Blocking radeonhd (was Re: Fedora 12 Graphics Issues: Cancel F13 and concentrate on fixing F12 ?)

[Hans Ulrich Niedermann]

On Tue, 1 Dec 2009 10:35:52 -0500
Bill Nottingham <notting at redhat.com> wrote:

[ radeonhd vs radeon ]

> So, if our X maintainers won't handle bugs with it, we have a working
> default alternative that is maintained upstream, and it's *known* to
> be broken in the default configuration, why ship it? If we're trying
> to focus on quality, I'm not sure why we'd ship something that's known
> broken.
> 
> Hans, are you OK if we block this from rawhide?

>From where I stand, there are a number of reasons both for and
against having a radeonhd package in Fedora. Most of those reasons will
have different importance for different people.

The reasons I see for having radeonhd in Fedora all boil down to
radeonhd and radeon containing different sets of bugs, and triggering
different sets of bugs in other software components (and probably also
hardware).

Often, those issues can be hard to find if the exact hardware is not
available to the developers, and thus take quite long to fix. See e.g.
http://airlied.livejournal.com/68550.html

There have always been cases of one driver working for people while
the other does not, and vice versa.

The complexity of the whole graphics system suggests this will
probably not change soon.

  For keeping radeonhd in Fedora

    K1. Giving users a working system using the other driver during the
        weeks or months needed to fix a bug in one of the drivers is
        good for users.

    K2. Easy availability of another driver to try makes locating the
        bug easier: Is the issue common to both drivers, or different
        or not present at all with the other.

  For blocking radeonhd from Fedora

    B1. Less work for me and Matej in bugzilla.

    B2. Less bugs mistakenly assigned to radeonhd by the reporters.

    B3. Lacking an alternative, the pressure to fix bugs with radeon
        would increase (and hopefully improve things).

    B4. radeonhd requires some nomodeset kernel parameter, depending on
        kernel version.

As to the KMS issue, I do not see where to communicate to users
that radeonhd needs KMS off but in README.fedora. radeonhd upstream do
not support KMS.

All that said, I have been mostly running radeon with nomodeset on my
F11 system (ThinkPad T60, X1400/rv515) for the last few months, so for
me personally, I would not lose much by radeonhd being removed from
Fedora. I have not had an opportunity to test the state of affairs on
F12 or even rawhide, and also have no R6xx, R7xx, R8xx chipsets, so I
cannot comment on any of that.

-- 
Hans Ulrich Niedermann