viirya opened a new issue, #7183:
URL: https://github.com/apache/datafusion-comet/issues/7183
### Describe the bug
Several JNI entry points in `native/core/src/execution/jni_api.rs` convert
their arguments with `unwrap()` / `expect()`, so a malformed argument panics
instead of returning a `CometError`:
- `convert_datatype_arrays` (used by `columnarToRowInit`,
`writeSortedFileNative` and `writeSortedPartitionsNative`) calls
`serde::deserialize_data_type(bytes).unwrap()`.
- `writeSortedFileNative` reads its `file_path` and `compression_codec`
strings with `JString::try_to_string(env).unwrap()`. `traceBegin`, `traceEnd`
and `logMemoryUsage` read their name argument the same way.
- `get_execution_context`, which `executePlan` and
`getShufflePartitionOffsets` use, calls `.expect("Comet execution context
shouldn't be null!")` on the handle that the JVM passes in.
`try_unwrap_or_throw` catches the panic and raises a `CometNativeException`,
so the JVM still gets an exception. However, that exception carries only the
panic message, such as `called `Result::unwrap()` on an `Err` value: ...`,
without saying which argument was bad. The backtrace is written to the
executor's stderr.
### Steps to reproduce
The JVM side never passes such arguments today, so this cannot be reproduced
through Spark. Reaching it requires a direct native call with a malformed
argument, for example a byte array that is not a serialized `DataType` passed
to `columnarToRowInit`, or a null `String` passed to `traceBegin`.
### Expected behavior
These conversions return a `CometError` through `?` with a message that
names the argument. The exception class stays `CometNativeException`, and the
conversion no longer goes through the panic path.
### Additional context
Exports that are not guarded at all (`getRustThreadId`,
`NativeBase.release`, `nativeSupportsPythonUdf`) are out of scope.
### Willingness to contribute
I can contribute a fix for this bug independently
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
