afs/rdf-delta

GraalVM Native Image

Closed this issue · 2 comments

mjulianotq commented

Considering the role this application fills, it's an excellent candidate for GraalVM's Native Image. I've given it a try, and it seems that the reliance on reflection of slf4j is causing a bit of a problem. I'm including the output of my attempt below. I want to understand this better myself, so I'll likely either open a pull request if one is needed or write a guide here on this ticket for inclusion wherever it may be appropriate.

Error: Unsupported features in 5 methods
Detailed message:
Error: com.oracle.graal.pointsto.constraints.UnresolvedElementException: Discovered unresolved method during parsing: org.slf4j.impl.StaticMDCBinder.getSingleton(). To diagnose the issue you can use the --allow-incomplete-classpath option. The missing method is then reported at run time when it is accessed the first time.
Trace:
        at parsing org.slf4j.MDC.bwCompatibleGetMDCAdapterFromBinder(MDC.java:99)
Call path from entry point to org.slf4j.MDC.bwCompatibleGetMDCAdapterFromBinder():
        no path found from entry point to target method

Error: com.oracle.graal.pointsto.constraints.UnresolvedElementException: Discovered unresolved method during parsing: org.slf4j.impl.StaticMarkerBinder.getSingleton(). To diagnose the issue you can use the --allow-incomplete-classpath option. The missing method is then reported at run time when it is accessed the first time.
Trace:
        at parsing org.slf4j.MarkerFactory.bwCompatibleGetMarkerFactoryFromBinder(MarkerFactory.java:61)
Call path from entry point to org.slf4j.MarkerFactory.bwCompatibleGetMarkerFactoryFromBinder():
        no path found from entry point to target method

Error: com.oracle.graal.pointsto.constraints.UnresolvedElementException: Discovered unresolved type during parsing: org.apache.log4j.Logger. To diagnose the issue you can use the --allow-incomplete-classpath option. The missing type is then reported at run time when it is accessed the first time.
Trace:
        at parsing io.netty.util.internal.logging.Log4JLoggerFactory.newInstance(Log4JLoggerFactory.java:38)
Call path from entry point to io.netty.util.internal.logging.Log4JLoggerFactory.newInstance(String):
        no path found from entry point to target method

Error: com.oracle.graal.pointsto.constraints.UnresolvedElementException: Discovered unresolved type during parsing: org.osgi.framework.FrameworkUtil. To diagnose the issue you can use the --allow-incomplete-classpath option. The missing type is then reported at run time when it is accessed the first time.
Trace:
        at parsing org.apache.logging.log4j.core.config.plugins.util.ResolverUtil.loadImplementationsInBundle(ResolverUtil.java:282)
Call path from entry point to org.apache.logging.log4j.core.config.plugins.util.ResolverUtil.loadImplementationsInBundle(ResolverUtil$Test, String):
        no path found from entry point to target method

Error: com.oracle.graal.pointsto.constraints.UnresolvedElementException: Discovered unresolved type during parsing: org.slf4j.ext.EventData. To diagnose the issue you can use the --allow-incomplete-classpath option. The missing type is then reported at run time when it is accessed the first time.
Trace:
        at parsing org.apache.logging.slf4j.EventDataConverter.convertEvent(EventDataConverter.java:33)
Call path from entry point to org.apache.logging.slf4j.EventDataConverter.convertEvent(String, Object[], Throwable):
        at org.apache.logging.slf4j.EventDataConverter.convertEvent(EventDataConverter.java:33)
        at org.apache.logging.slf4j.Log4jLogger.log(Log4jLogger.java:367)
        at org.apache.commons.logging.impl.SLF4JLocationAwareLog.debug(SLF4JLocationAwareLog.java:131)
        at com.amazonaws.http.IdleConnectionReaper.run(IdleConnectionReaper.java:194)
        at com.oracle.svm.core.thread.JavaThreads.threadStartRoutine(JavaThreads.java:519)
        at com.oracle.svm.core.posix.thread.PosixJavaThreads.pthreadStartRoutine(PosixJavaThreads.java:192)
        at com.oracle.svm.core.code.IsolateEnterStub.PosixJavaThreads_pthreadStartRoutine_e1f4a8c0039f8337338252cd8734f63a79b5e3df(generated:0)
afs commented

The patch server is a long-running process so the benefits of AOT are not so great (indeed, The GraalVM team reported that AOT is 5-% slower than JIT.)

It would be nice to have for the command line dcmd operations where start-up is noticeable.

The IT world seems to be moving towards more variety of instruction sets (e.g. ARM servers). Including instructions and template files would help users.