> > it's still more testing than you can get by manually firing up a few games that you > > have modified, and it's mostly done right before an intermediate update > > that's why I claimed those to be *usually* more stable > > It only picks up crashes or display changes in the first couple of seconds after > bootup, while it can find some issues it misses a whole load more.
I feel like we're going in circles. I never denied that the current regression test suites only catch a minority of possible breakages, still
1. far from intermediate releases not even these small regressions are caught 2. nobody has proposed and put in action a better (=more effective) solution
so I'm grateful to the persons who spend their time running these regression tests, no matter if they only catch 5% of the regressions: it's still more bugs that can be fixed before a release than without their tests
> > The testing takes a while as well, so changes can be made after it's started.
As I've said, I have the habit to stop adding changes close to intermediates releases (except for softlist ones) to avoid this. maybe more devs should adopt the same strategy, unless they think they are unable to commit the code in the days after the release...