7+ days of dogfood

Erik Cederstrand erik at cederstrand.dk
Mon Feb 11 13:36:18 UTC 2013

Den 11/02/2013 kl. 13.07 skrev Erich Dollansky <erichsfreebsdlist at alogt.com>:

> ok, I agree that developers could react faster some times. But, isn't
> it more important that the errors are caught at all?

Yes. As long as there is no alternative, the best motivation to fix a bug is when you just spent 20 minutes recovering your laptop from a crash.

I just believe more people would be motivated to run CURRENT if there was at least some basic runtime resting. Just like the tinderboxes save developer CPU cycles by saying "Don't bother with this revision, it doesn't compile", I think runtime tests would save real developer time by saying "Don't bother with this revision, it doesn't boot" or "Don't bother with this revision, gcc can't compile hello-world.cpp". Which doesn't imply that automated testing would catch all errors.

> So, the best is still if people like me are eating dog food and start
> complaining?
> Do not get me wrong here. I do not complain about the fact that there
> might be an error, I want to help poin-point the error with my
> complaint.

As I started out, I admire the work you and others are doing by running CURRENT and reporting and fixing errors. At the same time, I look to e.g. LLVM and their "no commit without a regression test" goal and think we could do way better :-)


More information about the freebsd-current mailing list