|
Ping Eric Sorenson. First of all, what is the desired behavior here?
To me it seems that If it is not possible to assert if a resource should be collected or not it would be best that the compilation errors out instead of just logging and continuing. The problem here is that the logic is indeed broken, and the result produced by simply skipping broken resources may be something different than what the user intended. Is it really a valid use case to skip broken resources and collect the rest?
It is clearly a problem that the previous error condition is logged again. Is that all that should be fixed here? Is the rest ok? In 5.0.0 we could fail the entire compilation instead of just logging (the "logging twice problem" then goes away).
|