On Mon, Jul 6, 2015 at 12:27 PM, Kevin Grittner <kgrittn@ymail.com> wrote:
> What would it take to get something like that which uses the
> check-world target instead of just the check target? Without the
> additional tests (like the isolation tests), some of these numbers
> don't reflect the coverage of regularly run tests. While it is of
> some interest what coverage we get on the 30 second `make check`
> runs, I would be a lot more interested in what our coverage is when
> the full 8.5 minute set of regression tests we have in git is run.
I agree. Obviously in some cases the coverage is bad because
concurrency is naturally required. It's even more glaring that
recovery codepaths are always all-red in coverage reports.
--
Peter Geoghegan