On Fri, 19 Dec 2008, Dave Miner wrote:

> Alok Aggarwal wrote:
>> Here is what I'm thinking of implementing in the AI engine
>> as far as dry run functionality goes -
>> 
>> - Parse and validate the manifest obtained from the AI
>>    webserver
>> - Set up installation parameters based on the information
>>    in the manifest
>> - Initiate target discovery to either select a target
>>    device specified in the manifest or pick a default target
>> - Initiate target instantiation in dry run mode to do
>>    some sanity checking
>> - Print out the installation parameters such as target
>>    device, IPS repo name/port, list of packages to be
>>    installed
>> 
>> Note that the IPS repo won't be contacted as part of this
>> process so if there are problems with name resolution, etc.
>> the dry run won't catch those.
>> 
>
> Seems like you could easily create a temporary package image area and then 
> run pkg install -n to it.

That's a useful thought and it will allow us
to validate repo information as well as check
validity of packages. I'll add it to the list.

>> Would this be sufficient for test purposes?
>>
>
> Test isn't the only consumer here; I'd expect that end users would be more 
> satisfied if you can dry-run all of the critical operations.

That's good to know.

Alok

Reply via email to