Re: PATCH: CITEXT 2.0 v3

"David E. Wheeler" <david@kineticode.com> writes:
>      lcstr = str_tolower(VARDATA_ANY(left), VARSIZE_ANY_EXHDR(left));
>      rcstr = str_tolower(VARDATA_ANY(right), VARSIZE_ANY_EXHDR(right));

>      result = varstr_cmp(
>          lcstr,
>          VARSIZE_ANY_EXHDR(left),
>          rcstr,
>          VARSIZE_ANY_EXHDR(right)
>      );

Not sure about a memory leak, but this code is clearly wrong if
str_tolower results in a change of physical string length (clearly
possible in Turkish, for instance, and probably in some other
locales too).  You need to do strlen's on the lowercased strings.
        regards, tom lane