| |||||||||||||
LCA: Graphics driver poniesLCA: Graphics driver poniesPosted Jan 28, 2010 13:57 UTC (Thu) by Tet (subscriber, #5433)Parent article: LCA: Graphics driver ponies The driver has just lost its user-mode support; it will be KMS only I do wish they wouldn't do this. It essentially prevents me from using a recent kernel. With my card, KMS causes a segfault in the server, but it works just fine without KMS. I understand why they want to do it. The non-KMS code is messy, and they don't want to maintain it. But it works, in places where KMS currently doesn't.
(Log in to post comments)
LCA: Graphics driver ponies Posted Jan 28, 2010 15:43 UTC (Thu) by rahulsundaram (subscriber, #21946) [Link]
The answer there is not to maintain two different systems with different bugs in each of them but fix the one that is there for the longer term
LCA: Graphics driver ponies Posted Jan 28, 2010 16:49 UTC (Thu) by nix (subscriber, #2304) [Link]
Especially given that the non-KMS code is not just 'messy' but
misdesigned. Multiple drivers in kernel and userspace arguing over a single video card is pretty much obviously a bad idea. If video cards on PCs had never been capable of displaying text, I suspect this design would never have lasted as long as it has, at least on Linux: this meant that most people could just use a text-mode console and pretend that there was only one graphics driver, the 2D X one, when in fact the other driver was in the video card...
LCA: Graphics driver ponies Posted Jan 29, 2010 21:08 UTC (Fri) by Tet (subscriber, #5433) [Link]
Agreed. Clearly maintaining both is not viable. All I'm proposing is not removing the non-KMS code until the KMS code is stable and usable. Importantly, that has to include older hardware (not obsolete, just not cutting edge). It seems there's been a reasonable amount of work on getting KMS working with newer video cards. But it needs to work on cards that are 3 or 4 years old, too, and based on my experience, it currently doesn't.
LCA: Graphics driver ponies Posted Jan 30, 2010 6:03 UTC (Sat) by rahulsundaram (subscriber, #21946) [Link]
Have you reported your bug? We can debate on what is the right threshold but it is also the case as long as there are two different systems end users tend to switch between them and never report bugs
LCA: Graphics driver ponies Posted Jan 31, 2010 16:21 UTC (Sun) by Tet (subscriber, #5433) [Link]
Yes. For better or for worse, I report every problem I find as a bug. The usual response seems to be to ignore it, wait for a year and then close the bug because it's for an unsupported release. But I have to keep trying...
LCA: Graphics driver ponies Posted Jan 30, 2010 18:48 UTC (Sat) by nix (subscriber, #2304) [Link]
It certainly should work with older hardware: my 2000-vintage system (now
reserved for compatibility checking) with its Radeon 9250 is quite happy with KMS.
However ATI have a crazy variety of cards, all wired up slightly
LCA: Graphics driver ponies Posted Jan 28, 2010 19:34 UTC (Thu) by jsbarnes (guest, #4096) [Link]
Please file a bug. A server segfault obviously shouldn't happen, but should also be pretty easy to fix.
LCA: Graphics driver ponies Posted Jan 28, 2010 22:02 UTC (Thu) by Thalience (subscriber, #4217) [Link]
As I understood it, the component that would lose UMS support is the X Driver (DDX). So you wouldn't be able to run a newer X on an old kernel.
LCA: Graphics driver ponies Posted Jan 28, 2010 22:59 UTC (Thu) by nix (subscriber, #2304) [Link]
... or on a non-Linux OS. I suppose OSes that can't be bothered to
implement KMS over the course of many years are probably moribund anyway. (The cynic in me would say that everything but Linux is pretty moribund as an X server platform these days...)
LCA: Graphics driver ponies Posted Feb 2, 2010 1:43 UTC (Tue) by Spudd86 (guest, #51683) [Link]
Apparently people ARE working on it for FreeBSD and Solaris, but FreeBSD doesn't have as many people as Linux so it might be a while
|
|||||||||||||