[Tizen General] Tizen platform development tools survey : dropping 32bit linux support

Carsten Haitzler (The Rasterman) tizen at rasterman.com
Tue Sep 24 02:21:25 GMT 2013


On Mon, 23 Sep 2013 16:21:08 +0300 Artem Bityutskiy
<artem.bityutskiy at linux.intel.com> said:

> Carsten,
> 
> I do agree with many of your points. But I really would like to refrain
> from diverging from the main point. Let me try to summarize like this.
> 
> The team briefly explained why they would like to drop 32-bit support:
> 
> a) They suspect there are no/very few users, and the survey is about
> learning about this for sure.

i answered that. the vast majority of the office around me uses 32bit and isn't
going to change in a hurry due to IT/security restrictions. this is 100's of
developers working on tizen all day, every day. not weekend hobbyists. :)

> b) There are specific bugs related to quemu which they have no resources
> to fix. These bugs are long-standing, and there is no indication that
> they are going to be fixed.

well that's fine. resources need allocation, but the point here i and clark
are making are that 32bit support is needed for various reasons, not just in 
self serving "i need it" ways, but in much broader ways too.

the support can be dropped, but it will have consequences.

> That's about it, AFAIU.
> 
> Now, "self-hosting" thing is great. "If mic/gbs works for PC, it should
> just work for Mobile" is even better. I am all for it. Would be really
> cool, honestly.
> 
> But I think the tools team is very practical. They have limited amount
> of resources, and hours in the day. They are trying to drop less
> important things in favor of more important.
> 
> I fully understand them. They have a feature which they carry, which
> consumes resources (both HW and human), but does not work for long time.
> They have a challenge - fix it or drop it. 
> 
> To make them fix it we should come up with good justification why it is
> important right now. Or just send them a link to patches in upstream
> quemu which fix the problem. Or something like that. Otherwise, why
> would they support the feature?
> 
> And then the other point I was trying to make is that there are many
> problematic areas, and I gave an example with signing images. Or look at
> how many troubles people have with mic / gbs and arm builds, check
> archives. Should the tools team invest more time to that, may be? I
> think yes.
> 
> So, judge yourself. My judgment is that I'd rather spent time on
> different things, which I personally believe are more important. I would
> be happy to also see mic/gbs to "just work" on Tizen PC/IVI/Mobile, but
> for me personally this is somewhere lower on the wish list.
> 
> But hey, this is a survey, everyone is free to express own opinion :-)
> 
> P.S. FYI, I am not from tools team, I am just the Tizen IVI kernel
> maintainer.

sure. that's the point of the request - get freedback on people needing/wanting
32bit x86 support. i responded. with more than just a selfish "i need it",
but with a more general and wide-reaching set of side-effects of dropping
such support. :)

personally i don't like the qemu misc binary emulation hack. it's cool and
neat, but really complicates a build env. i personally rather prefer a pure
cross-compile setup (ala OE) OR a full native build (so run qemu-system
emulation and a full os inside of it doing the builds - this also leads to a
nice bi-product of being trivial to "accelerate" by just slapping it on real
hardware and using a distcc cross-compile farm)

-- 
Carsten Haitzler (The Rasterman) <tizen at rasterman.com>


More information about the General mailing list