many years as a volunteer developer for OpenEmbedded and the Yocto Project. He has been an open source software and embedded hardware enthusiast for many years. He taught in a university setting for more than 5 years and has given many technical talks at conferences.

Abstract:

This paper will present how the Yocto Project can be used to create container images for use with container runtimes such as Docker. A brief introduction to container image formats will lead into more in-depth discussions about building container images. The Yocto Project build approach will be compared to popular tools currently in use to create containers images such as Dockerfiles, Cloud Build, and BuildKit. Topics might include licensing, ease of use, reproducible and build time. Based on these topics we will explore how the Yocto Project might be modified to improve the building of container images.

Speaker: Mark Asselstine, Software Architect, Wind River

Mark is a long-time Yocto Project contributor who has worked with embedded Linux for over 12 years. In his role as a software architect at Wind River, Mark works with customers on real world usecases for virtualization and containerization. This work has lead him to be a frequent contributor to the meta-virtualization and meta-cloud-services layers. He is also a member of the steering committee for the OverC project, a containerized Linux OS framework written as a Yocto Project layer.

Abstract:

This year Google announced the Coral Dev Board which is a low cost development board to quickly prototype on-device ML products. The development board is based on a removable system-on-module (SoM).

The official supported software for the Coral Dev Board is Mendel OS, which is a Debian derivative. This is a fairly common approach, to provide a Debian derivative, making it very easy to get started to prototype on the hardware.

For a production grade setup it is preferable to use Yocto/OE-core and in this presentation Mirza will go through the process of creating meta-coral which is a BSP layer supporting the Google Coral Dev Board.

This layer is heavily based on the available BSP components that are offered in Mendel OS. Mirza will also cover current status of meta-coral as in what works and what does not, and possible future development where he is looking for collaborators.

There will also be a live demonstrating successfully booting an Yocto/OE-core image on the Google Coral Dev Board.

Speaker: Mirza Krak, Northern Tech

Mirza Krak is an Embedded Linux Solutions Architect with seven years of experience in the field and is currently with the Mender.io open source project. Mirza was a Mender community member for a couple of years which led to him joining the Mender project full-time in 2018.

He is involved in various other open source projects and is a Linux kernel contributor.

Mirza’s expertise is within Board Support Package development which ranges from hardware bringup, bootloaders, Linux kernel and build systems (Yocto/OE-core). Mirza has spoken at various conferences including Embedded Linux Conference (US & EU), NDC Techtown, and other technology conferences.

Abstract

As storage size and processing power is increasing on the low end of embedded devices, using full distributions on embedded and IoT devices is becoming more pervasive. The ease of use, larger user communities, and ability to easily add precompiled packages could cause current users to drop Yocto in favor of one of the full size distributions. By providing a binary package feed, Yocto can reduce exposure to losing users to this. This will also have other benefits for Yocto users, such as enabling faster developer use of packages and reduce the amount of rebuilds.

This task is very much a work in progress. Jon Mason will go into the background of the dilemma, the work currently being done to implement this in Yocto, the design decisions and trade-offs, and future work.

Speaker: Jon Mason, Arm

Jon Mason is a Software Engineer with nearly 20 years experience in the industry. Jon joined Arm in October 2018 with his sole purpose being to make all Arm aspects of Yocto/OE as awesome as possible. Most recently, he was employed at Broadcom performing a variety of tasks including Linux kernel bring-up on new Arm SoCs, enabling hardware in u-boot, and porting Zephyr to a new Arm Cortex-A based SoC. Much of this development was done in the Yocto environment, which provided Jon a good background for his current role.

Outside of work, Jon maintains NTB and a few other drivers in Linux.

Abstract:

Using the Yocto Project master branch for a CI/CD (DevOps) approach has it’s challenges.  I will discuss moving from a stable release and uplift strategy to a continuous integration, continuous development and continuous delivery model using Yocto Project, OpenEmbedded and related community layers.  I will include information on my experiences and the challenges as both a contributor and consumer of the components, tracking contributions, integration of new content, testing challenges and release readiness practices.

Speaker: Mark Hatle, Xilinx