Closed gustavogaldinoo closed 6 months ago
Why isn't the existing check sufficient?
I recall the current code only checks benchmark-type consistency at the end of experiments and cannot generate a report if inconsistent.
This warns users earlier and helps save time and computation power.
Why isn't the existing check sufficient?
I recall the current code only checks benchmark-type consistency at the end of experiments and cannot generate a report if inconsistent.
This warns users earlier and helps save time and computation power.
For sure checking ahead of the experiment is important. But if we aren't doing this currently, there's a bug to track down here that this PR hasn't fixed right?
Why isn't the existing check sufficient?
I recall the current code only checks benchmark-type consistency at the end of experiments and cannot generate a report if inconsistent. This warns users earlier and helps save time and computation power.
For sure checking ahead of the experiment is important. But if we aren't doing this currently, there's a bug to track down here that this PR hasn't fixed right?
Ah actually I see where this might be fixed. Gustavo do you know why this didn't work previously?
Currently, Fuzzbench fails to generate a report if benchmarks are a mix between bug and coverage benchmarks.
This PR adds a method to check if benchmarks are a mix between bug and coverage benchmarks before starting the experiment, to spot the problem early and avoid wasting time.