Stress testing repeatedly runs code, often with multiple threads, to increase the chance of exposing bugs that depend on a particular scheduling sequence. More runs can increase confidence, especially for fine-grained tests, but cannot guarantee that a problematic schedule will occur. section 11.2.4
Interpret passing results cautiously:
- A test may never trigger the circumstances needed to reveal a bug, regardless of how often it runs. section 11.2.4
- Hardware and operating-system behavior may prevent relevant circumstances from occurring on the test system. section 11.2.4
- Large tests may cover too few of the possible scheduling permutations to provide strong confidence, even after many runs. section 11.2.4
Repeated queue tests
A fine-grained test of a thread-safe queue can be run repeatedly to build confidence; a single successful run provides less reassurance than many successful runs. section 11.2.4
