launchpad-dev team mailing list archive

[Robert Collins <robertc@xxxxxxxxxxxxxxxxx>]

On Tue, Aug 9, 2011 at 7:27 AM, Robert Collins
<robertc@xxxxxxxxxxxxxxxxx> wrote:
> On Tue, Aug 9, 2011 at 7:02 AM, Aaron Bentley <aaron@xxxxxxxxxxxxx> wrote:
>>> This is a classic priority inversion, and the normal scheduling fix is
>>> to grant the higher priority to the task holding the resource needed
>>> by the higher priority.
>>
>> I guess I'm reluctant to do priority inheritance because fixing bug B
>> may not be necessary or sufficient to unstick bug A.
>
> Thats fair enough; we need to make a judgement call each time this occurs.
>
> Here are my personal rules of thumb (which as I write this appear to
> just be capturing something more fuzzy at the back of my head :P):
> If I *am* going to work on B so that A is easier to work on, I *have*
> granted it higher priority, so I should record that in the bug
> tracker.
>
> If I *might* work on B for the same reason, I have not *yet* granted
> it higher priority, so I would merely note in the bug that it is an
> option for making A more tractable, but not raise the priority.

Aaron has starting using 'priority-inheritance' as a tag to indicate
when a bug is only (high|critical) because its a dependency of another
bug. I think this is great and have added that tag to our docs on the
wiki + the official tags list for auto-complete.

I've also update the bugtriage wiki page to explain that we use
dependency relations as a reason for a bug to be critical.

Cheers,
Rob