March 8th, 2007
So I’m still trying to figure out exactly _where_ it says ‘hacker’ on my job description, but ah well.
We now have most definitively the best graphics card detection kung fu in the Linux world. Boastful? Well, maybe. =) But the NVIDIA, ATI, Intel and Unichrome card detection in current Cooker is basically up to date with just about everything imaginable, after I’ve spent a fun filled few days trekking through the official pci.ids, ATI and NVIDIA driver documentation, and occasionally through random bits of source code looking for PCI IDs.
I also further rationalized the conventions for detecting and describing cards. Now, in XFdrake, ATI, NVIDIA and Intel all work the same way: by groups of cards / chips that use the same driver settings. So you have ATI pre-Radeon cards, ATI Radeon 9250 and earlier, ATI Radeon 9500 to X850, and ATI Radeon X1300 to X1950; NVIDIA Riva 128, NVIDIA TNT to GeForce 2, NVIDIA GeForce 3 to GeForce 4, and NVIDIA GeForce FX and later; and Intel 810 / 815 and Intel 830 and later. This is consistent and easy to understand. Only drawback really is that the naming scheme doesn’t cover ATI’s and NVIDIA’s professional card lines (FireGL and Quadro), but these should pretty much all be autodetected in the correct category anyway. I will try and think of a way to incorporate them into the names of the categories for 2008.0.
I have also been cheerfully cutting a swathe through the decade old cards that no-one uses any more. In current XFdrake there are individual entries for all sorts of massively obsolete cards which no-one has used for many years, and most of which map back to the same three drivers anyway. So I’ve just implemented generic categories for each main driver and got rid of the specific definitions for massively obsolete cards. So to anyone out there still using an ACX-based ISA graphics card…pick the generic VGA driver.
Or just treat yourself to a nice spanking new Rage64 or something.
During this I submitted a few patches to upstream pci.ids to provide extra information on NVIDIA devices and also a couple of new definitions for NVIDIA onboard sound I came across, so I’ve been doing my bit for the community.
I added support (as in, driver mapping) for several NVIDIA, Realtek and Intel onboard sound devices that were previously ‘unknown’, so in 2007.1, at least the correct driver will get loaded for these. Some will still not work correctly with the stock kernel because it’s only going to have ALSA 1.0.12 at time of release, unfortunately, but it will likely get updated to 1.0.14 in a post-release update, and kernel-tmb and kernel-linus will both have ALSA 1.0.14 at time of release.
Besides that, I updated the experimental packages we have for X server 1.3 and Intel driver 2.0 (with modesetting and randr 1.2 support, as I blogged about before) to today’s git status, which fixed 3D acceleration, updated the avant-window-navigator package to the latest SVN, and filed a few bugs on beryl for not working perfectly with gimmie. And I helped (or rather, commentated noisily) while Thierry attempted to add support for sdhci and tifm-based SD card readers to harddrake, although this doesn’t appear to be quite working yet.
Didn’t write a new newsletter yet because there’s sod all to write about until RC1 comes out.