On Nov 4, 2008, at 7:53 AM, Matthew Garrett <[EMAIL PROTECTED]> wrote:
> On Tue, Nov 04, 2008 at 07:39:11AM -0600, Victor Lowther wrote: > >> Ripping out quirks is definitly the wrong way to work with kernel >> mode >> setting. It is better to add new --quirk-none quirks once we know and >> have tested kernel and driver combinations on hardware that currently >> requires quirks. I needed something easier to work with than hal >> rules >> to test things with, and this is the yestbed I came up with. > > No. Quirks currently exist to work around missing functionality. Once > that functionality is implemented, continuing to provide the quirks is > working around bugs. It's more important to find and fix those bugs > than > it is to band-aid around them. On KMS systems, they're actively > damaging > - the kernel will program graphics mode, the quirks will program text > mode, X will ask the kernel what mode it's in, get the wrong answer > (because the kernel still thinks it's in graphics mode) and then > attempt > to display on a card that's set up in text mode. Did you miss the bits about taking kernel and video driver versions into account? It is not like old, broken software is going away, you know - not with multi year support contracts and all that. Or the bit about new --quirk-none quirks? Adding new quirks to the database that say "kernel rev x with video hardware y running driver z needs no quirks" is less invasive than ripping out quirks that are still needed on older video driver stacks. > > -- > Matthew Garrett | [EMAIL PROTECTED] _______________________________________________ Pm-utils mailing list [email protected] http://lists.freedesktop.org/mailman/listinfo/pm-utils
