Picocli extension not compatible with Java 24

[holly-cummins]

Describe the bug

To reproduce, install a starter app from code.quarkus, and then

sdk install java 24-tem
sdk use java 24-tem
quarkus ext add picocli
quarkus dev

It gives:

2025-07-11 12:29:53,629 INFO  [io.quarkus] (Quarkus Main Thread) code-with-quarkus stopped in 0.006s
Exception in thread "executor-thread-1" java.lang.IllegalAccessError: module java.base does not open java.lang to unnamed module @41dacfe1; to use the thread-local-reset capability on Java 24 or later, use this JVM option: --add-opens java.base/java.lang=ALL-UNNAMED
        at org.jboss.threads.JDKSpecific$ThreadAccess.<clinit>(JDKSpecific.java:32)
        at org.jboss.threads.ThreadLocalResettingRunnable.run(ThreadLocalResettingRunnable.java:13)
        at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
        at java.base/java.lang.Thread.run(Thread.java:1447)

Using the suggested mvn quarkus:dev -Djvm.args="--add-opens java.base/java.lang=ALL-UNNAMED" option does work around the issue (for now).

[quarkus-bot]

/cc @ebullient (picocli)

[jerboaa]

We have fixed this in #47637 Have you tried latest master? Or is this a duplicate of #47769 (I'm not sure)?

[holly-cummins]

Ah, thanks @jerboaa. I'd searched issues for picocli, but not for the exact error. Although I'm surprised that I'm still seeing it, since if the fix went in in May, it should be in a release by now.

What got generated on code.quarkus is <quarkus.platform.version>3.24.3</quarkus.platform.version>. I checked and there's no add-opens anywhere in the pom, and grep -r "Add-Opens" * is also not finding anything.

I checked the basic rest application, and it also doesn't have add-opens, but it does manage to run quarkus dev without a fatal error.

[maxandersen]

this happens even when not adding picocli.

Exception in thread "executor-thread-1" java.lang.IllegalAccessError: module java.base does not open java.lang to unnamed module @3da112f3; to use the thread-local-reset capability on Java 24 or later, use this JVM option: --add-opens java.base/java.lang=ALL-UNNAMED
	at org.jboss.threads.JDKSpecific$ThreadAccess.<clinit>(JDKSpecific.java:32)
	at org.jboss.threads.ThreadLocalResettingRunnable.run(ThreadLocalResettingRunnable.java:13)
	at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
	at java.base/java.lang.Thread.run(Thread.java:1447)
Exception in thread "vert.x-internal-blocking-1" java.lang.NoClassDefFoundError: Could not initialize class org.jboss.threads.JDKSpecific$ThreadAccess

	at org.jboss.threads.ThreadLocalResettingRunnable.run(ThreadLocalResettingRunnable.java:13)
	at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30)
	at java.base/java.lang.Thread.run(Thread.java:1447)
Caused by: java.lang.ExceptionInInitializerError: Exception java.lang.IllegalAccessError: module java.base does not open java.lang to unnamed module @3da112f3; to use the thread-local-reset capability on Java 24 or later, use this JVM option: --add-opens java.base/java.lang=ALL-UNNAMED [in thread "executor-thread-1"]
	at org.jboss.threads.JDKSpecific$ThreadAccess.<clinit>(JDKSpecific.java:32)
	... 3 more

just not seeing it before ctrl+c is done.

[holly-cummins]

this happens even when not adding picocli.

For me, quarkus ext rm picocli means that quarkus dev works without error. So picocli is making a big difference, by moving the error from shutdown (annoying) to startup (fatal). I think it's perhaps because the picocli tests are triggering app shutdown? Although I guess tests wouldn't run on quarkus dev without triggering, so maybe it's something else. But either way, with picocli, just doing quarkus dev is enough to see an apparently broken app.

[dmlloyd]

The quarkus command should be adding that flag --add-opens java.base/java.lang=ALL-UNNAMED IMO. If there are any apps not hitting this issue, it's more due to luck than anything.

[jerboaa]

The quarkus command should be adding that flag --add-opens java.base/java.lang=ALL-UNNAMED IMO. If there are any apps not hitting this issue, it's more due to luck than anything.

I believe #47769 is tracking this item (sorry not fluent in quarkus component speak)?

[holly-cummins]

In my earlier comment I'd assumed #47769 was merged, but actually, reading more closely, I see #47769 is an issue, not a PR, and we haven't even started fixing it. I'll add a comment to that issue explaining the impact and close this as a duplicate.