> On Tue, Aug 18, 2026 at 8:54 AM Tender Wang <tndrwang@gmail.com> wrote:
> The attached patch fixes this by checking the CASE argument against
> the equality operator and input collation of each WHEN comparison
> using the existing grouping_check_operand() logic.
I confirmed that this is a real wrong-result issue, and this approach
looks correct to me.
One small issue with the patch:
+ Assert(IsA(cw->expr, OpExpr));
+ opexpr = (OpExpr *) cw->expr;
cw->expr is not guaranteed to be an OpExpr. transformCaseExpr() calls
coerce_to_boolean() after constructing the "=" expression. With a
user-defined "=" operator returning a non-boolean type and an assignment
cast from that type to boolean, cw->expr becomes a FuncExpr wrapping the
OpExpr.
Therefore, the assertion can fail in an assert build, while the cast is
unsafe in a non-assert build. I think the non-OpExpr case should be
handled conservatively as a grouping conflict.
Best Regards,
Yuhang Qiu
