lets-order-some-fries commented on PR #68096:
URL: https://github.com/apache/doris/pull/68096#issuecomment-5777700960

   Thanks — both taken, in `814c7257`.
   
   **1. `convertSortOrdinalsToUnboundSlot`** — you're right, and it's the same 
bug one level up: the
   slot is *named* after the literal, so narrowing it renames it, and `order by 
4294967297` became a
   reference to a column called `"1"`. It now takes the literal's exact digits 
through the same
   `BigDecimal` accessor, so a `LARGEINT` keeps its digits too.
   
   **2. The test** — `assertThrows(Exception.class)` was too loose, agreed. It 
now asserts
   `NereidsException`, which is what `checkPlannerResult` surfaces in the 
sibling tests that use the
   same call (`AnalyzeCTETest`), **and** that the wrapping ordinal fails with 
the *same message* as the
   plainly out-of-range one — that equality is the property being fixed: the 
binder must not be able to
   tell the two statements apart.
   
   One thing I did **not** do: add plan-level sort-key assertions for the two 
`ORDER BY` call sites.
   I still can't run the FE test suite here, and I'd rather not ship a 
`PlanChecker` pattern match I
   haven't seen execute. Checkstyle passes on both commits. If buildall is 
green and you'd still like
   the sort-key checks, I'll add them against a working baseline.
   
   For what it's worth, the reason I can't build turned out to be useful: the 
darwin-arm64 prebuilt
   thirdparty is a release behind master in two tools — `thrift 0.16.0` where 
`fe/pom.xml` wants
   `0.24.0`, and `protoc 3.21.11` where the pom wants protobuf `4.33.4`. 
Symlinking system binaries
   over the first gets the reactor to 25 of 67 modules before `protoc-jar` 
stops honouring the override.
   


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to