[Sugar-devel] ANNOUNCE: Sugar 0.88 for the XO-1

[Peter Robinson]

On Wed, May 26, 2010 at 2:31 PM, Bernie Innocenti <bernie at codewiz.org> wrote:
> El Wed, 26-05-2010 a las 09:29 +0100, Peter Robinson escribió:
>
>> > I'm happy to hear this. F14 may bring interesting changes for the XO1,
>> > can we hope that support for the Geode won't have been dropped by
>> > then?
>>
>> There was a bug report [1] that was filed about it to do with glibc.
>> Since I last looked at the bug there was an update added mentioning a
>> kernel patch [2] to fix the problem. So by the look of it we could
>> actually get it supported in F-13 with little effort once that patch
>> has been accepted upstream. Yay!
>
> More than a bugfix, this is a work-around for the Geode not being fully
> compatible with the instruction set which Fedora is compiled for.
>
> Emulating the missing NOPL opcode in a kernel trap handler is going to
> be 10-100 times slower than the original instruction sequence. Let's
> just hope that GCC doesn't have the habit of generating any of these
> inside tight loops.

yea, or you can install your own koji infrastructure and setup a
i586/i383 secondary arch, rebuild all of fedora, provide hosting,
servers,storage and infrastructure  infrastructure for it.

TBH I don't know what changed between F-12 and F-13. It wasn't the
compile flag changes as I checked them so I'm wondering wondering why
its suddenly a problem. I have a mostly stable XO-1 running SOAS-2
without issues when fully updated (except for some black icons). So it
could be that for some reason we're suddenly triggering this where we
haven't in the past. Other suggestions are welcome. I'm really not up
to speed on random x86 assembler quirks between chips.

Peter