Support for JAR files with ResourceBundles as .properties at their root

[zakkak]

Fixes: gh-3535
Closes: quarkusio/quarkus#18305

[olpaw]

@zakkak thanks for suggesting a fix. To solve the problem correctly we need a different fix.

What we actually need is:

diff --git a/substratevm/src/com.oracle.svm.hosted.jdk11/src/com/oracle/svm/hosted/jdk11/ClassLoaderSupportImplJDK11OrLater.java b/substratevm/src/com.oracle.svm.hosted.jdk11/src/com/oracle/svm/hosted/jdk11/ClassLoaderSupportImplJDK11OrLater.java
index df9556d6504..ca0012e99af 100644
--- a/substratevm/src/com.oracle.svm.hosted.jdk11/src/com/oracle/svm/hosted/jdk11/ClassLoaderSupportImplJDK11OrLater.java
+++ b/substratevm/src/com.oracle.svm.hosted.jdk11/src/com/oracle/svm/hosted/jdk11/ClassLoaderSupportImplJDK11OrLater.java
@@ -80,9 +80,6 @@ public final class ClassLoaderSupportImplJDK11OrLater extends ClassLoaderSupport
             bundleName = specParts[0];
         }
         String packageName = packageName(bundleName);
-        if (packageName == null) {
-            throw new MissingResourceException("ResourceBundle does not seem to be a fully qualified class name.", bundleName, locale.toLanguageTag());
-        }
         Set<Module> modules;
         if (moduleName != null) {
             modules = classLoaderSupport.findModule(moduleName).stream().collect(Collectors.toSet());
@@ -107,8 +104,7 @@ public final class ClassLoaderSupportImplJDK11OrLater extends ClassLoaderSupport
     private static String packageName(String bundleName) {
         int classSep = bundleName.replace('/', '.').lastIndexOf('.');
         if (classSep == -1) {
-            /* The bundle is not specified via a java.class or java.properties format. */
-            return null;
+            return ""; /* unnamed package */
         }
         return bundleName.substring(0, classSep);
     }

This will make your reproducer from #3535 work but in addition will be able to correctly open up modules that have bundles in their unnamed package. I will create a PR and add you as a reviewer.

[zakkak]

Closing in favor of #3540