Problems with JDK 10

Stefan Marr java at
Thu Mar 22 23:14:27 UTC 2018


The OpenJDK bug systems seems to be for members only.
Could some possibly lend a hand in filing an issue or tell me how I could communicate this to the relevant javac people?

The issue:

Since JDK 10 (also confirmed with OpenJDK 11-ea+5), the code at the URL [1] results in:

src/som/primitives/ error: expression not allowed as annotation value
@Primitive(primitive = "threading:threadSpawn:with:",

The problem seems to be a somewhat odd parse error, which recognizes the name of a class, which needs to be generated by an annotation processors, as an expression.
This in turn causes javac to skip execution of any annotation processors. Specifically, javac aborts before the first round of annotation processing is started.

Steps to reproduce:

Take code at [1] extract and run

ant compile-som

It succeeds with JDK 9, but not with any later JDK.

The code is as small and self-contained as I could get it while still producing the error.
It comes with a precompiled library and annotation processor.
However, these are likely not influencing the bug.

The same parse error can be observed without the annotation processor being available.
The error goes away if the missing ToArgumentsArrayNodeFactory class is provided manually.

I would expect the behavior to be identical between Java 8 and 9.
Not sure what change would explain this behavior.

Currently, I don’t have a work around, so, this blocks use of Java 10.


Thank you

> On 22 Mar 2018, at 15:16, Stefan Marr <java at> wrote:
> Hi Doug:
>> On 22 Mar 2018, at 11:10, Doug Simon <doug.simon at> wrote:
>> This looks like a javac bug. Maybe turn on more flags to see what annotation processing is happening such as -XprintProcessorInfo and -XprintRounds.
> Ok, there is no annotation processing happing at all.
> I experimented a little more and there might be a problem with annotation processing, dependencies on generated classes, and possibly circular dependencies.
> I guess this is a javac bug.
> On my code base, javac seems to be in a highly problematic state.
> Even the error output is in some scenarios pointing to random things, which looks to me like internal data structures are corrupted.
> Thanks
> Stefan
> -- 
> Stefan Marr
> School of Computing, University of Kent

Stefan Marr
School of Computing, University of Kent

More information about the graal-dev mailing list