Re: intel driver and KMS
On 7 April 2010 14:31, Julien Cristau <jcristau@debian.org> wrote:
> On Mon, Apr 5, 2010 at 15:19:03 +0100, Pedro Ribeiro wrote:
>
>> It would be great if the upcoming Squeeze would support this platform
>> properly - all new laptops are using it and this would provide some
>> sort of future proof for the stable version!
>>
>> This is a tricky decision for you, because forcing users to switch to
>> KMS is not very nice... but I believe in the end this is better,
>> because upstream does not support UMS anymore, which would make your
>> life difficult for bug handling. In other words, the number of bugs
>> would rise sharply as you force users to switch to KMS, but hopefully
>> will come down with time and pay off over the support years of stable.
>>
> So basically you're saying 2.10 or 2.11 fixed some bugs. Which is all
> well and good, but then why not go with 2.12 in three months, because it
> will fix a couple of bugs introduced in 2.11. And introduce 10(?) new
> regressions.
Not really. It improved leaps and bounds Intel HD performance. Its not
about bug fixing.
And the intel driver has been pretty stable now since 2.9. I'm not
saying there are new bugs but there hasn't been big changes besides
the removal of UMS code.
> There will always be reasons to go with a newer release. We have the
> choice between a somewhat known quantity (2.9.x), where we can fix
> individual bugs that get reported, or going with the newer 2.10 or 2.11,
> which would bring a few bugfixes, and an unknown number of unknown
> regressions (and have less time to discover and fix those regressions).
> My opinion is it's too late for that. This might be different if we
> knew upstream were very careful to avoid regressions, and/or if there
> was a major feature that we absolutely want in squeeze and that was too
> hard to backport, and/or if we had a lot more resources for QA, testing
> and bug fixing, and/or if we knew 2.9 was majorly screwed, or something
> like that, but as far as I know none of these possible reasons are there
> (feel free to point out where I'm wrong).
The major feature I would like to see in Squeeze is basically good
support for new laptops with Core iX - keep in mind this platform will
probably be the most popular mobile platform in the world, just like
its predecessor (core 2 duo).
The fact that the Debian kernel is already using the 2.6.33 drm tree
makes it so much easier. By updating to 2.10 or 2.11 this would make
Squeeze much more future proof. Of course this would introduce new
bugs, but there is still at least 6 months to go according to the
release team so this is a critical time for this decision.
The big problem of Debian stable is always that it is too old for
older hardware. There is no way around this of course, its one of the
problems which stems from using well tested and stable versions.
I think that supporting Intel HD properly is absolutely critical for
the future success of Squeeze.
> Cheers,
> Julien
>
Regards,
Pedro
Reply to: