The changes to plan_install look good.

brock

jmr wrote:
> Thanks Brock:
>
> I made a few mods here to the install code even though we are not 
> currently using it to take the tuple and handle the exception_caught 
> gracefully.
>
> http://cr.opensolaris.org/~jmr/um_3981__3998_3950_v3_Oct16/
>
> JR
>
> Brock Pytlik wrote:
>   
>> LGTM.
>>
>> Brock
>>
>> jmr wrote:
>>     
>>> Brock, thanks for taking a look. Comments below:
>>>
>>> JR
>>>
>>> Brock Pytlik wrote:
>>>  
>>>       
>>>> updatemanager.py
>>>> Line 1129: Please don't just bump the version without making sure 
>>>> the code works with the bumped version. The reason we have API 
>>>> versioning is so that we fail early if the API is changed 
>>>> incompatibly rather than waiting till we happen to travel down the 
>>>> changed code path. In this case, plan_install now returns a tuple, 
>>>> not a boolean. As far as I can tell, this line hasn't been updated 
>>>> to accept the new return value.
>>>>     
>>>>         
>>> We only have Update All exposed and are not showing Install Updates 
>>> to the user, so this was not tested. I did bump the version and test 
>>> the Update All functionality which worked fine. I will make the 
>>> change here and test the Install Update using a debug mode switch to 
>>> turn this back on.
>>>
>>> What should I do on partial success of plan_install? Is it safe to 
>>> continue? Is it just a refresh that could have failed? We are not 
>>> doing a catalog refresh when the UM is launched from the notification 
>>> panel, as one has already been done externally for the notification 
>>> to be raised.
>>>   
>>>       
>> If you're not refreshing the catalog there, then there's no sense of 
>> partial success. The only way, currently, to get into an ambiguous 
>> state is if catalog refresh failed for one or more authorities.
>>     
>>>> I'm also unclear as to the purpose of __get_api_obj. If the api 
>>>> object isn't created on first trial, why would it succeed after that?
>>>>     
>>>>         
>>> It would not. I wrapped this as we can fetch the api object in a few 
>>> places now, depending on the code path. Previously I had done the api 
>>> fetch up front while getting the updates list, but then when I failed 
>>> with the version exception I could not output the error messages to 
>>> the details pane. Hence the change.
>>>
>>>   
>>>       
>> Ok
>>     
>>>> I'll give more feedback in a bit when I get back from a meeting.
>>>>
>>>> Brock
>>>>
>>>> jmr wrote:
>>>>    
>>>>         
>>>>> Hi,
>>>>>
>>>>> Simple patch:
>>>>> http://cr.opensolaris.org/~jmr/um_3981__3998_3950_v2_Oct16
>>>>>
>>>>> Bumping API version to 1. UM currently failing in the gate as the 
>>>>> underlying API has bumped the version and the call to:
>>>>>
>>>>> api.ImageInterface(...CLIENT_API_VERSION...) with version 0
>>>>>
>>>>> Throws a version mismatch exception.  Also changed how this error 
>>>>> was reported in the Details pane.
>>>>> http://defect.opensolaris.org/bz/show_bug.cgi?id=3998
>>>>>
>>>>> Range of minor UI changes to conform to UI Spec.
>>>>> http://defect.opensolaris.org/bz/show_bug.cgi?id=3981
>>>>>
>>>>> Change Copyright file to short version:
>>>>> http://defect.opensolaris.org/bz/show_bug.cgi?id=3950
>>>>> And removed SUNWipkg-um.import - David Comay told us it's no longer 
>>>>> needed.
>>>>>
>>>>> Thanks,
>>>>>
>>>>> JR
>>>>> _______________________________________________
>>>>> pkg-discuss mailing list
>>>>> [email protected]
>>>>> http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
>>>>>         
>>>>>           
>>> _______________________________________________
>>> pkg-discuss mailing list
>>> [email protected]
>>> http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
>>>   
>>>       
>
> _______________________________________________
> pkg-discuss mailing list
> [email protected]
> http://mail.opensolaris.org/mailman/listinfo/pkg-discuss
>   

_______________________________________________
pkg-discuss mailing list
[email protected]
http://mail.opensolaris.org/mailman/listinfo/pkg-discuss

Reply via email to