Menu
▾
▴
#1008 Fuzzing: Pseudo-randomized regression testing
We currently build SDCC nightly on the machines in the compile farm. Then we do regression testing there, either the full list or a third per night. The list contains essentially the targets in support/regression/ports, i.e. each stable port, and for some ports a few selected command-line options. This means that most combinations of command-line options never get tested in the farm.
I'd like to see a mechanism for additional testing, where a machine builds the regression tests which some pseudo-randomly selected command-line option combinations.
