Fix -XX:UseSVE=0 for non ARM JVM #2

jcsirot · 2025-01-20T13:30:32Z

The JVM option -XX:UseSVE=0 does not seem to exist for x64 JVM which prevents the JVM to start

… skeleton This annotation processor will collect type info into a file in order to register objects and functions when the module is called. Signed-off-by: Jean-Christophe Sirot <jcsirot@gmail.com>

Only in case `description` field of the annotation is not set. That way it's still possible to override it, specifically for Dagger, but on the other hand it's now possible to only use Java features and not provide fields so there's no duplication. Before: @ModuleFunction(value="echo", description="Returns a contain...") public Container containerEcho(String stringArg) { After: /// Returns a container that echoes... @ModuleFunction public Container echo(String stringArg) { Signed-off-by: Yves Brissaud <gh@lgtd.io>

Rename annotation classes so they are simpler: Module, Object, Function. As close as possible to the Dagger wording. I also moved the info objects (the ones used to generate the json file) to the dagger-java-sdk part so they can also be used to read the json file. Almost created the registry part of the entrypoint: it creates the right dagger module structure, but now it needs to create the right JSON output. This entrypoint can be used as the entrypoint of a module. It's intended to be generic (based on the json file for now, maybe on the annotation in the future). So the module side only needs to declare the mainClass to io.dagger.module.Entrypoint For instance by adding this to the pom.xml: <dependencies> <dependency> <groupId>io.dagger</groupId> <artifactId>dagger-java-sdk</artifactId> <version>1.0.0-SNAPSHOT</version> </dependency> <dependency> <groupId>io.dagger</groupId> <artifactId>dagger-java-annotation-processor</artifactId> <version>1.0.0-SNAPSHOT</version> <scope>provided</scope> </dependency> </dependencies> <build> <pluginManagement> <plugin> <artifactId>maven-compiler-plugin</artifactId> <version>3.13.0</version> <configuration> <verbose>false</verbose> <annotationProcessorPaths> <path> <groupId>io.dagger</groupId> <artifactId>dagger-java-annotation-processor</artifactId> <version>1.0.0-SNAPSHOT</version> </path> </annotationProcessorPaths> <annotationProcessors> <annotationProcessor>io.dagger.annotation.processor.DaggerModuleAnnotationProcessor</annotationProcessor> </annotationProcessors> </configuration> </plugin> </pluginManagement> <plugins> <plugin> <groupId>org.apache.maven.plugins</groupId> <artifactId>maven-shade-plugin</artifactId> <version>3.5.0</version> <executions> <execution> <phase>package</phase> <goals> <goal>shade</goal> </goals> <configuration> <transformers> <transformer implementation="org.apache.maven.plugins.shade.resource.ManifestResourceTransformer"> <mainClass>io.dagger.module.Entrypoint</mainClass> </transformer> </transformers> </configuration> </execution> </executions> </plugin> </plugins> </build> This should be good to create a jar with all the dependencies: ./mvnw clean package This includes the management of the annotations. And then the generated jar declares the right entrypoint, so this should be good to run with: dagger run java -jar dagger-module.jar This will not completely work as the module part is not done, so the dagger engine doesn't know about the module and so the registration can't be done. But that's a different story. The goal here was to be able to: - document with annotations - create a first basic entrypoint - declare as much as possible in the SDK so it's easy to integrate Signed-off-by: Yves Brissaud <gh@lgtd.io>

This is a very simple one, just provided so that modules don't have define the `dag` field. In some way this is just to enforce some conventions. Signed-off-by: Yves Brissaud <gh@lgtd.io>