Details
-
Bug
-
Status: Open
-
Major
-
Resolution: Unresolved
-
1.37.0
-
None
-
None
-
None
Description
I am seeing deadlocks during initialization of interface RelDataTypeSystem and class RelDataTypeSystemImpl. The command ./gradlew :core:test --tests RelDataTypeSystemTest hangs about 90% of the time for me. Here's part of the thread dump running under JDK 21.
"ForkJoinPool-1-worker-7" #30 [1161671] daemon prio=5 os_prio=0 cpu=8.26ms elapsed=19.26s tid=0x00007d08b0000ff0 nid=1161671 waiting on condition [0x00007d098b3fd000] java.lang.Thread.State: RUNNABLE at org.apache.calcite.sql.type.SqlTypeFixture.<init>(SqlTypeFixture.java:28) - waiting on the Class initialization monitor for org.apache.calcite.rel.type.RelDataTypeSystem at org.apache.calcite.sql.type.RelDataTypeSystemTest$Fixture.<init>(RelDataTypeSystemTest.java:143) at org.apache.calcite.sql.type.RelDataTypeSystemTest.testCustomDecimalModReturnTypeInference(RelDataTypeSystemTest.java:310) at java.lang.invoke.LambdaForm$DMH/0x00007d0934160400.invokeVirtual(java.base@21/LambdaForm$DMH) ... at java.util.concurrent.ForkJoinWorkerThread.run(java.base@21/ForkJoinWorkerThread.java:188) "ForkJoinPool-1-worker-8" #31 [1161672] daemon prio=5 os_prio=0 cpu=7.33ms elapsed=19.26s tid=0x00007d08a8000ff0 nid=1161672 waiting on condition [0x00007d098b2fc000] java.lang.Thread.State: RUNNABLE at org.apache.calcite.rel.type.RelDataTypeSystem.<clinit>(RelDataTypeSystem.java:38) - waiting on the Class initialization monitor for org.apache.calcite.rel.type.RelDataTypeSystemImpl at org.apache.calcite.sql.type.SqlTypeFixture.<init>(SqlTypeFixture.java:28) at org.apache.calcite.sql.type.RelDataTypeSystemTest$Fixture.<init>(RelDataTypeSystemTest.java:143) at org.apache.calcite.sql.type.RelDataTypeSystemTest.testDecimalModReturnTypeInference(RelDataTypeSystemTest.java:229) ...
Occurs in current main (commit a8802c7).
Not a high priority. Just documenting that this bug exists. Hopefully it only exists when running tests.
Attachments
Issue Links
- is related to
-
CALCITE-6464 Type inference for DECIMAL division seems incorrect
- Closed
- relates to
-
CALCITE-6560 Allow types with negative scale
- Closed