[IMPALA-10703] PrintPath() crashes with ARRAY in ORC format

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: Impala 3.4.0
• Fix Version/s: Impala 4.1.0
• Component/s: Backend
• Labels: complextype orc
• Epic Color: ghx-label-3
• Assignee: Amogh Margoor
• Reporter: Gabor Kaszab
• Created: 14/May/21 13:16
• Updated: 03/Dec/21 05:45
• Resolved: 23/Aug/21 10:43

Description

Repro steps:

• Issue only happens in debug build as apparently there is a DCHECK failing.
• You have to launch Impala with --log_level=3 option to increase the log level.
• Then running this query crashes Impala:

  select inner_arr.ITEM.e from functional_orc_def.complextypestbl tbl, functional_orc_def.complextypestbl.nested_struct.c.d.ITEM inner_arr;
  

 

Backtrace (relevant part):

#7  0x000000000280c2b4 in impala::PrintPath[abi:cxx11](impala::TableDescriptor const&, std::vector<int, std::allocator<int> > const&) (tbl_desc=..., path=...) at /home/gaborkaszab/shadow/Impala-upstream/be/src/util/debug-util.cc:237
#8  0x0000000002a69eeb in impala::HdfsOrcScanner::ResolveColumns (this=0x10e79000, tuple_desc=..., 
    selected_nodes=0x7fe54980a7d0, pos_slots=0x7fe54980a780)
    at /home/gaborkaszab/shadow/Impala-upstream/be/src/exec/hdfs-orc-scanner.cc:452
#9  0x0000000002a69cf7 in impala::HdfsOrcScanner::ResolveColumns (this=0x10e79000, tuple_desc=..., 
    selected_nodes=0x7fe54980a7d0, pos_slots=0x7fe54980a780)
    at /home/gaborkaszab/shadow/Impala-upstream/be/src/exec/hdfs-orc-scanner.cc:449
#10 0x0000000002a6a547 in impala::HdfsOrcScanner::SelectColumns (this=0x10e79000, tuple_desc=...)
    at /home/gaborkaszab/shadow/Impala-upstream/be/src/exec/hdfs-orc-scanner.cc:497
#11 0x0000000002a67720 in impala::HdfsOrcScanner::Open (this=0x10e79000, context=0x7fe54980b260)
    at /home/gaborkaszab/shadow/Impala-upstream/be/src/exec/hdfs-orc-scanner.cc:237
#12 0x00000000029f19c9 in impala::HdfsScanNodeBase::CreateAndOpenScannerHelper (this=0xd280800, partition=0xaac3d80, 
    context=0x7fe54980b260, scanner=0x7fe54980b258)
    at /home/gaborkaszab/shadow/Impala-upstream/be/src/exec/hdfs-scan-node-base.cc:874
#13 0x0000000002baab86 in impala::HdfsScanNode::ProcessSplit (this=0xd280800, filter_ctxs=..., 
    expr_results_pool=0x7fe54980b500, scan_range=0xac59c00, scanner_thread_reservation=0x7fe54980b428)
    at /home/gaborkaszab/shadow/Impala-upstream/be/src/exec/hdfs-scan-node.cc:480
#14 0x0000000002baa28a in impala::HdfsScanNode::ScannerThread (this=0xd280800, first_thread=true, 
    scanner_thread_reservation=8192) at /home/gaborkaszab/shadow/Impala-upstream/be/src/exec/hdfs-scan-node.cc:418
#15 0x0000000002ba95f2 in impala::HdfsScanNode::<lambda()>::operator()(void) const (__closure=0x7fe54980bc28)
    at /home/gaborkaszab/shadow/Impala-upstream/be/src/exec/hdfs-scan-node.cc:339

This DCHECK fails:
https://github.com/apache/impala/blob/a47700ed790c2415e52a85e40063bed53a7cb9e8/be/src/util/debug-util.cc#L237

Check failed: path[i] == 1 (5 vs. 1)

There was a similar issue recently, but here a different DCHECK fails:
https://issues.apache.org/jira/browse/IMPALA-9918