Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

[Posted October 27, 2010 by corbet]

LinuxDevices.com has an overview of the Yocto Project just announced by the Linux Foundation. "Unlike build systems based on shell scripts or makefiles, the Yocto Project automates the fetching of sources from upstream sources or local project repositories, says the project. Its customization architecture is said to allow the choice of a wide variety of footprint sizes as well as control over the choice or absence of components such as graphics subsystems, visualization middleware, and services. Yocto is based on the GNOME-derived Poky Linux, a well established platform-independent, cross-compiling build system that uses the same architecture as the OpenEmbedded build system. "

to post comments

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 28, 2010 2:29 UTC (Thu) by pabs (subscriber, #43278) [Link] (5 responses)

Hmm, makes you wonder why they didn't just join OpenEmbedded.

What Do We Want? LWN Supporter Subscription!

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 28, 2010 3:12 UTC (Thu) by xxiao (guest, #9631) [Link]

Poky is at an awkward position since acquired by intel who then pushes meego.

Meego uses opensuse's build framework.
Poky uses openembedded's build framework instead.
Intel drops ubuntu/debian(from moblin days) and switches to rpm/opensuse.
Canonical is behind Linaro which both are ARM friendly.
...

I agree why not just help Openembedded, but then Montavista and MentorGraphics are both using Openembedded which are Intel/Windriver's competitors.And then there is Android which is mostly about ARM again, Intel got to do something...

confused?

I will stick to openembedded/debian/ubuntu/openwrt, which each seems have a real community, and is dominant in embedded space.

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 28, 2010 7:26 UTC (Thu) by rossburton (subscriber, #7254) [Link]

Using "raw" OpenEmbedded in a real product is near impossible because it changes too much, you'll need to fork and stabilise it to match your own schedule. (similar to how Ubuntu fork and stabilise Debian Unstable)

Of course this is what Poky does. There is huge amounts of co-operation between OE and Poky with new packages in OE migrating to Poky, and new features being worked on in Poky and then merging back to OE when they are stable.

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 28, 2010 20:35 UTC (Thu) by zeddii (subscriber, #70881) [Link] (2 responses)

I can comment as someone who's been working with OE/poky/other Linux build
systems for quite a few years, and from someone who's been involved with
multi arch embedded kernel development for just as long.

For me, Yocto compared to OE is all about (the attempt) to pick the technologies and parts that are the best of the alternatives. Standing behind those choices and doing the work to make sure that they all play nicely together.

Using the kernel as an example, it has consistently proven itself to be
difficult to maintain a consistent feature set across boards and architectures, that are of interest to embedded devices. Getting to a prompt is nice, but being able to take advantage of tracing, debug and leveraging bug fixes that another board, on another architecture ran into is a massive time saver.

Having a place to collect, maintain and develop the kernel features of interest to embedded development is key, with the hope that these features can eventually merge into the mainline kernel.

The alternative is grabbing a quasi random kernel, seeing what sort of support for an esoteric board, suffering through silent boot death, getting your initrd to load or your ethernet device to NFS boot and then finally making it to the prompt .. to find out that you are getting random segfaults due to some hard to find cache management issue.

Having a place to go and get a good known base, is a great place to start. And having that known good place without some level of an umbrella project is hard to achieve.

My Canadian 2cents worth.

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 28, 2010 20:52 UTC (Thu) by jacmet (subscriber, #19734) [Link] (1 responses)

We already have such a place, it's called kernel.org.

Keeping a bunch of patches in yet another build system instead of pushing these things upstream doesn't sound like progress to me.

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 29, 2010 16:51 UTC (Fri) by zeddii (subscriber, #70881) [Link]

kernel.org is absolutely the place to do kernel development, embedded or not.
No disagreements there. But having a place to bounce ideas around, or get multi architecture soak time in a off-cadence or ongoing basis is more what I was referring to. The goal is to not carry any patches, and improve the same kernel, which is the kernel.org master.

Thanks for triggering the need for clarification!

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 28, 2010 18:07 UTC (Thu) by dvhart (guest, #19636) [Link]

I've been working on the Yocto Project for the last six weeks, helping the team get ready for the recent launch. So yes, the following comes from an Intel employee, but I hope you'll find it to be without bias.

I think rossburton summed up the "why Poky and not just OpenEmbedded" well. There are several reasons, but the focus on a known set of working packages is a big one in my opinion.

As to the perceived conflict between MeeGo and the Yocto Project, there is obviously some overlap in that they are both embedded Linux projects. However, they have very different goals. MeeGo is a platform, it defines various APIs, the UI experience, and even things like an application store. As such, MeeGo is targeted at personal embedded devices (phones, tablets, mids, and netbooks - to a lesser degree).

The Yocto Project focuses on what could be dubbed "deeply embedded" devices. Those devices where the OS is relatively invisible to the user. These sorts of devices don't require a the same common platform APIs or user experience - as they don't expect 3rd party vendors to be writing applications on them (and hundreds of other similar devices without modification). They vary greatly in resources, and demand an extreme level of customization. This where the Yocto Project comes in, helping developers pick and choose which components they need in their Linux deployment, and removing much of the pain associated with rolling their own distribution.

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 30, 2010 0:19 UTC (Sat) by stevem (subscriber, #1512) [Link] (1 responses)

Yawn, yet another build system?

Yocto Project aims to standardize embedded Linux builds (LinuxDevices.com)

Posted Oct 31, 2010 3:38 UTC (Sun) by xxiao (guest, #9631) [Link]

not really, as Poky is basically just openembedded-polished.
I understand that intel is eager to influence the embedded open source space, but its meego, poky, windriver(which has its own build system), all together does not seem very streamlined.
a donation to openembedded or openwrt probably will be much better, in my opinion. those're already very successful community projects. and, embedded is not PC, you can't really standardize, people are going to choose whatever fits the best, no matter how beautiful the Yocto-umbrella is.