Re: problem: world boot strapping from 7-C to HEAD with CDDL on sparc64?

On Thu, 26 Jun 2008, Ruslan Ermilov wrote:

> On Tue, Jun 24, 2008 at 08:57:54AM +0000, Bjoern A. Zeeb wrote:
>> On Mon, 23 Jun 2008, John Birrell wrote:
>>
>>> On Mon, Jun 23, 2008 at 05:46:00PM -0400, Ed Maste wrote:
>>>> All that said, I'd really like to be able to continue building head on
>>>> RELENG_6.  (In my case I have a rather fast RELENG_6 box so I have been
>>>> doing my pre-commit testing there.)  Do the changes we're discussing
>>>> here involve anything that makes that difficult?
>>>
>>> Possibly. :-)
>>>
>>> It depends on whether your RELENG_6 box is up-to-date with respect to the
>>> RELENG_6 branch.
>>>
>>> I think that as current diverges more from 7 it will be harder to keep the
>>> tree building on 6.
>>
> Are we talking about changing the build-/cross- tools?  If not,
> it shouldn't be affected.
>
>> I think the problem here simply is that you are relying on the headers
>> to be installed on the base system instead of in src|obj copies. I do
>> not know if/how we do it for bootstrapping though.
>>
> If it's about build- and cross-tools, then yes, they rely on
> the host compiler, headers, etc. because they will be run on
> _this_ computer, using its CPU and the current kernel.

So if a bootstrap tool cannot be build - what's the solution?

The problem is way beyond building HEAD on 6 as we just found out
here -- you can no longer update a 6.2-STABLE to 7-STABLE either
without going through a latest 6-STABLE or building WITHOUT_CDDL
and then rebuilding on 7. Either way its two builds.

If I am not mistaken not even 6.3-RELEASE could be source updated to
7-STABLE. And 6.3-RELEASE is the only supported RELENG_6 release at
the moment.  This is from one STABLE branch to the next STABLE branch.
I think this is unacceptable, no matter what and how.

-- 
Bjoern A. Zeeb              Stop bit received. Insert coin for new game.