Ernst, M. D., W. G. Griswold, Y. Kataoka and D. Notkin, Dynamically discovering pointer-based program invariants, Technical Report UW-CSE-9911 -02, University of Washington, Seattle, WA (1999).  Home/Search   Document Details and Download   Summary   Related Articles   Check

.... result from splitting data into parts based on some condition and comparing the resulting invariants over each part; if mutually exclusive invariants appear in each part, they may be used as predicates in implications, and unconditional invariants in each part are composed into implications [EGKN99] In this research, we ignore those invariants that are inexpressible in ESC Java s input language; for example, many of the sequence invariants are ignored. For each variable or tuple of variables in scope at a given program point, each potential invariant is tested. Each potential unary ....

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointer-based program invariants. Technical Report UW-CSE-99-11-02, University of Washington, Seattle, WA, November 16, 1999. Revised March 17, 2000.

....points in the program, such as a method precondition or postcondition, an object invariant, or the condition of an assert statement. Our experiments use the Daikon invariant detector, whose output includes conditionals, pre and poststate variables, and properties over a variety of data structures [EGKN99]. Thus, the output is often a high quality specification that matches human written formal specifications or can be proved correct [NE01b, NE01a] Even lesser quality output forms a partial specification that is nonetheless useful [KEGN01] The specifications are unsound: the properties are ....

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointer-based program invariants. Technical Report UW-CSE-99-1102, University of Washington, Seattle, WA, November 16, 1999. Revised March 17, 2000.

....Tools Daikon [2] extracts likely algebraic invariants from information gathered during the program s execution. For example, Daikon can infer invariants such as y = 2x . Daikon handles heap structures in a limited fashion by linearizing them into arrays under some specific conditions [3]. Our work di#ers in that we handle heap structures in a much more general fashion and focus on referencing relationships as opposed to algebraic invariants. Womble [8] and Chava [10] both use a static analysis to automatically extract object models for Java programs. Both tools use information ....

M. D. Ernst, Y. Kataoka, W. G. Griswold, and D. Notkin. Dynamically discovering pointer-based program invariants. Technical Report UW-CSE-99-11-02, University of Washington, November 1999.

No context found.

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointerbased program invariants. Technical Report UW-CSE-99-11-02, University of Washington, Seattle, WA, November 16, 1999.

No context found.

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointer-based program invariants. Technical Report UW-CSE-99-11-02, University of Washington, Seattle, WA, November 16, 1999.

No context found.

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointer-based program invariants. Technical Report UWCSE -99-11-02, University of Washington, Seattle, WA, November 16, 1999.

No context found.

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointerbased program invariants. Technical Report UW-CSE-99-11-02, University of Washington, Seattle, WA, November 16, 1999.

....such as if p 6= null then p:value x and p:value limit or p:left 2 mytree . Conditional invariants result from splitting data into parts based on the condition and comparing the resulting invariants; if the invariants in the two halves differ, they are composed into a conditional invariant [EGKN99] For each variable or tuple of variables in scope at a given program point, each potential invariant is tested. Each potential unary invariant is checked for all variables, each potential binary invariant is checked over all pairs of variables, and so forth. A potential invariant is checked by ....

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointerbased program invariants. Technical Report UW-CSE-99-11-02, University of Washington, Seattle, WA, November 16, 1999.

....Pointerbased invariants are obtained by linearizing graph like data structures. Conditional invariants result from splitting data into parts based on the condition and comparing the resulting invariants; if the invariants in the two halves differ, they are composed into a conditional invariant [EGKN99] For each variable or tuple of variables, each potential invariant is tested. Each potential unary invariant is checked for all variables, each potential binary invariant is checked over all pairs of variables, and so forth. A potential invariant is checked by examining each sample (i.e. tuple ....

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointerbased program invariants. Technical Report UW-CSE-99-11-02, University of Washington, Seattle, WA, November 16, 1999.

....field value of type int. The result is a sortedness invariant over a realistic and useful (albeit somewhat complicated to describe) collection. In other data structures, Daikon found similar invariants (such as sortedness of a tree or membership in a collection) details are provided elsewhere [EGKN99]. 6 REDUNDANT INVARIANTS Invariants that are logically implied by other invariants need not be computed or reported to the user. Eliminating implied invariants greatly reduces the time and space costs of invariant inference; in practice, without this improvement Daikon often fails to compute ....

....value from being counted multiple times when testing statistical confidence prevent chance properties from being reported as justified invariants. Elsewhere we report on how to discover invariants over recursive pointer based structures and conditional invariants that are not universally true [EGKN99]; an example is p = NULL or p:left 2 mytree. We are also exploring a richer user interface for Daikon, which would support some graphical displays (when appropriate) and give the programmer more control over both instrumentation and display of invariants. We also intend to perform more and larger ....

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointerbased program invariants. Technical Report UW-CSE-99-11-02, University of Washington, Seattle, WA, November 16, 1999.

....splitting of the traces; exceptions to invariants being tested; splitting on special values, such as common constants (like zero and one) or extremal values found earlier; and using static analysis to identify potential predicates for splitting. Preliminary results are reported elsewhere [EGKN99] 11 Conclusions This research demonstrates the feasibility and effectiveness of discovering program invariants based on execution traces. This technique automatically detected all the stated invariants in a set of formally specified programs; furthermore, the invariants detected in a C program ....

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointer-based program invariants. Technical Report UW-CSE-99-11-02, University of Washington, Seattle, WA, November 16, 1999. 33

....be of runtime type MyObject, which has field value of type int. The list of those slots of the elements of the reachable ListNodes is non decreasing. In other data structures, we found similar invariants (such as sortedness of a tree or membership in a collection) details are provided elsewhere [EGKN99]. 7 REPEATED VALUES Daikon reports only invariants that pass a statistical confidence test; properties that could easily have occurred by chance are not reported, as they are likely to be accidents of the data. For instance, given 0 x; y 10, if there are only three (x; y) pairs, then the ....

.... Our most significant progress has been in the area of pointer based data structures, where we can now extract invariants such as p = NULL or p:left 2 mytree; elsewhere we report on how we discover this class of disjunctive properties over pointers and fields in pointer based data structures [EGKN99]. We are also taking initial steps towards a richer user interface for Daikon, which would support some graphical displays (when appropriate) and would allow the programmer far more control over what variables and program points are monitored, as well as much more direct control over which ....

Michael D. Ernst, William G. Griswold, Yoshio Kataoka, and David Notkin. Dynamically discovering pointer-based program invariants. Technical Report UWCSE -99-11-02, University of Washington, Seattle, WA, November 16, 1999.

No context found.

Ernst, M. D., W. G. Griswold, Y. Kataoka and D. Notkin, Dynamically discovering pointer-based program invariants, Technical Report UW-CSE-9911 -02, University of Washington, Seattle, WA (1999).

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC