Running Fuzzers

Run only fuzzers:

bash

ast fuzz
ast fuzz parser
ast fuzz parser --runs 10000 --seed 42
ast fuzz parser --fuzzer parse-error
ast fuzz parser --runs 1.5x
ast fuzz parser --runs +10%

Run tests and fuzzers together:

bash

ast test --fuzz
ast test --fuzz --fuzz-runs 10000 --fuzz-seed 42
ast test --fuzz parser --fuzzers parse-error,serialize-roundtrip
ast test --fuzz --fuzz-runs +100000

Useful Flags

--runs <value>
--seed <n> (pins the seed; default uses a random seed each campaign)
--fuzzer <name[,name...]>
--fuzzers <name[,name...]>
--suite <name[,name...]> (fuzz alias)
--suites <name[,name...]> (fuzz alias)
--jobs <n>
--build-jobs <n>
--run-jobs <n>
--mode <name[,name...]>

<value> can be:

an absolute run count like 500
a multiplier like 1.5x
an additive percentage like +10%
an additive integer like +100000

Notes

fuzz targets currently build as bindings
xfuzz(...) registers a skipped fuzz target
--fuzzer and --fuzzers filter by normalized fuzz("name", ...) target names
CLI run overrides apply to each fuzzer's effective base count, including fuzzers that set operations via either fuzz(..., operations) or .generate(..., operations)
fuzz results are summarized separately from normal test totals