Manual Reference Pages  - selt (n)

NAME

selt - SE-LTL model checker

Part of Tina Toolbox for analysis of Petri nets and Time Petri nets.

CONTENTS

Synopsis
Description
Options
Interacting With The Nd Stepper
Selt Se-ltl Language And Commands
Examples
See Also
Authors

SYNOPSIS

selt [-help]

selt ktzfile [-f formula | formfile] [-prelude ltlfile] [-ltl2ba | -spot]
[-q | -v] [-b | -c | -p | -s | -g] [-S scnfile] [-wp n]
[outfile] [errorfile]

DESCRIPTION

selt model-checks state-event LTL formulas on a kripke transition system given in ktz format.

If some formula is specified (by flag -f or by providing formfile), then the result of evaluation of the formula is printed according to the output mode and verbosity flags, and selt exits.

If no formula file is specified, then selt starts an interactive session, evaluating commands entered by the user on standard input (see "selt se-ltl language and commands" below).

OPTIONS

-help Recalls options.
 

Ktz input:
 

ktzfile
  The kripke transition system on which formula are model-checked, in ktz format (extension .ktz).

-wp (0|1|2)
  This flag removes (0), preserves (1), or forces at each state (2) the temporal divergence property possibly present in the kts file (see tina’s -wp flag for details).

LTL input:
 

-f formula
  Passes to selt the formula to be verified.

formfile
  Specifies a file holding the formulas to be verified. Must have extension .ltl

-prelude ltlfile
  Specifies a file containing selt commands to be evaluated on entry, before any formula provided by -f, by formfile, or interactively. This flag is useful to load SE-LTL libraries. ltlfile must have extension .ltl

-ltl2ba Convert ltl formula to buchi automata using ltl2ba (default). ltl2ba binaries are bundled with tina distributions.

-spot Convert ltl formula to buchi automata using ltl2tgba from the SPOT model checking library. ltl2tgba must be installed at your site, from the sources available at (http://spot.lip6.fr).

Verbosity level:
 

-v prints banner and evaluation times for LTL commands (default).

-q No banner nor times are printed. This flag is useful for batch operation.

Output mode flags:
 

-b When evaluating a formula, just prints its truth value, without counter example.

-c When a formula evaluates to FALSE, prints in addition a counter example in condensed form (default).

-p When a formula evaluates to FALSE, prints in addition a counter example in full form.

-s When a formula evaluates to FALSE, prints in addition a counter example in the scn formmat of the nd stepper.

-g When a formula evaluates to FALSE, does not print a counter example, but builds the full graph obtained as the synchronization of the kts and of the bucchi automaton translating the negation of the formula, instead, in ktz form (to be provided).

Other flags:
 

-S scnfile
  When a formula evaluates to FALSE, writes a counter example in .scn format of in file scnfile (creating the file if it does not exist, and overwriting it otherwise). This flag is useful to replay counter examples in the nd stepper when modelchecking an existing ktz description and a description of the net the .ktz file describes the behavior of is available (see "interacting with the nd stepper" below).

Output destination:
 

outfile
  Where results are written. The output format is determined by the file type, according to the table below. If absent or specified by "-", then results are written on standard output in the format specified by the output flag. If both an outfile and some output flag are present, then the format defined by the output flag supersedes that determined by the outfile extension.

file extension          output format
--------------------------------------------------------------
.scn                    nd stepper .scn format
.ktz                    .ktz format
any other               textual output

Errors destination:
 

errorfile
  Where error messages are written. By default, errors are printed on standard error.

INTERACTING WITH THE ND STEPPER

For model-checking Petri nets or Time Petri nets, tool selt is interfaced with the nd editor of the Tina toolbox.

When the .ktz of the behavior of some net is not available, or rebuilding it is cheap, SE-LTL properties of the net can be model-checked as follows: load the net into nd, then build the ktz description of its behavior (using the reachability tool, with the default output mode and construction), then invoke the modelchecker from the menu in the ktz window, and finally select "stepper" in the nd tools menu. Any counter example computed by the model-checker will then be loaded automatically into the stepper, for replay.

When the .ktz of the behavior of some net is available and rebuilding it is expensive, then invoke selt with option -S. The file passed with -S will always hold the last counter example built. In parallel, load the net description into nd and select "stepper" in the nd tools menu. For replaying a counter example, load the .scn file specified with -S.

SELT SE-LTL LANGUAGE AND COMMANDS

Selt implements a version of the SE-LTL linear time temporal logic as described in paper:

S. Chaki, E. M. Clarke, J. Ouaknine, N. Sharygina, and N. Sinha, "State/Event-based Software Model Checking", 4th International Conference on Integrated Formal Methods (IFM’04), Springer LNCS 2999, pages 128--147, April 2004.

For building Buchi automata from LTL files, selt relies on ltl2ba, a tool described in the paper below and available at http://www.liafa.jussieu.fr/~oddoux/ltl2ba

P. Gastin and D. Oddoux, "Fast LTL to Buchi Automata Translation", 13th Conference Computer-Aided Verification (CAV’2001), Springer LNCS 2102, pages 53--65, July 2001.

1. Lexical matters:

- A identifier is either:

Any place or transition identifier allowed in .net or .ndr descriptions, that is: any sequence of letters, digits, underscores "_" and primes "’", or any sequence of characters enclosed in braces in which "{", "}" (except the outer ones) and "\" are prefixed by "\"

Any sequence of symbols from the list ~ ‘ ! @ # $ % ^ & * - + = : ? | / < > [ ];

A qualified identifier: an identifier prefixed by S., E. or L.

e.g. hello, _p4’_, 123, >=<, or {variable x45}, are legal identifiers.

- The commands are: op, infix, prefix, forget, verb, output, source, quit, assert. Command names may not be used as operator or variable names.

- When analyzing identifiers, the scanner advances as right as possible. So, in a juxtaposition of identifiers, two symbolic or two alphanumeric unbraced identifiers, or e.g. an alphanumeric identifier and a command name, must be separated by a space. But no space is necessary between identifiers of different kinds or between a parenthesis (or ";") and an identifier.

- justaposition bind tighter than infixes and associate to the left, infixes and prefixes associate to the right. That is, if f is a 3-ary operator in functional notation, [], <>, -, are prefixes, and /\, \/, are infixes:

                        [] p1 => <> - p3 /\ f u v w
  parses as     ([] p1) => ((<> (- p3)) /\ (f u v wp))
                        
                        f - <> p1 (f p0 p1 (p4 /\ p5)) \/ f u v w
  parses as     (f (- <> p1) (f p0 p1 (p4 /\ p5))) \/ (f u v w)

and           f - <> p1 f p0 p1 p4 \/ f u v w parses as     (f (- <> p1) f p0 p1 p4) \/ (f u v w) which is ill-typed

- infixes have precedence in 0..5 (see below). Infixes with higher precedence bind tighter than those with lower precedence.

2. The initial environment:

It is made up of (pushed in that order):

- The atomic state and event propositions. They have the names captured in the .ktz file, i.e. those of the places and transitions of the Petri net if the .ktz file was generated by tina.

- Then, the logic and arithmetic primitives, constituted of:

   constants:   T (true), F (false),
                dead (deadlock property)
                div (temporal divergence property)
                sub (state is partially known)

prefixes:    [] (always), <> (eventually), () (next),              - (logic negation), ~(arithmetic negation)

infixes:     U (until), V (release), of precedence 0              => (implies), <=> (equivalent), of precedence 1                 /\ (and), \/ (or), of precedence 2                 <=, >=, =, le, lt, ge, gt, of precedence 3                 +, of precedence 4                 *, of precedence 5


                 - Then the user defined operators.

Since the syntactic classes of atomic state propositions, atomic event propositions, logic primitives, integer, and user defined operators, overlap, we must have some way of disambiguating identifiers. For this:

- unqualified identifiers are bound to the command with than name, if any, or otherwise to the last pushed environment entry with that name.

- identifiers qualified by S (e.g. S.p1) are bound to the atomic state proposition with that name with the qualifier removed (e.g. p1);

- identifiers qualified by E (e.g. E.t1) are bound to the atomic event proposition with that name with the qualifier removed (e.g. t1);

- identifiers qualified by L (e.g. L./\) are bound to the logic primitive with that name with the qualifier removed (e.g. /\);

So, atomic propositions (found in the ktz file) sharing their name with some atomic propositions in a different group (state or event) or with some command (e.g. op), or some logical primitive (e.g. -), or the name of which is an integer (e.g. 3), must be referred to in formulas by their qualified form (e.g. S.op, E.-, or S.3);

3. Fixity:

Identifiers declared infix (binary logic primitives or user defined operators declared by "infix") must be used in infix notation;

Identifiers declared prefix (unary logic primitives or user defined operators declared by "prefix") must be used in prefix notation (in a juxtaposition of identifiers, prefix operators associate with the right expression);

Other operators or primitives accessed by their qualified names must be used in functional notation. E.g. if a1, a2, a3 are parenthesized expressions, and f has arity 3, then:

        a1 /\ a2 and L./\ a1 a2 are legal (and equivalent)
    but a1 L./\ a2 is ill-typed

        f - a1 a2 a3 is legal (parses as ((f (- a1)) a2) a3) but f L.- a1 a2 a3 T is ill-typed (parses as (((f L.-) a1) a2) a3)

4. Commands and effects:

Commands must terminate with ";". In formula files, the final ";" may be omitted (EOF plays that role). The effects of commands are as follows ("exp" is any ltl expression, x, y, f, xi are identifiers):

exp;
        evaluates LTL expression exp;

assert exp "whentrue" "whenfalse"; evaluates exp then prints string whentrue (resp. whenfalse) if exp holds (resp. does not) instead of the default message TRUE (resp. FALSE).

op f x1 ... xn = exp;         declares an operator f or arity n (n >= 0), to be used in         functional notation;

infix [n] x f y = exp;         declares a binary operator f in infix notation. n is an         optional integer in 0..5 specifying precedence.

prefix f x = exp;         declares a unary operator f in prefix notation;

forget f1 ... fn;         Removed items names f1 ... fn from the environment, and their         fixity information;

source [file | "file"];         reads at toplevel the contents of file. The file name is         optionally surrounded by string quotes (this is necessary if         the name includes spaces);

verb [true | false | debug];         verbosity level. Tne initial setting follows from the command         line flags -v | -q (default -v);

        true (default): prints the banner, prompts, results of         commands, and evaluation times;

        false: just prints the results of evaluation of LTL         expressions (useful in batch mode);

        debug: may print extra information (mainly for developper);

output [proof | fullproof | quiet | stepper | graph]         specify effects and results of evaluations of LTL         expressions. The command line flags -c, -p, -b, -s, -g         specifies the initial setting (default -c);

        proof (default): evaluations return TRUE, or FALSE with a         counter example in condensed form;

        fullproof: evaluations return TRUE, or FALSE with a counter         example in full form (may be very large);

        quiet: evaluations just return TRUE or FALSE, without counter         examples.

        stepper: as output fullproof except the counter example is in         stepper format (typically used in batch mode);

        graph: not implemented yet

quit             leaves selt (also ^D on Unix targets).

5. Example commands:

p1;                     (eqv. p1 lt 0 \/ p gt 0)
- p1;                   (eqv. p1 = 0)
p1 /\ p2 >= 2;          (eqv. p1 /\ (p2 >= 2))
op x > y = x >= y + 1;  (declares arithmetic operator >, in infix form)
p1+p2;                  (eqv. p1+p2 >= 0, eqv. p1 \/ p2)
p1+p2 > p3;
p1*p2 = 0;
infix y follows x = [] (x => <>y); 
                (declares logical operator "follows:, in infix form)
(t3 \/ p5) follows (t1 /\ p1>=p2);
                (eqv. [] ((t1 /\ (p1>=p2)) => <> (t3 \/ p5)));

EXAMPLES

selt abp.ktz -f "[] (t1 => <> t3)" -q -b
selt abp.ktz -p

SEE ALSO

nd(n), tina(n), sift(n), tedd(n), plan(n), struct(n), ktzio(n), ndrio(n), muse(n), pathto(n), play(n), formats(n)

AUTHORS

Bernard Berthomieu, with contributions by Francois Vernadat, Pierre-Olivier Ribet, LAAS/CNRS, 2000-2012, {Bernard.Berthomieu|Francois.Vernadat}@laas.fr.


Tina Tools selt (n) Version 3.5.0
Generated by manServer 1.07 from src/selt.n using man macros.