It would be nice if the analysis would detect situations in which all compilers fail (at least, fail in the same way), and assign them a 'potential fuzzer bug' status.
One unknown here is how to represent this status:
one separate status code (which would mean that we lose information about exactly what kind of way the bug exhibited);
shadowed status codes, one for each original code (bit messy and redundant);
separate tracking of 'basic' status codes and analysis inferences (which would complicate our present way of analysing overall subject statuses across compilers, but that'd probably need to be made more sophisticated anyway);
This is a similar issue to what happens when we filter statuses; ideally, we'd carry around the specific pre-filtering status, apply a 'this status is filtered' tag, and then the dashboard would use that while more in-depth analysis readers might specify what it was filtered from.
It would be nice if the analysis would detect situations in which all compilers fail (at least, fail in the same way), and assign them a 'potential fuzzer bug' status.
One unknown here is how to represent this status:
This is a similar issue to what happens when we filter statuses; ideally, we'd carry around the specific pre-filtering status, apply a 'this status is filtered' tag, and then the dashboard would use that while more in-depth analysis readers might specify what it was filtered from.