You probably don't want to use this. I spent a few minutes writing the most naive possible fuzzer, to see if it would turn up any bugs. Turns out, this terrible method can generate bugs faster than I can debug them. However, the general approach and the specific implementation are pretty lame. If you're looking at this, you probably want something like Csmith or jsfunfuzz.
I'm planning to, eventually, write a real fuzzer to see if something that generates random expressions turns up more bugs than this. Suggestions / pull requests welcome.
8353: parser loses track of location when parse is called
8352: underspecified type for writemime
8341: EOF mixed up in unicode hangs parser
8329: changing number of BLAS threads before operation causes segfault
8286: exponent of large number causes crashes with SIGABRT
8277: code near exception incorrectly optimized out
8266: hang/bad result on ndigits
8241: hang/segfault on long symbol
8239: segfault on code_native/code_llvm
searchindex returns an unclear error message:
julia> searchindex("q8NWiRVEW3NRTGXW6At7qN5aU2dhJ0ZGHSMfcH50k1Xsn1HIiPNRO0NXVhafnLkEd55","wKPx3QydhlLKz3fZQK3fCVhl6ZwMoFDq18ti1Ozcswi\
Av0Yq4k9utZgoN4KhdvIJiUjQcTIFZyD1cNmdsf1PcmmXMeh10N1",-234321342423423234534534543)
ERROR: error compiling searchindex: error compiling _searchindex: cannot convert type to a julia type
in searchindex at string.jl:307
sprandn with bad args can take an arbitraily long time before erroring out.
displayable psuedo-randomly hangs.
addprocs of negative Int128 causes something crazy to happen.
replace on string containing bad characters errors out instead of skipping past bad character. See sandbox/3.
Special casing of some operators
Address space randomiziation causes some methods to be randomized across invocations of julia.