Open lfoppiano opened 6 years ago
I am not sure that running the PMC 1942 PDF tests with CI is a good idea: heavy process that can take more than one hour on a good machine. It's really something tied to development and tuning. In addition there is no mechanism to store and compare metrics for this end-to-end evaluation over time, so the automation could not be exploited at this time.
Personally I think there is even value to have that in the absence of automatic metric comparison. Once you notice a degradation you could more easily narrow it down later. Of course it would be nice have an automatic comparison like there is for coverage. Perhaps an alternative would be to run over a subset of the PMC 1942 tests or run it delayed rather than on every commit. Well, just an idea anyway.
Some ideas gathered with @de-code: