Is the current way to count the number of actual data races the right way to do it?

Question

Is the current way to count the number of actual data races the right way to do it?

jcp19 opened this issue 5 years ago · 1 comments

Check if the data race candidates are actually a pair of locations in the code. That is the only way to ensure that each data race candidate is counted only once.

Answer 1 · 2019-07-31T15:14:29.000Z

It seems that most of the detected data races are the same in the sense that two different CausalPair correspond to the same pair of locations even though they contain different events. It is the pair of locations that seems to matter. As such, I implemented a countDataRaces method in the RaceDetector class that counts the unique pairs of locations. The evidence is anecdotal but the results seem more adequate our test case.