> On 9 Sep 2024, at 17:29, Jacob Champion <jacob.champion@enterprisedb.com> wrote:
> pg_utf8_string_len() doesn't check the remaining string length before
> calling pg_utf8_is_legal(), so there's a possibility of jumping a
> couple of bytes past the end of the string. (The overread stops there,
> because the function won't validate a sequence containing a null
> byte.)
>
> Here's a quick patch to fix it. I didn't see any other uses of
> pg_utf8_is_legal() with missing length checks.
Just to make sure I understand, this is for guarding against overreads in
validation of strings containing torn MB characters? Assuming I didn't
misunderstand you this patch seems correct to me.
--
Daniel Gustafsson