[Xorg-driver-geode] proposal: always build with -march=geode

[Writer, Tim]

On Tue, Jul 13 2010, Gaetan Nadon <memsize at videotron.ca> wrote:

> On Tue, 2010-07-13 at 17:11 -0400, Writer, Tim wrote:
>
>     Hmmm. I see a desire to build upstream Xorg from source and being an
>     unsuspecting builder/newcomer as mutually exclusive. While this
>     particular error message is a little opaque, the solution is straight
>     forward and only an apt-get/yum/YaST, etc. away. I think most developers
>     will have no trouble understanding that building 32-bit anything on a
>     64-bit platform requires at least a basic 32-bit development
>     environment. Once we have 32-bit builds working correctly on 64-bit
>     systems, it would be useful to add the above error to a FAQ (is there a
>     FAQ for xorg-driver-geode?) but what else can we do. Most (all?) Linux
>     distributions are packaged this way, with separate 64-bit and 32-bit
>     development environments. Even if we would prefer it another way, what
>     can we reasonably do about it?
>
>
> I agree. Building on as many platforms as possible should not be a
> goal for geode.

Agreed. I thought that there was fairly broad agreement in this thread
that building a working i386 driver on x86_64 is a useful common case
that will benefit many developers.

> Because only a dozen or so out of the 240 xorg packages do not build
> on large number of platforms, some people don't know if it is a real
> build error and report problems.
>
> The patches I originally suggested were meant to address this low priority
> "feature".

I thought they addressed the less common case of building a non-working
but compilable i386 driver on x86_64. Maybe I was confused.

> This seem to have resurrected past proposals which are outside my
> competencies.
>
> Bottom line, there is a requirement to build 32 bit driver on 64 bit
> host and this takes precedence.
> I am ok with aborting the config with some message to avoid bug
> report.

Me too. My main concern was that we avoid building non-working drivers
because that seems to have more traps for newcomers.

> There is still the second I mention to verify.

I'm not following you.

Tim