gp pre-release and OS5 problems (was Re: Call for testers!)

[Matt Davey <mcdavey mrao cam ac uk>]

On Thu, 2006-04-13 at 18:47 -0400, David A. Desrosiers wrote:
> On Thu, 2006-04-13 at 00:24 +0100, Matt Davey wrote:
> > apologies.  I'm tired.  In the gp-patch in the previous mail I
> > should have changed a different instance of PI_PF_NET to
> > PI_PF_DLP.  The previous patch can safely be applied too. 
> 
> With a few pilot-link patches (including yours in bug #1639 on our
> side), it crashes immediately on the very first conduit (Address).

You could test basic gp operation by disabling affected conduits.

>  I
> suspect this is going to be a big problem, because AddressDB.pdb doesn't
> physically exist on my Treo650, nor do any of the "Classic" databases. 

Ouch.  I guess we should at least avoid the crash.  Does that occur on
the pilot-link or gp side?  Are you referring to the evolution
e_address_conduit or something else?

> I've been tracking this down for a few days now, and it looks like the
> PACE layer (which transparently translates between the "Extended" (new)
> format databases and the "Classic" (legacy) formats), doesn't exist
> during HotSync, and so no translation takes place. This means a query to
> open a db with the CreatorID of 'addr' will fail, since that doesn't
> actually exist on the device. 

Is there any workaround that could be added to the conduits, after
detecting the OS platform and pilot-link version?  Otherwise we're in
pretty bad shape: the evo conduits would have to give up if a OS5
platform is detected and the pilot-link version is less than X :(

> I'll test with an OS4 device. I don't think we're going to be able to
> support "virgin" OS5 devices at all, if we can't solve this problem.

This is bad news.  I guess at a minimum we need to avoid the crash and
handle the error in affected conduits.  I'd presume you don't want to
delay 0.12.0 until we can support OS5.

Matt

Matt Davey		Only the dead have seen the end of war.
mcdavey mrao cam ac uk 				-- Plato