[Piuparts-devel] short review on daves branches

[Andreas Beckmann]

On 2012-05-26 20:04, Dave Steele wrote:
> On Sat, May 26, 2012 at 9:56 AM, Andreas Beckmann <debian at abeckmann.de> wrote:
>> *  remotes/dave/rrdeps
>> I'm curious, cherry-picked and tested it  :)
> ...
>> - should this be done for dependency-failed-testing,
>> dependency-does-not-exist and dependency-cannot-be-tested, too?
>> - the numbers should be separated by ', '
>>
> 
> I need a reading from the jury on this. Right now it only displays the
> counts for failed-testing and cannot-be-tested, and sorts by block
> count in those cases. I can:
> 
>   - Add the numbers to the other states, and sort in all
>     cases by reverse dependency instead of block count (does block
> count count for a
>     package that hasn't actually failed yet?) (successfully-tested
> sorts by dependency
>     count)
>   - Add the numbers to other states, and go ahead and sort by block
> count everywhere
>     (successfully-tested sorts alphabetically)
>   - Leave it like it is, with reporting only on the failed pages
>   - Make a more invasive change to handles cases (display and sorting) uniquely

We probably need to test it and see what gives most information ...
rdeps+blocked is probably only interesting for failures, but it should
be ok to apply this to all error states

For waiting states it may be tempting to test the biggest blockers first
- but only in a single slave scenario. otherwise the first slave gets
all the interesting packages, so some randomization is probably better.
And for better waiting processing it would be more interesting to
measure the depth of the dependency tree ... as thats the minimum number
of runs we need to process the stuff

for successful states counts could be displayed, but sorting should be
kept alphabetically


Andreas