Closed nkanderson closed 3 days ago
Not to give ourselves more work... but a script to generate random test files would be pretty easy.
Yeah, I think some randomly generated patterns would be great. I also think we should try to identify any specific patterns that represent some particular functionality we need to test. I'm not sure I understand the events well enough yet to come up with an example relevant to our context, but more generally, thinking about whether there are sequences of states we go through that lead to a bad or hung state if they're not handled properly. If we have some named test cases to run through, we can probably identify the issue more quickly if they fail.
Yeah, they might actually need to be random with some constraints to be realistic/useful. Again, something I think will come into focus in the coming days.
Superseded by #42
When we understand the trace events well enough, we should come up with a handful of event sequences (maybe 2-4 to start with) to test with. They should probably focus on the simplest cases to begin with.