Message2814

Author silvan
Recipients malte, moritz, raznis, silvan
Date 2013-12-17.17:07:54
Content
One case where this can happen with the current code is if we do
bisimulation-based shrinking and group by h, and the number of different h
values is larger than the target size. In this case the initialization of the
bisimulation already creates too many abstract states.

For now, the assertion has been commented out, since it's not clear what the
best strategy is:
- shrink if we have too many different h values, making the shrink
non-h-preserving, or
- live with the fact that the target bound isn't respected in this case

ParcPrinter is a good domain to test this because it has such large action costs
and hence so many different h values.
History
Date User Action Args
2013-12-17 17:07:54silvansetmessageid: <1387296474.44.0.793103345292.issue250@unibas.ch>
2013-12-17 17:07:54silvansetrecipients: + silvan, malte, moritz, raznis
2013-12-17 17:07:54silvanlinkissue250 messages
2013-12-17 17:07:54silvancreate