Re: I created a faster JSON parser

[Herman , Géza]

Christopher Wellons <wellons@nullprogram.com> writes:

> > I'd glad if you can give some advices: which fuzzy-testing 
> > framework
> > to use, which introductory material is worth reading, etc.
>
> I'm partial to AFL++, and it's what I reach for first. It also 
> works
> with GCC. It has two modes, with persistent mode preferred:

Thanks so much for the description!  I created a standalone 
version of my parser (I attached it), and used "afl-clang-fast -o 
json json.c -fsanitize=address,undefined" and afl-fuzz to test it. 
It's been running for an hour, the tester didn't find any problems 
yet.

I discovered a funny clang bug: it incorrectly optimizes around 
setjmp in do_test(): when json_parser_init runs, it stores the 
workspace pointer in a register.  And if there is an error during 
JSON parsing, it will always free the pointer which is in that 
register.  But in the meantime (I mean, after json_parser_init, 
and before the error is thrown), the parser could have updated 
it. So free() will be called on an already freed block.  I had to 
add a dummy printf("free!\n"); to circumvent this optimization.

json.c
Description: Text Data