Error due to LPG option `-same_object` about app-trapper HOT 8 CLOSED

Hi Sebastian,

First of all I apologize for a (very) late response. For the former example, as you suggest, the online LPG option was unlucky (as for the most of the benchmarks). I have updated the example (problem-1 of the Airport domain) for which the online LPG version works.

Now, I believe that it is in a more presentable shape.

from app-trapper.

Now, when I delete that option from the commands in def.h, the system works, LPG runs, plans are generated, though at some point LPG gives segementation fault:

Computing mutex... 
 Unreacheble Goal 39 :  AT(P2 L01)
 Unreacheble Goal 39 :  AT(P2 L01)
The problem is unsolvable since at the fixpoint level the goals are mutex or not reachable


LPG: Segmentation Fault - Seed 748533630
Checking plan: soln.tmp
Bad plan file!


Failed plans:
 soln.tmp 

FAIL

from app-trapper.

If I remember well, LPG which is used with the APP is modified such that it generates "endstate.txt" files. I'll have a look on different versions of LPG that might support the -same_object option. The question is, however, whether it can solve the "Segmentation fault" problem as it seems to be happening after realizing that the problem is unsolvable (at least according to the log).

from app-trapper.

I have checked different versions of LPG and only the one we have in /tools/ gives segmentation fault. It happens when the problem is (trivially) unsolvable. I suspect it has something to do with a modification providing the "endstate.txt" file. Although I believe that the segmentation fault issue is harmless in this case, I might provide a different version of LPG to be used for planning.

I also checked the option "-same_object" which seems to be a typo as it works with "same_objects" (although I don't think that omitting this option makes any harm).

from app-trapper.

Thanks @lchrpa ! I just came back from summer break, and looking at this as seems to be the only major issue remaining.

You are right the issue seems to arise when the problem is not solvable, as it prints:

The problem is unsolvable since at the fixpoint level the goals are mutex or not reachable

However, if so why do we have it as our "example"? Am I missing something? Is it because the online version is "unlucky" and does not find it?

The offline version seems to find a solution (../src/pp-lpg), so the problem is itself solvable, which is good! The online seems to fail to find a solution but also crashes at the end.

I ran the online version 10+ times and never got a solution, I wonder what we want to leave for the "example". I guess it would be good to have an example which the online is able to solve, at least sometimes right?

What do you think?

from app-trapper.

Great Lukas! No need to apologize, I appreciate you are still interested to clean this up together with me, I would not be able to do it myself at all! :-)

I tried the new example and works now:

Solution number: 1
Total time:      0.01
Search time:     0.00
Actions:         18
Duration:        18.000
Plan quality:    18.000 
Total Num Flips: 1
Distance from input plan: 0 
     Plan file:       plan_pfile.pddl_1.SOL


SUCCESS

I think I will provide this example and the previous non-working to show that the online is often "unlucky". Done!

Thanks!

from app-trapper.

ops, I should leave it open so that I don't forget to have this and the other (non-working unlucky example)

from app-trapper.

This is done.

from app-trapper.