Generated Format Testdata
Background
As one can see from the test coverage reports we lack of in depth testing on the demuxers and parsers. Testing those is important to be able to deliver standard compliant and robust implementation. Shipping a huge ammount of testfiles with the tests is not feasable.
Task
The scope of the project would be to use a tool to generate a set of testdata form a formal format description. One such tool that could be used is peach fuzzer. The idea here is to generate not all possible mutationl of a file (which is close to impossible), but instead create a subset that is likely to trigger the majority of bugs. To give an example, one would create variants of wav files that contain 0,1,2,17,255 audio-channels - 1 and two would be likely cases, 0 and 255 corner cases and 17 a random value in the middle.
Resources
- GStreamer test coverage
- peach fuzzer Mentor: Stefan Kost - ensonic on IRC
Back to TaskList