On Mon, Apr 4, 2016 at 2:25 PM, Charles-Antoine Couret
Le 04/04/2016 20:42, Justin Forbes a écrit :
> Generally I would not be opposed to something like this being tied to an option in
.config, so users can choose whether or not they wish to disable retests
It's a good idea, I will improve my patch. ;)
> (I personally retest frequently).
What is the interest to retest with the same kernel?
Several reasons, testing for race conditions, testing after certain
conditions have been met, testing with different VM configurations on
the same VM.
> Of course the default behavior there would need to be allowing
them (unchanged), but you can add disable_retest=1 or something to disable them.
Or maybe inverse? Default case without retest.
It depends if the use case of retest is popular or not, in fact.
Actually, this has nothing to do with the case being popular, it is
more about changing behavior on existing test systems. People should
be able to update without changing the behavior they are used to.
> Instead of making this a new script, incorporate it into
runtests.sh with the .config checking and you can use $testset to verify. You should
probably also use $logdir in case people override this locally.
Irrelevant: A RPM package as planned for this application?
I have built one, but it is a weird package in that it needs to end up
somewhere easily writeable by whoever is running the test, we actually
build things at runtime, etc. Now that we have a config file it might
be worth revisiting.