This is not a bug - but a 'design' issue which makes the androsim.py results
less relevant.
What steps will reproduce the problem?
1. Take two Dex in particular with some empty methods
private blah() {}
2. Compare the Dex with androsim.
Androsim will say that all empty methods are "identical". Which is technically
true, but does not generally convey any special meaning on the similarity of
files.
3.
What is the expected output? What do you see instead?
I would prune empty methods and not consider them for comparison.
Might be worth also re-thinking what paths should be compared. For example most
apps will have some android.annotation.TargetApi and will thus have similarity
there, but it is usually meaningless.
Such an improvement might be considered for future versions.
What version of the product are you using? On what operating system?
Linux. Androguard from hg, v2.
Original issue reported on code.google.com by aaforti...@gmail.com on 2 Dec 2013 at 3:36
Original issue reported on code.google.com by
aaforti...@gmail.com
on 2 Dec 2013 at 3:36