fault coverage testing, first implementation (EXPERIMENTAL)
authorChristian Thaeter <ct@pipapo.org>
Mon, 12 Jul 2010 21:05:09 +0000 (23:05 +0200)
committerChristian Thaeter <ct@pipapo.org>
Fri, 16 Jul 2010 04:08:07 +0000 (06:08 +0200)
commitaa621b1515c6086278e10e44477bad78bdae51bb
tree9520abfd6d44c307da71fb72b31a7c1a23624c40
parent69c89f65fe62bbbaf8576ad78ef682b4821723a3
fault coverage testing, first implementation (EXPERIMENTAL)

This creates a facility to permute over instrumented faults. This is done
by hashing backtraces, every path reaching a failure point is recorded.

Programs using this fault coverage checking first parsing all logfiles
provided in the env var NOBUG_COVERAGE to learn about the previous state.

Then running the program will log at every instrumented fault and initially
fail at any newly discovered failure point and pass at the last known
failure point. Running a program multiple times always feeding back the
logs from the previous run will eventually permute through all possible way
calling errors and handling them.
Makefile.am
src/nobug.c
src/nobug.h
src/nobug_coverage.c [new file with mode: 0644]
tests/Makefile.am
tests/coverage.sh [new file with mode: 0755]
tests/test_nobug_coverage.c [new file with mode: 0644]