We wrap the RegionValues returned by ParamModelChecker in a ParamResult
object, which stores additional meta-information that allow for the
result to be tested.
In test mode, if there is only a single region, tests against the given
expected RESULT expression.
git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@11639 bbc10eb1-c90d-0410-af57-cb519fbb1720
constValues.getValueOf() throws an Exception if the constant value
does not exist. This may happen, e.g., for parametric model checking,
where the parametric constants will have no values.
So, we first check if the constant value exists. If this is not the
case then the RESULT line is clearly no match.
In the future, it might make sense to handle parametric constants
differently than non-existent constants, e.g., to provide a warning
for a nonsensical RESULT line in the latter case.
git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@11601 bbc10eb1-c90d-0410-af57-cb519fbb1720
Instead of before, where only
// RESULT: integer
or
// RESULT: numerator/denominator
were allowed for result expressions, we now support arbitrary expressions
over the model/property parameters, e.g.,
// RESULT: 4.3*floor(q*3)
Caveat: Currently, there can be no whitespace in the expression...
Additionally, when model checking in exact mode, the expected value is
also computed using Expression.evaluateExact.
git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@11557 bbc10eb1-c90d-0410-af57-cb519fbb1720
Now, the *first* RESULT that matches the model/property constants
is returned, previously it was the *last*. Inverting the order seems
to better match user expectations.
git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@11548 bbc10eb1-c90d-0410-af57-cb519fbb1720
Add handling for ExpressionITE to Simplify.
Incidentally, this fixes a bug in PTA handling, where guards with an
ITE operation where not properly simplified and thus unsupported, e.g.,
s=0 ? true : false
as guard.
git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@11214 bbc10eb1-c90d-0410-af57-cb519fbb1720
If the expected result specified via // RESULT
is an error, we treat a PrismNotSupportedException just
as any other ordinary exception. This allows verifying
the content of a PrismNotSupportedException error message.
Otherwise, we do the usual special handling as before.
git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@11173 bbc10eb1-c90d-0410-af57-cb519fbb1720
Expressions such as
P=?[ X F P>0[F<=4 s=7] ]
were problematic before, as the LTL check for time bounds would recurse into the P-subformula
and complain without need.
git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@11036 bbc10eb1-c90d-0410-af57-cb519fbb1720
In SVN 10565, custom equals/hashCode methods were introduced for the
various ASTElements/Expressions, providing semantic equality. In the
implementation of ConvertForJltl2ba, a Hashtable was used to store
SimpleLTL formulas for the already handled Expression objects. With
the new equals/hashCode, this can lead to SimpleLTL formulas that
share subtrees, resulting in a DAG instead of a tree.
The SimpleLTL.simplify() function behaves incorrectly for DAGs, e.g.,
for "! (X (s1=7)) | (X (X s1=7))", producing wrong output formulas.
We refactor ConvertForJltl2ba to do a simple recursive transformation
from Expression-based LTL to SimpleLTL. Optionally, sharing of
identical subtrees can be enabled.
git-svn-id: https://www.prismmodelchecker.org/svn/prism/prism/trunk@10894 bbc10eb1-c90d-0410-af57-cb519fbb1720
Joachim Klein