Merge branch 'v0.16.x_bugfix' into v0.16.x

de.gebit.integrity.runner/src/de/gebit/integrity/runner/DefaultTestRunner.java

@@ -1321,6 +1321,20 @@ protected SuiteSummaryResult callSuiteSingle(Suite aSuiteCall) {
 						// If this happens, an abortion has happened on the fork due to an AbortExecutionException.
 						// TODO make this nicer, it's kind of ugly to create a fake object with null values
 						abortExecutionCause = new AbortExecutionCauseWrapper(null, null);
+
+						if (tempResult == null && tempForkResultSummary == null) {
+							// We may not have any result at this point, as the fork has aborted without providing us
+							// one. In order to ensure that at least the exception that triggered the abortion is logged
+							// in the counts (and thus a user who typically only looks at the total counts is alerted to
+							// the problem), we generate a result here with one exception. This might be wrong actually
+							// (for example there could have been a successful test before the exception, which may not
+							// be counted in this case), but that's basically what's meant by "test result total numbers
+							// may be inaccurate" which is printed out in this case, and the current structure of the
+							// master-fork sync protocol makes it hard to perfectly fix this inaccuracy in cases of
+							// sudden execution path deviations. Thus, this "forced result" was deemed a good-enough
+							// solution. This fixes issue #145: https://github.com/integrity-tf/integrity/issues/145
+							tempResult = new SuiteSummaryResult(0, 0, 1, 0, tempSuiteDuration);
+						}
 					}
 
 					// and afterwards we'll switch back to real test mode