[Development] QArrayDataObs/Pointer UB post-mortem

[Ulf Hermann]

>     Apparently we are quite often doing changes we believe to be
>     refactorings, but that actually aren't. How do we deal with that?
> 
> 
> This one is easy: It says it right there on the front page of https:// 
> refactoring.com/: "Refactoring is a part of day-to-day programming" [...]

The theory is clear. However, in practice, mistakes happen and 
accidental side effects slip through. Of course I'm very careful to 
avoid this, and we have a lot of tests to prevent accidental behavior 
changes. Yet, I'm not perfect, nor do I know anyone who hasn't been 
wearing a brown paper bag instead of the Two Hats at some point.

In my experience these mistakes in refactorings are more frequent than 
breakage from bad merge conflict resolutions for bug fixes. Most of the 
bug fixes I see are small, localized affairs while refactorings usually 
touch code in many places. As a result, mistakes in bug fixes are 
generally easier to spot than in refactorings.

That's why I try to pick back as little code changes as possible, even 
at the cost of some merge conflicts and somewhat messier code in older 
branches. Sometimes you cannot avoid a refactoring before a fix. That's 
fine then, and then we need to pick back the refactoring, too. But most 
of the time, I can avoid it.

best,
Ulf