Iceberg-spark-runtime-4.1_2.13-1.11.0 fails in Spark-only runtime when Flink classes are absent

[PenneyHuang]

Apache Iceberg version

1.11.0 (latest release)

Query engine

Spark

Please describe the bug 🐞

Using the official iceberg-spark-runtime-4.1_2.13-1.11.0.jar in a Spark-only runtime fails when Flink table classes are not present.

FormatModelRegistry eagerly tries to register org.apache.iceberg.flink.data.FlinkFormatModels during static initialization. That code references Flink      
classes such as org.apache.flink.table.data.RowData.                                                                                                        
                                                                                                                                                            
If those Flink classes are absent, Spark reads fail with:                                                                                                   
                                                                                                                                                            
  ─ text                                                                                                                                                    
  java.lang.NoClassDefFoundError: Could not initialize class org.apache.iceberg.formats.FormatModelRegistry                                                 
  ...                                                                                                                                                       
  Caused by: java.lang.NoClassDefFoundError: org/apache/flink/table/data/RowData                                                                            
      at org.apache.iceberg.flink.data.FlinkFormatModels.register(FlinkFormatModels.java:30)                                                                
      at org.apache.iceberg.formats.FormatModelRegistry.registerSupportedFormats(FormatModelRegistry.java:207)                                              
      at org.apache.iceberg.formats.FormatModelRegistry.<clinit>(FormatModelRegistry.java:69)                                                               
                                                                                                                                                            
This happens in the normal Spark reader path (BaseBatchReader), even after disabling Auron/native execution.                                                
                                                                                                                                                            
Steps to reproduce                                                                                                                                          
                                                                                                                                                            
1. Start Spark 4.1                                                                                                                                          
2. Add iceberg-spark-runtime-4.1_2.13-1.11.0.jar                                                                                                            
3. Read any Iceberg table in Spark SQL                                                                                                                      
                                                                                                                                                            
Expected behavior                                                                                                                                           
                                                                                                                                                            
A Spark-only runtime should be able to use the Spark runtime jar without requiring Flink runtime classes.                                                   
                                                                                                                                                            
Notes                                                                                                                                                       
                                                                                                                                                            
This looks similar to #16720, but with missing Flink classes instead of missing ORC classes.                                                                
                                                                                                                                                            
It seems FormatModelRegistry should skip optional engine-specific registrations when transitive dependencies are absent.

Willingness to contribute

• I can contribute a fix for this bug independently
• I would be willing to contribute a fix for this bug with guidance from the Iceberg community
• I cannot contribute a fix for this bug at this time

[wombatu-kun]

Duplicate of #16720: same root cause, just the Flink registrar instead of ORC. FormatModelRegistry.registerSupportedFormats() catches only NoSuchMethodException, so the NoClassDefFoundError thrown inside FlinkFormatModels.register() (when org.apache.flink.table.data.RowData is absent) escapes the static initializer and permanently poisons FormatModelRegistry for every subsequent read.

Open PR #16611 already addresses this: it widens that loop's catch to NoSuchMethodException | NoClassDefFoundError | ExceptionInInitializerError, and because the loop iterates every entry in CLASSES_TO_REGISTER (including org.apache.iceberg.flink.data.FlinkFormatModels), the Flink case is covered there too. Suggest tracking this as a duplicate of #16720 and closing once #16611 merges.