James Carlson wrote:
> Darren J Moffat writes:
>>>     I would prefer to leave this as a straight 2.2.4 for 2.0
>>> replacement and open a subsequent case for the EOL/EOF of Apache 1.x.
>> That is fine with me.
> 
> With the clarification that this case seeks Minor release binding, and
> the fact that 2.0 is in S10, what's the upgrade story for users going
> from S10 to S10+1?
> 
> Since we're reusing the same path, it sounds like applications may
> just break after upgrading, and there's no way to do a transition.  Is
> that correct?

Yes this is correct. The current proposal *overwrites* the existing 
/usr/apache2 
binaries with binary incompatible bits [corresponding directories in /etc and 
var will not be overwritten destructively]. The result is that *existing* 
customer modules will break. There is no transition path.

> Does this case also cover upgrading subversion from libapr-0.9 to
> libapr-1.2?  If not, then is delivery of this project dependent on a
> future case that does this upgrade?  We've already approved subversion
> in Solaris (PSARC 2006/563), so I think this project is incomplete if
> it doesn't address the compatibility issue.

Subversion should link against the new apr-1.2 bits, not against the old 
apr-0.9 
bits.

> The only answer I've seen so far is from Stefan Teleman saying that
> subversion would have to go through a "full regression test" --
> whatever that means -- 

That means that it should run all the tests provided with Subversion and 
ascertain that all of them pass.

> but no indication of who would actually do the
> work to switch this over, and whether that delivery is part of this
> project, or if the existing SUNWsvn just stops working.

In the currently proposed scenario SUNWsvn will break.

--Stefan

-- 
Stefan Teleman
Sun Microsystems, Inc.
Stefan.Teleman at Sun.COM


Reply via email to