| Summary: | Using constant arrrays as parameter to templetes gives faulty behavior | ||
|---|---|---|---|
| Product: | UPPAAL | Reporter: | Jacob Illum <illum> |
| Component: | Engine | Assignee: | Gerd Behrmann <behrmann> |
| Status: | CLOSED FIXED | ||
| Severity: | normal | ||
| Priority: | P2 | ||
| Version: | 3.4.6 | ||
| Hardware: | PC | ||
| OS: | All | ||
| Architecture: | |||
| Attachments: |
The system that causes problems
The system that causes problems |
||
Created attachment 23 [details]
The system that causes problems
Created attachment 24 [details]
The system that causes problems
A fix for 3.4 has been checked into CVS. I still need to check the development branch. I checked the development branch. The bug is not present there, but the test case triggered another problem. A fix for that problem has also been checked into CVS now. |
When a constant array is given a an input parameter to a template, then testing the value of this array is unreliable, i.e. a boolean array can be matched against neither true nor false. Furthermore, if a variable is added to the system the behavior changes, even if the variable is not used in any component. See the attached system for detail: Here P takes a constant array as an input parameter. A process is created with P({true,false}). The initial state of P is A, here there is a single transition to the state B. In B we have two transition back to A depending on whether the value of the first entry of the array is true or false. However, the simulator claims the system is deadlocked. When adding a global variable i and not using it in P, the behavior changes. Now, a transition is available in B, however, it is the wrong one.