Why would Daikon generate the same invariants for two sets of traces?

[Soha Hussein]

Hi,

I am trying to use Daikon to generate invariants in incremental iterative mode, that is initially I create a set of traces T1 using Chicory, run Daikon then get invariant set S1. In the next iteration, I create a new set of traces T2, such that T1 \subset T2, again I do that by running Chicory on new test input to get invariant set S2. The problem I am seeing that S1 is identical to S2, even though T2 has more non-repeated tests that T1. Now I am not sure why Daikon has not learned from the new tests in T2. Is there any explanation for that?

Thank you in advance,

Soha

[Michael Ernst]

Soha-

> Invariant set S1 is identical to invariant set S2, even though traces T2 has more non-repeated tests than traces T1.

It is possible that the additional traces in T2 do not add any more information than was in T1.
Or, maybe any new invariants implied by the new traces are not statistically justified.

Please see Section 9.1.15 of the manual, which discusses why bigger traces may lead to fewer invariants or more invariants:
http://plse.cs.washington.edu/daikon/download/doc/daikon.html#Bigger-traces-cause-invariants-to-appear
That manual section doesn't explicitly say that the set of invariants might be identical, but that is also a possibility.

Are there specific invariants you expected to see?  The manual gives some strategies for investigating why they did not appear.
If you are stymied, you can shore your traces with us and we'll try to help.

                    -Mike

--
You received this message because you are subscribed to the Google Groups "Daikon discuss" group.
To unsubscribe from this group and stop receiving emails from it, send an email to daikon-discus...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/daikon-discuss/4ad30e9c-385e-41b5-90a4-3831c6a3d915n%40googlegroups.com.