Chronicle-Queue
Chronicle-Queue copied to clipboard
OpenHFT
LastAppendedTest fails when -DdisableProxyCodegen=true
see also this test - net.openhft.chronicle.wire.method.MethodWriterProxyTest
which is ignored:
I suggest we remove disableProxyGen as it is broken anyway, it was added in 2020
I see no reason to keep this property, +1 for removal.
P. S. It's enabled in Chronicle-Queue-Demo\order-processor\pom.xml - we should check if removal will impact the demo.
@glukos I have seen method reader/writer code generation not working when running with mvn:exec. I assume that is why that property is set in chronicle-queue-demo. There are other examples (e.g. chronicle-services benchmarks) where we commit the generated code to git to work around this.
Good news: I just tried removing that property from chronicle-queue-demo and ran the examples with mvn:exec and there were no problems.
I tried to run one of the chronicle services benchmarks without pre-generated code and it failed - but it worked once I bumped chronicle-core version in chronicle-bom (to bring in @alamar 's fix to core: https://github.com/OpenHFT/Chronicle-Core/pull/388). See also https://github.com/ChronicleEnterprise/Chronicle-Services/pull/326)
This is a very complicated area and we have some cleanup to do. As the concerns are in Wire, please see https://github.com/OpenHFT/Chronicle-Wire/issues/460
We can't just remove disableProxyCodeGen - we still need to support proxy method readers/writers - see https://github.com/OpenHFT/Chronicle-Wire/pull/473.
We could fix proxy method readers/writers for this test, but I can't see the cost/benefit as they are legacy and with the above PR, generated method readers/writers support everything they need to.
I'll mark this issue as low priority and we can close once proxy method readers/writers are gone (or we can close now with wontfix now)