New webrev:
http://cr.opensolaris.org/~bpytlik/ips-3799-v2/

Shawn Walker wrote:
> Brock Pytlik wrote:
>> Webrev:
>> http://cr.opensolaris.org/~bpytlik/ips-3799-v1/
>>
>> Bug:
>> 3799 need to start progress tracker going early on
>
> api_versions.txt:
>   line 2: s/useing/using/
fixed
>
>   line 5: missing '.' to match lines 6, 7?
fixed
>
> image.py:
>
>   line 1ff: it would be nice to have a consistent variable name for 
> the progress tracker, in this changeset we have: prtrack, 
> progresstracker, and progtrack.
Sure, making the added variables use progtrack.
>
>   line 2101: you moved evaluate_start here, but isn't this needed for 
> make_uninstall_plan, and other cases as well since it isn't in 
> evaluate() anymore?
Interestingly, it appears that it's not really needed. However, it is 
the correct thing to do, at least to add it to make_uninstall_plan and 
repair. Can you elaborate what other cases you had in mind?
>
> imageplan.py:
>   line 421: you removed evaluate_start, but not evaluate_done?  I'm 
> not certain about this change...it should work, but it seems odd.
For now I'm inclined to leave it in place. The function currently has 
the information needed to call evaluate_done. The point was to move the 
spinner earlier, so I think it's ok. If you want to make the argument 
that I should add another phase to the progress tracker and leave 
evaluate_start where it is, that's fine. Just please tell me what that 
phase should display to the user and I can move to that.

Brock
>
> Otherwise, looks fine.
>
>> Previously, on my machine, it was approximately 8 seconds before the 
>> CLI provided any feedback when installing a package, and about 14 
>> seconds when doing an image-update. (The bug has a more detailed 
>> break down of where time was spent and when the spinner started.) 
>> With this fix, the CLI displays output to the user in approx 1 
>> second. Mostly, the change was to just start the spinner sooner, but 
>> a change was also made to provide feedback while refreshing the 
>> catalogs as that could be a non-trivial wait on a slow network.
>
> Thank you for fixing this as the unknown wait during the catalog and 
> initial manifest fetching was one of my big complaints.
>
> Cheers,

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

Reply via email to