Bug propagation and debugging in asymmetric software structures

We address the issue of how software components are affected by the failure of one of them, and the inverse problem of locating the faulty component. Because of the functional form of the incoming link distribution of software dependence network, software is fragile with respect to the failure of a random single component. Locating a faulty component is easy if the failure only affects its nearest neighbors, while it is hard if it propagates further.

Figure 1

Cumulative distribution of incoming links (circles) and outgoing links (squares) between packages in a computer running RedHat 8.0. Empty and full symbols are obtained with rpm -q - -whatrequires and rpm -q - -requires, respectively. The continuous line has a $-1$ slope.

Figure 2

Number of outgoing links vs number of incoming links for the packages of REDHAT 8.0.

Figure 3

Cumulative distribution of incoming links in LINUX (circles), MOZILLA (squares), APACHE (diamonds), and MYSQL (triangles). The dashed line is $1∕q$.

Figure 4

Cumulative distribution of outgoing links in LINUX (circles), MOZILLA (squares), APACHE (diamonds), and MYSQL (triangles).

Figure 5

Basin size distribution of failure propagation on function call graphs. Continuous lines are for bug influence basins and dashed lines are for debugging basins.

Figure 6

Basin size distribution of failure propagation on package-package dependency graph. The continuous line is for bug influence basins and dashed line is for debugging basins.