[Piuparts-devel] testing sid with --warn-on-leftovers-after-purge, too

[Andreas Beckmann]

On 2013-01-02 13:57, Holger Levsen wrote:
> Hi Andreas,
> 
> On Mittwoch, 2. Januar 2013, Andreas Beckmann wrote:
>>> hm, do we really need to test almost 40k extra packages just to get this
>>> difference?!
>> The good thing: seeding is easy
> 
> sure, but thats just the seeding. After that we would constantly test 40k 
> packages more. Do you have an idea how to avoid this?

How loaded is piatti? do you have a load average graph over the last,
say, year?

I just got an idea, although it's low priority for me to implement it
(but we should bug it or TODO it to not forget about it)

Example:

[hard]
precedence = 1
...

[quick]
precedence = 5
...
seed-section = hard


flags must be compatible s.t. [hard] in this example does at least the
same as [quick], e.g. by just dropping --warn-on-leftovers-after-purge
the pass/ ogs from [quick] need to be a superset of the pass logs from
[hard]

recycle timeouts need to be compatible, i.e. [quick] needs to recycle
(pass/) logs at the same interval (or better a little bit (1-2 days)
slower) than [hard]

seed-section would need to be evaluated by the master
* during reserve(), before handing out a package to a slave, check
whether the seed-section has a pass log and copy (hardlink) this to the
current section instead of handing it out to a slave


The only useful flag combination for such a setup would be

hard: --fail-on-leftovers-after-purge [--install-remove-install]
quick: --warn-on-leftovers-after-purge



Andreas