Re: pg_parse_json() should not leak token copies on failure

On Wed, Oct 2, 2024 at 10:45 AM Andrew Dunstan <andrew(at)dunslane(dot)net> wrote:
> Generally looks good. Should we have a check in
> setJsonLexContextOwnsTokens() that we haven't started parsing yet, for
> the incremental case?

Good catch. Added in v4.

> > At the moment, we have a test matrix consisting of "standard frontend"
> > and "shlib frontend" tests for the incremental parser. I'm planning
> > for the v4 patch to extend that with a "owned/not owned" dimension;
> > any objections?
> >
>
> Sounds reasonable.

This is also done, along with runs of pgindent/perltidy.

I've added a 0002 as well. While running tests under Valgrind, it
complained about the uninitialized dummy_lex on the stack, which is
now fixed. (That bug was introduced with my patch in 0785d1b8b and is
technically orthogonal to 0001, but I figured I could track it here
for now.) This is also how I found out that my existing fuzzers
weren't checking for uninitialized memory, like I thought they were.
Grumble...

Thanks,
--Jacob