For the past year we have been documenting the Intel Poulsbo Linux driver and how it is a bloody mess on the basis of it being a binary-only driver (in comparison to their fully open-source stack for their other IGPs) that is not well maintained, is not easy to procure outside of Ubuntu, and is ridden by other problems. Intel though is not solely at fault because the GMA 500 "Poulsbo" chipset is actually a product of PowerVR. A month ago we learned that a new Poulsbo Linux driver is under development and that it would use Gallium3D and be a big improvement over the current Poulsbo driver stack, albeit it would remain closed-source. A month has passed since sharing the news and this morning we finally got our hands on new Intel Poulsbo hardware...
I've used the existing IEGD releases on similar hardware. My main gripes are that Linux support seems to be an afterthought and everything is handjammed in. You are supposed to create a configuration with some windows tool and then export it to a tar file that you copy to your linux target. Then you run it as root and it does a bunch of random makefile and shell script stuff, asking you questions along the way.
Looking at the xorg config format, it seems like they are trying to make it look like a registry.
For myself, I made a clean makefile for the userspace bits that installs them to proper places with install and accepts DESTDIR. The kernel module is a bit more difficult to deal with as they have attempted to support multiple kernel versions in a particularly bizarre way. They copied the headers from a whole bunch of kernels and keep them in various directories. I ripped all those and made it include headers from the actual kernel and added a simple Kbuild and Makefile instead of their Frankenstein setup. Using real kernel headers makes it much easier to build on 2.6.32.
I'm pretty sure the IEGD 10.2 is newer. However, I'm not sure what the differences are between the iged xorg driver for poulsbo and the psb xorg driver for poulsbo. Are they two different drivers, or do the ubuntu packages just rename the xorg driver?
Also, the target system is an embedded system that is from scratch (distroless) and the filesystem image for it is constructed on a development machine. That's why the IEGD scripts that attempt to find out the version of things on the host box and then install to the host box annoy me so.
I do not understand where the author of the article gets the idea that Moblin would be the upstream/origin/place of publishing of such a driver, if it existed.
It just does not make sense to have a Linux distribution being this source....
I think this comes from the driver that came with canonical netbooks and could not be obtained independently, or at least, thats the way I understand it. Just like ATI sometimes releases a driver through Ubuntu before the driver is available on their site.
There is a great deal of confusion surrounding this driver, who makes what version, who distributes what version, etc. Including the strangeness of Intel requiring an account to download a driver (free account).