I believe whenever the annotator runs to either build the target itself or to build downstream deps, it expects the compilation run to succeed, i.e., complete without compilation errors. (NullAway needs to be run at the WARNING level for this to occur.) If this is true, then annotator should detect when a build command fails, print an appropriate error message, and terminate the run. Right now, the annotator ignores build failures, which often leads to an incrutable failure later in the annotator run, like crashing due to the absence of some expected tsv file.
This could be implemented by directly checking the return code of the executed command in executeCommand or at callers of this function if that's better.
I believe whenever the annotator runs to either build the target itself or to build downstream deps, it expects the compilation run to succeed, i.e., complete without compilation errors. (NullAway needs to be run at the WARNING level for this to occur.) If this is true, then annotator should detect when a build command fails, print an appropriate error message, and terminate the run. Right now, the annotator ignores build failures, which often leads to an incrutable failure later in the annotator run, like crashing due to the absence of some expected tsv file.
This could be implemented by directly checking the return code of the executed command in
executeCommandor at callers of this function if that's better.