Very occasionally (e.g. Travis CI job #351.1) tests fail very randomly that doesn't make any sense.
My suspicion is since fill_random doesn't check if the new random value is indeed different from original value, tests which assume fill_random will always corrupt data may fail very rarely (1/256 chance) as a result.
Very occasionally (e.g. Travis CI job #351.1) tests fail very randomly that doesn't make any sense.
My suspicion is since
fill_randomdoesn't check if the new random value is indeed different from original value, tests which assumefill_randomwill always corrupt data may fail very rarely (1/256 chance) as a result.