Closed Depetrol closed 3 months ago
I think the right fix is that the RTI should be code generated like any of the federates and deployed along with them. There should be no such thing as a stand-alone, separately installable RTI. This will also enable progress towards a more decentralized and fault tolerant RTI.
@edwardalee, what is proposed by @Depetrol can be implemented in just a couple of hours of work and fixes an immediate issue that has been left dangling for a long time. I think it's a good idea to apply this patch and discuss further possible refactoring efforts separately.
Also see related discussion and work in progress here: https://github.com/lf-lang/reactor-c/pull/464
Let me transfer the issue to the reactor-c repo.
Ok, but it’s only a temporary fix. The next time the RTI is updated, the same issue will arise.
Ok, but it’s only a temporary fix. The next time the RTI is updated, the same issue will arise.
No, the point is that the issue won't arise any longer (as long as you're using Docker) because each push to master
will result in an updated image being pushed to DockerHub.
Automatically published images can be found here: https://hub.docker.com/r/lflang/rti/tags
Currently the docker RTI image in Docker hub is 6 months ago. This can cause unintended failures in the docker target with message type mismatches. I'm proposing we add a continuous deployment pipeline that:
master
branch aslflang/rti:latest
lflang/rti:<version>