Hi Shawn.
Thanks for your reply.
Shawn Walker wrote:
> Jack Schwartz wrote:
>> Hi everyone.
>>
>> Sometimes installer environments lack a needed driver to complete the
>> installation. The Driver Update project adds packages with such
>> drivers to the booted environment before starting the installation.
>> For AI, manually-added drivers are specified in the manifest. IPS
>> packages are specified using P5I files. (Drivers can also be
>> searched-for, like the current DDU does today.)
>>
>> Question: What is the best way to handle the situation where the
>> beginning environment specifies a publisher with a name and URL-A,
>> and a P5I file given in the manifest specifies the same publisher
>> name with URL-B?
>>
>> The packagemanager (b127) neatly skirts around this issue by not
>> allowing such a clash. Should Driver Update do the same?
>>
>> I ask because manually running pkg set-publisher to a new publisher
>> (e.g. going to pkg.opensolaris.org/dev from p.o.o./release) and then
>> doing a pkg image-update is allowed. (... or did I hear that this
>> just changed?)
>>
>> If the answer is yes and Driver Update is allowed to do the update,
>> should the new publisher be kept on the system, or should the
>> publisher be reverted back to the old one? Either way, some
>> installed package will be put out of sync with the current
>> publisher. (I would think that in most cases, one of the two URLs
>> would be a more evolved version of the other, and that the one being
>> set by Driver Update is the newer one; but these could be invalid
>> assumptions.)
>
> If there's a conflict with an existing publisher, there's no way to
> guarantee that the new URI is correct or that the old one wouldn't
> work just as well.
>
> Dan, I, and Stephen have discussed how to deal with these conflicts
> but I don't believe we have good answers yet.
>
> In the short term, my personal suggestion would be to abort on
> conflicts, or continue and warn the user that the operation may not
> succeed.
Of your two suggestions, probably the best thing to do would be to log
the warning and try to continue. If one publisher is a more evolved
version of the other, the likelihood of success is high. And if they are
different, then the warning is still there to show what is happening.
Thanks again,
Jack
>
> I do not believe that you want to silently, automatically change the
> origins or mirrors for the publisher.
>
> Cheers,
