The fileset classes not present in resulting index
mvysny opened this issue · comments
I have the following maven project: pom.xml is as follows:
<?xml version="1.0" encoding="UTF-8"?>
<project xmlns="http://maven.apache.org/POM/4.0.0"
xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance"
xsi:schemaLocation="http://maven.apache.org/POM/4.0.0 http://maven.apache.org/xsd/maven-4.0.0.xsd">
<modelVersion>4.0.0</modelVersion>
<groupId>com.github.mvysny.vaadin-jandex</groupId>
<artifactId>vaadin-jandex</artifactId>
<version>14.4.6-SNAPSHOT</version>
<properties>
<maven.compiler.source>8</maven.compiler.source>
<maven.compiler.target>8</maven.compiler.target>
<vaadin.version>14.4.6</vaadin.version>
</properties>
<dependencyManagement>
<dependencies>
<dependency>
<groupId>com.vaadin</groupId>
<artifactId>vaadin-bom</artifactId>
<version>${vaadin.version}</version>
<type>pom</type>
<scope>import</scope>
</dependency>
</dependencies>
</dependencyManagement>
<dependencies>
<dependency>
<groupId>com.vaadin</groupId>
<artifactId>vaadin</artifactId>
<optional>true</optional>
<exclusions>
<!-- Webjars are only needed when running in Vaadin 13 compatibility mode -->
<exclusion>
<groupId>com.vaadin.webjar</groupId>
<artifactId>*</artifactId>
</exclusion>
<exclusion>
<groupId>org.webjars.bowergithub.insites</groupId>
<artifactId>*</artifactId>
</exclusion>
<exclusion>
<groupId>org.webjars.bowergithub.polymer</groupId>
<artifactId>*</artifactId>
</exclusion>
<exclusion>
<groupId>org.webjars.bowergithub.polymerelements</groupId>
<artifactId>*</artifactId>
</exclusion>
<exclusion>
<groupId>org.webjars.bowergithub.vaadin</groupId>
<artifactId>*</artifactId>
</exclusion>
<exclusion>
<groupId>org.webjars.bowergithub.webcomponents</groupId>
<artifactId>*</artifactId>
</exclusion>
</exclusions>
</dependency>
</dependencies>
<build>
<plugins>
<!-- in order to build Jandex index for Vaadin jars, we need to unpack them first -->
<plugin>
<artifactId>maven-dependency-plugin</artifactId>
<version>3.1.2</version>
<executions>
<execution>
<id>unpack-dependencies</id>
<phase>process-sources</phase>
<goals>
<goal>unpack-dependencies</goal>
</goals>
</execution>
</executions>
<configuration>
<includeGroupIds>com.vaadin</includeGroupIds>
<!-- <outputDirectory>${project.build.directory}/classes</outputDirectory>-->
</configuration>
</plugin>
<plugin>
<groupId>org.jboss.jandex</groupId>
<artifactId>jandex-maven-plugin</artifactId>
<version>1.0.8</version>
<executions>
<execution>
<id>make-index</id>
<goals>
<goal>jandex</goal>
</goals>
<!-- phase is 'process-classes by default' -->
</execution>
</executions>
<configuration>
<fileSets>
<fileSet>
<directory>${project.build.directory}/dependency</directory>
</fileSet>
</fileSets>
</configuration>
</plugin>
</plugins>
</build>
</project>Running mvn clean package on this project will result in an empty Jandex index being produced (the jandex.idx only has 18 bytes).
Running with mvn clean package -X shows that Jandex plugin actually walks over the Vaadin classes, it simply won't include those in the resulting index:
...
[INFO] Indexed com.vaadin.flow.internal.BrowserLiveReload$Backend (0 annotations)
[INFO] Indexed com.vaadin.flow.internal.UsageStatistics$1 (0 annotations)
[INFO] Indexed com.vaadin.flow.internal.StateNode$1 (0 annotations)
[INFO] Indexed com.vaadin.flow.internal.UsageStatistics (0 annotations)
[INFO] Indexed com.vaadin.flow.internal.UsageStatistics$UsageEntry (0 annotations)
[INFO] Indexed com.vaadin.flow.internal.StateNode$FeatureSetKey (0 annotations)
[INFO] Indexed com.vaadin.flow.internal.ResponseWriter (1 annotations)
...
However, uncommenting the <outputDirectory>${project.build.directory}/classes</outputDirectory> will cause the dependency plugin to unpack all classes into target/classes/ causing the Jandex plugin to index the classes correctly: the jandex.idx now has ~600kb. The problem is that all classes are now packaged into the resulting jar file.
I had a quick look at this.
The problem is that the index file is created in the fileset directory target/dependency/META-INF/jandex.idx and not in target/classes/META-INF/jandex.idx where the jar-plugin would pick it up.
I'm not sure how this is supposed to work, but I would have expected all fileset indexes to be merged into one big index file...? @n1hility @stuartwdouglas
Anyway, @mvysny a workaround for you would be to configure jar-plugin to either include target/dependency/META-INF/jandex.idx or to exclude all (Vaadin) classes.
Excellent catch, I have completely missed the target/dependency/META-INF/jandex.idx file. I would also expect just one big index file to be produced.
I went with the other workaround to exclude all classes (or rather only to include the jandex.idx file):
<plugin>
<!-- workaround for https://github.com/wildfly/jandex-maven-plugin/issues/25 -->
<artifactId>maven-jar-plugin</artifactId>
<version>3.2.0</version>
<configuration>
<includes>
<include>**/jandex.idx</include>
</includes>
</configuration>
</plugin>
I just wanted to add a test for the fileSets feature (because I rewrote the Mojo to use actual Java annotations instead of javadoc annotations) and found this issue. So yea, currently, an extra index is created for each fileSet, which I'm not sure is intentional or not. I'd personally also expect a single index for all fileSets.
I'm working on moving the Jandex Maven plugin into the Jandex codebase and I guess we can change this, as part of Jandex 3.0 release, and document as breaking change :-)
I'd personally also expect a single index for all
fileSets
I changed my mind on that: In my current project we (for some time) had an index just for the test branch of a module that was included into the respective test-jar. We switched to beans.xml because we now produce two test-jars (with different content) and so one test index doesn't cut it.
But now, due to quarkusio/quarkus#19030 (comment), we'd actually need to create two distinct test index files (and an adjustment in Quarkus) to fully achieve what we "need".
So I think we need more flexibility: combined index file, n separate index files etc.
Whoah that's an interesting one. Here's my take on it:
- single execution of the
jandexgoal should IMHO produce a single index - if you need to produce multiple indices (for some reason, I don't like the word "indexes" 🤣), you should have multiple executions of the
jandexgoal
WDYT?
Some more config would be required for that actually (being able to configure not just the file name, but also full file path), but that should be doable.
Well, right after I submitted my previous comment, it occured to me that I had not considered mutliple executions! So I think that's a 👍 😉
I've also realized that it won't be that easy to include different index files in different test-jars with standard maven, but that's not the problem of this plugin.
for some reason, I don't like the word "indexes"
Haha, I tried to avoid it as well in my last comment, but I failed in my first one.
FWIW, another 👍 for one execution -> one index.
I also opened a separate but somewhat related enhancement issue wildfly/jandex-maven-plugin#33 in case you feel like to have extra spare time!
Done in #195.