The GitHub Actions job "CodeQL" on hive.git/parquet-io-cache has succeeded. Run started by GitHub user abstractdog (triggered by abstractdog).
Head commit for run: 78b4620bd00fcb7a4775fc207981fa4c3b8e9ad9 / Denys Kuzmenko <[email protected]> HIVE-30059: The stock Parquet reader caches with the daemon's layout, as the native reader does - LlapCacheAwareFs built its ParquetCacheLayout from the conf given to registerFile, the job's, while ParquetEncodedDataReader builds it from the daemon's; when the job's hive.llap.io.encode.alloc.size differs from the daemon's (set only on the daemon, or by a session SET) the two readers cut a chunk into different pieces: a stock read re-caches an evicted tail as one buffer, and a concurrent miss fails the query with "Found a block with different length at the same offset" - registerFile takes the layout; VectorizedParquetRecordReader builds it from the daemon's cache conf - testStockReadRecachesAnEvictedTailAsItsPieces fails with the layout built from the job conf Report URL: https://github.com/apache/hive/actions/runs/38711125413 With regards, GitHub Actions via GitBox --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
