Ouch. From opera.ucsd.edu/paper/fse11.pdf
Conversation
Replying to
I suspect survivorship bias. If you didn't get it right in the initial stable release then you're also relatively likely to not get it right second time around. But their basic methodology seems shaky anyway, so it's hard to know where to begin.
1
1
1
I think it might be something a bit more fundamental than that. One of the issues with concurrency bugs is that it's often not practically possible to write regression tests. Which obviously also influences the quality of merged bug fixes, as they'll not really be validated.
1
2
I think it's strange to focus on incorrect fixes and top-down tooling while saying nothing about the importance of design. The practical context is also not addressed by paper - obviously some bugs are far worse than others, in all kinds of ways, and for all kinds of reasons.
1
2
I estimate that there is a 93.7% chance that the streetlight effect is involved: en.wikipedia.org/wiki/Streetlig