[GLUTEN-6887][VL] Daily Update Velox Version (2026_09_17) (#13043)

* [GLUTEN-6887][VL] Daily Update Velox Version (dft-2026_09_17)

Upstream Velox's New Commits:
fdfbc1c1a by Suryadev Sahadevan Rajesh, perf(encoding): Prune SubIntSplit boundary candidates (#19010)
447c01129 by Krishna Pai, misc!: Reconcile fbcode and GitHub source trees (#19057)
16a0305ea by Natasha Sehgal, feat: Add legacy_timestamp_with_timezone query config
5bf9a8ba8 by RindsSchei225e, refactor(nimble): Remove the raw size stats consistency check (#18898)
07fd55cb3 by Suryadev Sahadevan Rajesh, perf(encoding): Skip constant SubIntSplit sections when decoding (#19051)
a597b2d39 by Maria Basmanova, feat(rpc): Look up RPC function signatures without stubs (#19052)
bf41c616b by abhinavmuk04, fix(trace): Redact credentials from task trace metadata (#19053)
3c38b3161 by Amit Dutta, refactor: Remove backward compatibility macro
52986e2cb by kevinwilfong, fix(fuzzer): Parenthesize derived tables in DuckDB reference SQL (#18931)
3bcef3b0f by tanjialiang, refactor(serializer): Remove kLegacy nimble serialization format (#18872)
b08de36f0 by hu hengrui, perf: Prefetch in RowContainer::listRows on x86
db3421f83 by Shruti Shivakumar, fix(cudf): Index BETWEEN column bounds correctly
1074fc019 by Shruti Shivakumar, test(cudf): Disable CPU fallback in GPU tests
b83ea4c16 by Ke, feat(fuzzer): Report query-shape coverage
18bb55199 by Jimmy Lu, feat: Propagate worker count and splits through a nested fixed point (#18958)
c2366f300 by Suryadev Sahadevan Rajesh, perf(encoding): Direct-indexed frequency counting for narrow SubIntSplit segments (#19009)
e647cd3d1 by Suryadev Sahadevan Rajesh, perf(encoding): Trim constant bit planes before the SubIntSplit split DP (#18908)
584c4fa05 by Ke, feat(tablet): Add V2 chunk stats writer
5c21e9f47 by Andrii Rosa, build: Bump FB_OS_VERSION to v2026.09.07.00 (#18968)
032350cc7 by Krishnakanth, fix(json): Collapse duplicate keys in json_parse (keep last) (#18212)
c30a0cdc5 by Suryadev Sahadevan Rajesh, feat(encoding): Zigzag-delta pre-transform for SubIntSplit (#18803)
ba26f411a by Xiaoxuan, perf(nimble): Optimize nullable materialization (#19024)
5efabda9b by HighW4y2H3ll, refactor: Use string_view data()+size() instead of end() with raw pointers
43941fb64 by LingBin, fix(functions): Avoid string_view iterator-to-pointer use(libc++)
b9956f766 by Wenxin Cheng, fix(wave): Fix kernel_ld_test build break from CUDA 13 cuCtxCreate_v4 signature
e680a48ae by peterenescu, fix(sparksql): Support to_json template for disabling scalar check (#19030)
025d471ed by Deepthi Bose, feat(function): Update functions to support UNKNOWN type
7ff6c21c7 by Andrii Rosa, build: Migrate S3 test storage from MinIO to pgsty/silo
0e9f3babd by XiaoDu, fix(nimble): Apply layouts to predefined FlatMap keys (#18970)
e79a05ee7 by Jimmy Lu, feat: Convergence criterion as a plan sequence (#18854)
a917d59bb by tanjialiang, refactor(sequence): Migrate Sequence Storage callers off nimble kLegacy (#18948)
ec3ce27fe by Naveen Mahadevuni, refactor(iceberg): Add folly dependency to velox_hive_iceberg_test
ae7e5b4be by Maria Basmanova, fix(expression): Extract filters through widening casts
ee0d6ec79 by HighW4y2H3ll, refactor: Do not assume `std::string_view::end()` is a raw pointer
0b7e4cf29 by HighW4y2H3ll, refactor: Compare `pos` against `data() + size()` instead of `std::string_view::end()`
6d4fcd76d by Ke, feat(fuzzer): Add configuration coverage framework (#19005)
342ba1a02 by tanjialiang, perf(index): Devirtualize per-stream bufferRef, hoist locals in collectStripeStreamViews (#18988)
13de20e15 by Zac, feat(nimble): Add IndexLookup::KeyCursor for full-index scans
0e156dca5 by tanjialiang, perf(tablet): Devirtualize StripeGroup::streamMajor readAt via dispatch helper (#18987)
ea8f4019c by RindsSchei225e, perf(nimble): Fuse the string size scan into the min/max scan
b12ed29f1 by Bradley Dice, fix(dwio): Join cache executor before shutdown
e3164fdfd by PRASHANT GOLASH, perf(nimble): Optimize FSST reader validation and output materialization (#18937)
48be91e65 by tanjialiang, perf(tablet): Persist DirectDataInput load-path scratch vectors (#18986)
208c508a9 by tanjialiang, perf(tablet): Guard DirectDataInput::load per-group sort with is_sorted (#18985)
ca9595f05 by tanjialiang, perf(encoding): MainlyConstant count skips Trivial<Bool> materialize (#18984)
99dbc8cca by Krishnakanth, fix(exec): Gate multi-operator plan-node stats by operator boundary
059a0d648 by macduan, perf(nimble): Add a dense bulk visitor path for ALP decoding
7ae3bfc58 by tanjialiang, perf(encoding): Reuse RLE sliceRuns scratch via ScopedVector + bufferPool (#18983)
a4de5771b by tanjialiang, perf(encoding): Slice BlockBitPacking offsets via SliceEncoding delta (#18982)
27fc77e37 by Mariam AlMesfer, fix(sparksql): Make unix_timestamp/to_unix_timestamp ANSI-aware
cd2891814 by Yuxuan Chen, Use `string_view::data()` instead of `begin()` for the `memcpy` source
f6105f168 by Yuxuan Chen, Use `data()` instead of `std::string_view::begin()` where a raw pointer is required
38436f6fc by Kevin Tang, fix(geospatial): raise the bing tile zoom-span limit as a user error (#18993)
17fbf01d2 by Dong Wang, misc: Fix typos in FileSplitReader.h
6bd17de0b by macduan, perf(dwio): Retrieves the ScanSpec directly  when creating column loaders
04eccf71e by tanjialiang, fix(encoding): Handle zero-bit-width FBW in copyPackedBits
563634c34 by tanjialiang, perf(encoding): Apply SliceEncoding value delta to sorted position streams (#18979)
d76ab551e by Shrinidhi Joshi, feat(exchange): Add file durable exchange backend (#18871)

Author: GlutenPerfBot

ep/build-velox/src/get-velox.sh

@@ -18,8 +18,8 @@ set -exu
 
 CURRENT_DIR=$(cd "$(dirname "$BASH_SOURCE")"; pwd)
 VELOX_REPO=https://github.com/IBM/velox.git
-VELOX_BRANCH=dft-2026_09_14
-VELOX_ENHANCED_BRANCH=ibm-2026_09_14
+VELOX_BRANCH=dft-2026_09_17
+VELOX_ENHANCED_BRANCH=ibm-2026_09_17
 VELOX_HOME=""
 RUN_SETUP_SCRIPT=ON
 ENABLE_ENHANCED_FEATURES=OFF

gluten-ut/spark34/src/test/scala/org/apache/spark/sql/GlutenDateFunctionsSuite.scala

@@ -16,6 +16,9 @@
  */
 package org.apache.spark.sql
 
+import org.apache.gluten.utils.BackendTestUtils
+
+import org.apache.spark.SparkException
 import org.apache.spark.sql.catalyst.util.DateTimeUtils
 import org.apache.spark.sql.functions._
 import org.apache.spark.sql.internal.SQLConf
@@ -188,8 +191,16 @@ class GlutenDateFunctionsSuite extends DateFunctionsSuite with GlutenSQLTestsTra
             Seq(Row(secs(ts5.getTime)), Row(null)))
 
           // invalid format
+          // Velox rejects the unsupported specifier 'b' while building the Joda
+          // formatter, as vanilla Spark does, though it reports a different
+          // exception type. The legacy formatter returns null instead of failing.
           val invalid = df1.selectExpr(s"to_unix_timestamp(x, 'yyyy-MM-dd bb:HH:ss')")
-          checkAnswer(invalid, Seq(Row(null), Row(null), Row(null), Row(null)))
+          if (legacyParserPolicy == "legacy" || !BackendTestUtils.isVeloxBackendLoaded()) {
+            checkAnswer(invalid, Seq(Row(null), Row(null), Row(null), Row(null)))
+          } else {
+            val e = intercept[SparkException](invalid.collect())
+            assert(e.getMessage.contains("Specifier b is not supported"))
+          }
         }
     }
   }

gluten-ut/spark35/src/test/scala/org/apache/spark/sql/GlutenDateFunctionsSuite.scala

@@ -16,6 +16,9 @@
  */
 package org.apache.spark.sql
 
+import org.apache.gluten.utils.BackendTestUtils
+
+import org.apache.spark.SparkException
 import org.apache.spark.sql.catalyst.util.DateTimeUtils
 import org.apache.spark.sql.functions._
 import org.apache.spark.sql.internal.SQLConf
@@ -203,8 +206,16 @@ class GlutenDateFunctionsSuite extends DateFunctionsSuite with GlutenSQLTestsTra
             df.selectExpr("to_unix_timestamp(ts, 'invalid-format')"),
             Seq(Row(secs(ts1.getTime)), Row(secs(ts2.getTime))))
 
+          // Velox rejects the unsupported specifier 'b' while building the Joda
+          // formatter, as vanilla Spark does, though it reports a different
+          // exception type. The legacy formatter returns null instead of failing.
           val invalid = df1.selectExpr(s"to_unix_timestamp(x, 'yyyy-MM-dd bb:HH:ss')")
-          checkAnswer(invalid, Seq(Row(null), Row(null), Row(null), Row(null)))
+          if (legacyParserPolicy == "legacy" || !BackendTestUtils.isVeloxBackendLoaded()) {
+            checkAnswer(invalid, Seq(Row(null), Row(null), Row(null), Row(null)))
+          } else {
+            val e = intercept[SparkException](invalid.collect())
+            assert(e.getMessage.contains("Specifier b is not supported"))
+          }
         }
     }
   }

gluten-ut/spark40/src/test/scala/org/apache/spark/sql/GlutenDateFunctionsSuite.scala

@@ -206,8 +206,16 @@ class GlutenDateFunctionsSuite extends DateFunctionsSuite with GlutenSQLTestsTra
             df.selectExpr("to_unix_timestamp(ts, 'invalid-format')"),
             Seq(Row(secs(ts1.getTime)), Row(secs(ts2.getTime))))
 
+          // Velox rejects the unsupported specifier 'b' while building the Joda
+          // formatter, as vanilla Spark does, though it reports a different
+          // exception type. The legacy formatter returns null instead of failing.
           val invalid = df1.selectExpr(s"to_unix_timestamp(x, 'yyyy-MM-dd bb:HH:ss')")
-          checkAnswer(invalid, Seq(Row(null), Row(null), Row(null), Row(null)))
+          if (legacyParserPolicy == "legacy" || !BackendTestUtils.isVeloxBackendLoaded()) {
+            checkAnswer(invalid, Seq(Row(null), Row(null), Row(null), Row(null)))
+          } else {
+            val e = intercept[SparkException](invalid.collect())
+            assert(e.getMessage.contains("Specifier b is not supported"))
+          }
         }
     }
   }

gluten-ut/spark41/src/test/scala/org/apache/spark/sql/GlutenDateFunctionsSuite.scala

@@ -16,6 +16,9 @@
  */
 package org.apache.spark.sql
 
+import org.apache.gluten.utils.BackendTestUtils
+
+import org.apache.spark.SparkException
 import org.apache.spark.sql.catalyst.util.DateTimeUtils
 import org.apache.spark.sql.functions._
 import org.apache.spark.sql.internal.SQLConf
@@ -203,8 +206,16 @@ class GlutenDateFunctionsSuite extends DateFunctionsSuite with GlutenSQLTestsTra
             df.selectExpr("to_unix_timestamp(ts, 'invalid-format')"),
             Seq(Row(secs(ts1.getTime)), Row(secs(ts2.getTime))))
 
+          // Velox rejects the unsupported specifier 'b' while building the Joda
+          // formatter, as vanilla Spark does, though it reports a different
+          // exception type. The legacy formatter returns null instead of failing.
           val invalid = df1.selectExpr(s"to_unix_timestamp(x, 'yyyy-MM-dd bb:HH:ss')")
-          checkAnswer(invalid, Seq(Row(null), Row(null), Row(null), Row(null)))
+          if (legacyParserPolicy == "legacy" || !BackendTestUtils.isVeloxBackendLoaded()) {
+            checkAnswer(invalid, Seq(Row(null), Row(null), Row(null), Row(null)))
+          } else {
+            val e = intercept[SparkException](invalid.collect())
+            assert(e.getMessage.contains("Specifier b is not supported"))
+          }
         }
     }
   }