On 11/6/23 9:50 AM, Peter Dimov via Boost wrote:
Robert Ramey wrote:
Test coverage is important, but tests written beforehand can never reasonably cover every usage scenario on every (as of yet unknown) platform.
That's why the correct methodology is, when a downstream use is broken by a change, to add tests that make sure breakage doesn't occur again.
Agreed. Over 24 years that means one has a lot of tests. This is the only way to make sure one isn't just cycling the same changes in and out. That is, it's the only way to guarantee that library quality is constantly improving: Kaisan
I'm almost 76 year old. Is it really a good idea to depend on me for the future integrity of the Boost release?
Certainly not. But that's not the question. The question is which approach produces higher quality releases, and so far each time you've said "I think that doing X will improve the quality" you've been, in my opinion, wrong.
LOL - Hows the current approach working for us? Robert Ramey