toolset issueshttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues2018-03-07T16:54:27+01:00https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/15Make ToString work properly for Expressions classes and automaton components2018-03-07T16:54:27+01:00Gereon Foxgfox@mpi-inf.mpg.deMake ToString work properly for Expressions classes and automaton componentsIf you hover over a C# expression in the graphical debugger, it uses object.ToString() to give an informative summary of the result of evaluating that expression.
However, for C# objects that represent Modest expressions, all ToString do...If you hover over a C# expression in the graphical debugger, it uses object.ToString() to give an informative summary of the result of evaluating that expression.
However, for C# objects that represent Modest expressions, all ToString does is return the root operator symbol. For example, the object tree representing the expression "(x + 7) * a[i]" is summarized to the very uninformative string "*".
We already have formatting for Expressions, so it should not be that big of a deal to make ToString return proper summaries.Sanny SchmittSanny Schmitthttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues/16Make automata modifiable2018-02-28T15:28:42+01:00Michaela Klauckklauck@depend.uni-saarland.deMake automata modifiableAt the moment you can only create an automaton from which you know already everything, i.e. all locations, edges, ... It would be nice to be able to construct an automaton on the fly by adding locations and edges one after another.
Even ...At the moment you can only create an automaton from which you know already everything, i.e. all locations, edges, ... It would be nice to be able to construct an automaton on the fly by adding locations and edges one after another.
Even better: Be able to modify the automaton as you like, i.e. allow to delete parts of it and check its consistency directly.Sanny SchmittSanny Schmitthttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues/20Replace reflective parameter objects by JANI classes2018-01-19T00:06:09+01:00Arnd HartmannsReplace reflective parameter objects by JANI classesParameter configurations to analysis engines, conversions etc. currently implement IParameterObject, exposing all the actual parameters as properties and the parameter metadata as types and attributes in the code. We will replace this by...Parameter configurations to analysis engines, conversions etc. currently implement IParameterObject, exposing all the actual parameters as properties and the parameter metadata as types and attributes in the code. We will replace this by using the `ParameterDefinition` and `ParameterValue` classes that follow the jani-interaction specification.Sanny SchmittSanny Schmitthttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues/22Use properties instead of fields for the state struct in compiled exploration2018-03-02T15:58:55+01:00Arnd HartmannsUse properties instead of fields for the state struct in compiled explorationCurrently, the locations and state variables of a model are compiled as fields into the state struct type in compiled exploration (in the Exploration project). This is inflexible since the way that values are stored cannot be changed ind...Currently, the locations and state variables of a model are compiled as fields into the state struct type in compiled exploration (in the Exploration project). This is inflexible since the way that values are stored cannot be changed independently of how the values are accessed. We should instead implement properties for the fields and only access (get and set) them via these properties. As a second step, this will allow more advanced storage, e.g. bit-packing multiple state variables into a single (short, int, long etc.) field of the struct.https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/27Properties should not be part of an automata network2017-12-07T12:04:40+01:00Gereon Foxgfox@mpi-inf.mpg.deProperties should not be part of an automata networkThis is a rather philosophical issue, and it is a humble test baloon. On a meta-level, I am posting it to find out how people react to it and whether we care about these kinds of questions at all:
Today I noticed again that **in order t...This is a rather philosophical issue, and it is a humble test baloon. On a meta-level, I am posting it to find out how people react to it and whether we care about these kinds of questions at all:
Today I noticed again that **in order to construct an automata network, you have to give properties**, i.e. the AutomataNetwork class forces you to pass it a set of properties when an automata network is to be created. This in itself is IMHO already bad design because in the mental concept of a network of automata, properties are just absent: It is perfectly natural to think of an automata network without ever bothering about the concept of property at all.
However, I also have two more technical arguments to support my point:
1. If you construct an automata network, but don't know the properties to check yet, you will later have to create a copy of the network with those properties as an argument to the constructor. I find this counter-intuitive and - at least in theory - wasteful. I would rather expect the model checker to take two completely separate arguments, i.e. the automata network and then a set of properties that are defined over the set of variables one finds in that network.
2. The fact that the type `Property` occurs in the definition of the type `AutomataNetwork` creates a tight coupling between those types. I think that in general one rather strives towards *loose* coupling because reducing dependencies among different parts of the code makes it easier to maintain and easier to reuse for a new purpose (which people like Michaela, Gregory, or myself are doing as their primary task).
I can imagine that despite the above counter-arguments, this design might have been chosen because it allows an easier interface to the model checker (which thus has to take only one single argument, and not several), but because of those counter-arguments, I would like to overhaul it.
In case the feedback for this suggestion of mine is positive, I intend to do two things:
1. Find out whether resolving this issue is indeed worth the gain and then have Sanny resolve it.
2. Keep an eye open for and collect more issues of this kind. I have a feeling that they occur in a number of places, but so far I never took care to remember them. However I just realized that often when Gregory or Michaela ask me for help and also when I myself am struggling, problems are at least partly connected to tight coupling (may it be actual or perceived). I therefor consider tight coupling a technical debt that you're sometimes willing to make, but that we're paying quite some interest on. So let's pay back some of it.https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/28Disabling console output, especially the progress bar2017-12-08T23:29:54+01:00Michaela Klauckklauck@depend.uni-saarland.deDisabling console output, especially the progress barAs far as the Modest people in Saarbrücken know, there is no way to disable the console output, especially the progress bar. For some purposes, like e.g. for debugging, but also for other applications where no statistics etc. are used, i...As far as the Modest people in Saarbrücken know, there is no way to disable the console output, especially the progress bar. For some purposes, like e.g. for debugging, but also for other applications where no statistics etc. are used, it would be great to have a flag or a parameter to enable/disable the output. More concrete this is sometimes necessary, because the progress bar can be interleaved by other outputs, which then are partially overwritten. In addition the progress bar causes crashes of the execution if the console window is resized/moved to another window during execution on linux.https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/29Remove dependency of motest on Simulation2018-03-01T06:51:19+01:00Arnd HartmannsRemove dependency of motest on SimulationThe motest executable project currently depends on the Simulation project. This is strange, since simulation and model-based testing would appear to be two rather independent things. The reason appears to be that `MotestParams` derives f...The motest executable project currently depends on the Simulation project. This is strange, since simulation and model-based testing would appear to be two rather independent things. The reason appears to be that `MotestParams` derives from `SimulationAnalysisToolchain.AnalysisParams`. I don't think that should be the case; the parameters for model-based testing are clearly not a superset of the parameters for simulation (does model-based testing do e.g. scheduler sampling or rare event simulation?). I suggest removing the dependency of motest on Simulation.Alexander Graf-BrillAlexander Graf-Brillhttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues/30Removing commit 01e5e3042018-01-03T11:44:48+01:00Arnd HartmannsRemoving commit 01e5e304We decided to edit or, probably better, remove commit 01e5e3045b5ee3e40bb9538a8b9eb82470e584be from master due to the problem with the commit message. This issue is to find out who is affected by that (i.e. who has already pulled that co...We decided to edit or, probably better, remove commit 01e5e3045b5ee3e40bb9538a8b9eb82470e584be from master due to the problem with the commit message. This issue is to find out who is affected by that (i.e. who has already pulled that commit), and set up a strategy to deal with it.
Questions:
1. Who has already pulled in commit 01e5e3045b5ee3e40bb9538a8b9eb82470e584be?
2. Of those, who has _merged_ that commit into their own development branch (and not _rebased_ their development branch on a master including that commit)?https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/31String Representation of States in mcsta Scheduler2018-12-18T18:39:14+01:00Gregory StockString Representation of States in mcsta SchedulerI recently encounter a problem concerning the variable order in the states of a scheduler generated by `mcsta`.
The position of some variables in different schedulers for the "same" model (the structure is identical, only some initial va...I recently encounter a problem concerning the variable order in the states of a scheduler generated by `mcsta`.
The position of some variables in different schedulers for the "same" model (the structure is identical, only some initial variable/array values change) vary.
Context:
I use `modes` to simulate a modest model according to the `mcsta` scheduler in order to receive the optimal trace implicitely calculated by the model checker.
I currently rely on the position of the variables because otherwise I cannot distinguish between the `location` of all instances of the same process given that they all have the same name, e.g. `Experiment.location`.
In the previous model version, only the position of the global clock `gc` deterministically changed between the representation in the scheduler and the internal representation in `modes`[^1] / order in the output trace.
This allowed for a simple fix that deterministically moved the affected entries to their correct place.
This is yet not possible anymore as not even the order *within the schedulers* is predictable.
Example (first entry of the two schedulers):
```
+ State: (AttitudeControl.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Battery.location = 0, Sun.location = 0, Main.location = 0, GlobalInvariant.location = 0, GlobalSync.location = 0, a = 1, ratio = 0, AttitudeControl.a_dst = 0, gc = 0, new_time = 0, AttitudeControl.slewingEnd_time = 0, Battery.old_time = 0, ac_lock = False, slewing = False, insolation = False, Sun.updateAfterInsolationEnd = False, cost = 0, l = 134784000, sun_c = 0, Battery.r = 0, c = [0, 0, 0, 0, 0, 0, 0], nee = [0, 0, 0, 0, 0, 0, 0], pa = [False, False, False, False, False, False, False])
|============================================================================================================================================================================================================================|
+ State: (AttitudeControl.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Experiment.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Exp_Provider.location = 0, Battery.location = 0, Sun.location = 0, Main.location = 0, GlobalInvariant.location = 0, GlobalSync.location = 0, a = 1, ratio = 0, AttitudeControl.a_dst = 0, ac_lock = False, slewing = False, insolation = False, Sun.updateAfterInsolationEnd = False, gc = 0, cost = 0, l = 134784000, new_time = 0, sun_c = 0, AttitudeControl.slewingEnd_time = 0, Battery.r = 0, Battery.old_time = 0, c = [0, 0, 0, 0, 0, 0, 0], nee = [0, 0, 0, 0, 0, 0, 0], pa = [False, False, False, False, False, False, False])
```
How is the variable order in the scheduler states determined? And is it possible to keep it fixed for the same model (only with different variable assignments) or even to ensure the same order within `mcsta`'s schedulers and `modes`' traces?
[^1]: for example the `toString()` of `ref T currentState` in `Simulation/Runs/STASimulationRun.cs`https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/32Location names in ToString() representation of states (T: struct)2018-03-07T15:22:40+01:00Michaela Klauckklauck@depend.uni-saarland.deLocation names in ToString() representation of states (T: struct)When trying to step through a compiled network of states (state struct `T`) by printing the visited states using `ToString()`, I noticed that the variables are named as specified in the input model (e.g. in the `Jani`-file), such that th...When trying to step through a compiled network of states (state struct `T`) by printing the visited states using `ToString()`, I noticed that the variables are named as specified in the input model (e.g. in the `Jani`-file), such that the variable-value-mapping can be easily identified (e.g. `counter = 0` for a variable named `counter` introduced in the `Jani`-model). But the locations of the automata are numbered in the order they are specified in the input file and their names are not used in the state representation, such that they can not be recognized directly. For example a declaration of locations:
``"locations": [
{
"name": "loc10",
"transient-values": []
},
...
{
"name": "loc0",
"transient-values": []
},
``
will in the end result in states, where `automaton.location = 0` in the `ToString()` representation indicates that the location is `loc10`, because `loc10` is the first declaration of a location. It would be much more intuitive if the current location in each automaton would be indicated by the location's name (e.g. `automaton.location = loc0`), like it is done for variables.Sanny SchmittSanny Schmitthttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues/41Move code formatting rules to solution / project files2018-03-07T12:07:19+01:00Felix FreibergerMove code formatting rules to solution / project filesAs discussed in the meeting, we want to align the coding guidelines with Visual Studio defaults and settings stored within the solution or project files.
Steps needed for this:
* [x] investigate how and which settings can be configure...As discussed in the meeting, we want to align the coding guidelines with Visual Studio defaults and settings stored within the solution or project files.
Steps needed for this:
* [x] investigate how and which settings can be configured for the project (@fefrei)
* [ ] make appropriate adjustments to the coding guidelines (@gfox, @ahartmanns)
* [ ] apply the adjusted guidelines to the projectGereon Foxgfox@mpi-inf.mpg.deGereon Foxgfox@mpi-inf.mpg.dehttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues/42Make interval arithmetic handle empty intervals gracefully2018-02-28T07:42:49+01:00Gereon Foxgfox@mpi-inf.mpg.deMake interval arithmetic handle empty intervals gracefullyRight now if you have two intervals `a` and `b` and one of them is empty, the operations `a + b` and `a - b` throw an `ArgumentOutOfRangeException.`. Making these operations simply return an empty interval instead would simplify calling ...Right now if you have two intervals `a` and `b` and one of them is empty, the operations `a + b` and `a - b` throw an `ArgumentOutOfRangeException.`. Making these operations simply return an empty interval instead would simplify calling code and actually fits the semantics of intervals well ("The sum of `a` and `b` is the interval of all values `x + y` for `x` from `a` and `y` from `b`").
I therefor suggest we make these operation return empty intervals in said case. I already tried to check if the exceptions are handled anywhere, which does not seem to be the case.
Any reason not to do it?Gereon Foxgfox@mpi-inf.mpg.deGereon Foxgfox@mpi-inf.mpg.dehttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues/44Support complex initial state specifications2018-03-12T10:12:27+01:00Michaela Klauckklauck@depend.uni-saarland.deSupport complex initial state specificationsWhen restricting the initial state using an arithmetic expression (over global variables `q1, q2, q3` in this case), the error message:
> error: Complex initial states specifications are not yet supported.
is thrown.
It would be nice ...When restricting the initial state using an arithmetic expression (over global variables `q1, q2, q3` in this case), the error message:
> error: Complex initial states specifications are not yet supported.
is thrown.
It would be nice to extend the range of supported expressions because in e.g. PRISM case studies they are used very often.
` "restrict-initial":{
"exp":{
"op":"≥",
"left":{
"op":"+",
"left":{
"op":"+",
"left":"q1",
"right":"q2"
},
"right":"q3"
},
"right":1
}
},`
The currently supported syntax for initial state restrictions is checked in `DissectInitialStatesRestriction()` in `AutomataNetwork.cs`. And the error message is thrown in ` CreateSetInitialStateAndTransients()` in `CompiledAutomaton.cs`https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/45Support more than one initial state2018-03-12T13:38:26+01:00Michaela Klauckklauck@depend.uni-saarland.deSupport more than one initial stateModest does currently not support more than one initial state. This means that initial state restrictions using disjunctions or shrinking variables to a certain range are not allowed. Therefore this issue is in some sense related to #44....Modest does currently not support more than one initial state. This means that initial state restrictions using disjunctions or shrinking variables to a certain range are not allowed. Therefore this issue is in some sense related to #44.
A workaround would be to add a dummy initial state with tau transitions to all other real initial states. But since PRISM supports this feature, all JANI translations from such PRISM models will result in Jani files not directly (without a preprocessing step adding the dummy initial state) supported by Modest.https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/51Use NetworkCompilerHelper instead of compiling manually in motest2018-07-08T16:55:47+02:00Arnd HartmannsUse NetworkCompilerHelper instead of compiling manually in motestmotest currently manages all the steps of compiling an `AutomataNetwork` manually. This prevents better encapsulation of the compilation process in the Exploration project. It should be possible with very limited effort to make motest us...motest currently manages all the steps of compiling an `AutomataNetwork` manually. This prevents better encapsulation of the compilation process in the Exploration project. It should be possible with very limited effort to make motest use `NetworkCompilerHelper.Compile` and related methods instead.Alexander Graf-BrillAlexander Graf-Brillhttps://dgit.cs.uni-saarland.de/modest/toolset/-/issues/57Recursive process calls fails because of too strict checking2019-02-25T14:12:07+01:00Hans van der LaanRecursive process calls fails because of too strict checkingWhen running the code below, we get the following error: figure-3-15-process_decl.modest:(16,28): error: Cannot read local variables or process parameters immediately before a recursive process call.
If I interpreted Arnd's critique cor...When running the code below, we get the following error: figure-3-15-process_decl.modest:(16,28): error: Cannot read local variables or process parameters immediately before a recursive process call.
If I interpreted Arnd's critique correctly, in this case we should be able to make a recursive process call since we don’t change the variable.
```
//---Property which should always hold---
const int check = 1;
property Term = Pmin(<> (check == 1)) == 1;
//---------------------------------------
action snd_data, rcv_ack, timeout;
exception err;
int n = 3;
process Sender(int n)
{
snd_data {= n = n - 1 =};
alt {
:: rcv_ack
:: timeout; alt {
:: when(n > 0) Sender(n)
:: when(n == 0) throw(err)
}
}
}
Sender(2)
```https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/82Case study cashier.modest cannot be parsed by mcsta2019-05-21T14:22:14+02:00Hans van der LaanCase study cashier.modest cannot be parsed by mcstaThe cashier.modest case study from the modes models cannot be parsed by mcsta.
We get the following error when we try to check the model with mcsta:
```
cashier.modest:(40,15): error: Expected "S", "T" or "X".
```The cashier.modest case study from the modes models cannot be parsed by mcsta.
We get the following error when we try to check the model with mcsta:
```
cashier.modest:(40,15): error: Expected "S", "T" or "X".
```https://dgit.cs.uni-saarland.de/modest/toolset/-/issues/95Missing DisplayNames in DotExporter and PythonExporter2021-03-29T10:37:48+02:00Peter SmitMissing DisplayNames in DotExporter and PythonExporterBoth the Dot- and PythonExporter's parameters classes are missing the DisplayName attribute on their properties.Both the Dot- and PythonExporter's parameters classes are missing the DisplayName attribute on their properties.