VeloxRuntimeError when reading parquet file with only meta data

[zhixingheyi-tian]

Describe the bug
This issue is exposed from :

  GlutenDataFrameJoinSuite:
   "SPARK-24690 enables star schema detection even if CBO disabled"

Error：

Job aborted due to stage failure: Task 0 in stage 2.0 failed 1 times, most recent failure: Lost task 0.0 in stage 2.0 (TID 3) (sr250 executor driver): java.lang.RuntimeException: Exception: VeloxRuntimeError
Error Source: RUNTIME
Error Code: INVALID_STATE
Reason: (365 vs. 365)
Retriable: False
Expression: footerLength + 12 < fileLength_
Function: loadFileMetaData
File: ../../velox/dwio/parquet/reader/ParquetReader.cpp
Line: 64
Stack trace:
# 0  
# 1  
# 2  
# 3  
# 4  
# 5  
# 6  
# 7  
# 8  
# 9  
# 10 
# 11 
# 12 
# 13 
# 14 
# 15 
# 16 

To Reproduce

val path = "/tmp/data/"
spark.range(1).selectExpr("id AS c", "id AS f")
  .write.mode("overwrite").parquet(s"$path")
spark.read.parquet(s"$path").show()

Analysis

This is because Spark will generate two parquet files. One has one row data. The other has no data, but with meta data.
Velox parquet reader will encounter issue, when parse the second parquet file.

[zhixingheyi-tian]

Now， it is fixed by oap-project/velox#105