Re: [aspectj-users] pointcut to match "return" statements

[Alexander Kriegisch <Alexander@xxxxxxxxxxxxxx>]

A) Returning null is not necessarily a programming error. Not handling null results gracefully might be though.

B) How should it be possible to statically determine a non-trivial (dynamically created) return value during compile time? If you want to catch typical, statically determinable bugs using a byte code analyser, I recommend Findbugs.

-- 
Alexander Kriegisch


> Am 06.01.2014 um 15:50 schrieb Sean Patrick Floyd <seanpfloyd@xxxxxxxxxxxxxx>:
> 
>> On 06.01.2014 15:26, Alexander Kriegisch wrote:
>> You can neither match "return null" nor assignments to local variables, only assignments to member variables.
>> 
>> What you can do is match returning calls or executions and check their results dynamically like this:
>> 
>>     after() returning(Object result) : execution(!void *(..)) {
>>         if (result == null)
>>             System.err.println(thisJoinPoint.toString() + " must not return null");
>>     }
>> 
>> As you can see, I am only matching non-void methods because void methods always implicitly return null. You can further refine the pointcut if you want to restrict it to being applied to certain packages or classes or if you want to match boolean methods returning false, int methods returning 0 etc.
> I'm aware of that, but I am looking for a way to do check statically in a policy enforcement aspect (at compile time).
> Detecting programming errors at runtime is not as reliable.
> _______________________________________________
> aspectj-users mailing list
> aspectj-users@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/aspectj-users