US 2008/0313602 A1Published application

Bounded program failure analysis and correction

AssigneeMicrosoft Corporation

InventorsNikolai Tillmann, Jonathan Paul de Halleux, Wolfram Schulte

Life of this patent

Estimating the term…

Loading patent…

Abstract

In one embodiment, a computer system determines that a previously run test scenario configured to test a software program has failed to produce an expected result due to one or more semantic errors, generates error trace code configured to monitor the called component, processes the test scenario using the error trace code, and analyzes error trace information to determine the point at which the semantic error occurs in the called component. In an alternative embodiment, a computer system detects a semantic error in a software component of a software program, constructs an error condition that may include source code representing a minimum condition under which the error occurs, generates an object invariant based on the error condition that represents an opposite condition to that represented by the error condition, and automatically generates source code change recommendations using the object invariant that prevent the semantic error from reoccurring in subsequent test scenarios.

Claims20

Dependent claims

At a computer system, a method for automatically determining the point at which a semantic error occurs in a software component of a software program using an error trace corresponding to a single component of the software program, the method comprising:

an act of determining that a previously run test scenario configured to test a software program has failed to produce an expected result due to one or more semantic errors, the software program including at least one calling component and at least one called component, the calling component configured to call the called component, the call including one or more values to be used by the called component as inputs;
based on the determination, an act of generating error trace code configured to monitor the called component, the monitoring including identifying input values received in the call from the calling component and return values generated by the called component, the error trace code comprising a plurality of checks configured to detect unexpected result values for each software component;
an act of processing the test scenario using the error trace code, the error trace code detecting unexpected result values at the called component, the unexpected result values indicating a violation of specified allowable behavior for the called component;
and
an act of analyzing error trace information provided by each of the plurality of checks of the error trace to determine the point at which the semantic error occurs in the called component.

2 more elements

Dependencies are inferred from the claim text.