> On Jul 6, 2020, at 3:48 PM, Jed Brown <[email protected]> wrote:
>
> Barry Smith <[email protected]> writes:
>
>> We should be more clever than checking just examples but also check source
>> code. If, for example, only KSP is changed then tests could skip the
>> sys/vec/mat levels saving time.
>>
>> If we did this in the pipeline they could be significantly faster; all the
>> time saved not testing sys to mat for Matt's changes :-)
>>
>> Is there a fallacy here?
>
> I think we could pretty readily do this at the granularity of package.
> Significantly harder (would need to interface with a coverage tool) to do it
> at finer granularity.
>
> What fraction of our testing time is currently at petscmat and below?
No idea, but could easily be 1/3
