From a53dc7f70f67e4b0bbc75c8f68e0570267f036ec Mon Sep 17 00:00:00 2001 From: Alex Groce Date: Wed, 23 May 2018 14:25:19 -0700 Subject: [PATCH] more info on using analyzeswarm --- README.md | 6 +++++- 1 file changed, 5 insertions(+), 1 deletion(-) diff --git a/README.md b/README.md index 92eda03..1bf38e0 100644 --- a/README.md +++ b/README.md @@ -530,7 +530,11 @@ A more systematic way to go about directed swarm testing is to try: to generate triggers and suppressors for ALL coverage targets hit during a run, grouped into equivalence classes (targets with the same set of triggers and suppressors) and ranked by the least-hit target in -each equivalence class. Just iterating through the generated +each equivalence class. The output will be stored in files beginning +with ``: a set of files named `.N.probs` that can +used with `--swarmProbs`, and a single `.class` file, with all the targets, +triggers, suppressors, and minimum frequencies for classes, plus +pointers to the related probability files. Just iterating through the generated probability files for the classes for the rarest targets is a good way to go about directed swarm testing. The `0.5` above can be any cutoff, above which targets hit by at least that fraction of tests are