Loading extensions conflicts with JVM in debug mode

[juja0]

When extensions (like httpfs) are loaded via duckdb-jdbc driver when the JVM is running in debug mode, it somehow breaks instrumentation agent attachment.

This manifests as a crash when libraries like bytebuddy-agent attempt attaching to the running JVM. Consequently, other libraries and frameworks like mockito and spring that depend on bytebuddy-agent also crash.

Sample code to demonstrate the issue: https://github.com/juja0/duckdb-httpfs-java-debug-issue

Note: The code in the link above needs to be executed in debug mode to trigger the issue.

Observations

• Initially noticed in duckdb_jdbc 1.3.2.0 but also happens in 1.4.4.0
• Unsure if it happens on versions prior to duckdb_jdbc 1.3.2.0
• Happens when calling ByteBuddyAgent.install
• ByteBuddyAgent version doesn't seem to matter. Happens on a fairly old version and also in the latest version.
• Happens only if duckdb loads httpfs extension before installing ByteBuddyAgent. If the order is flipped, it works fine.
• Happens on httpfs and aws extensions but not on ducklake extension
• Happens only when run in debug mode.

[staticlibs]

Hi, thanks for the report! Just I am unable to reproduce this so far. I wonder if this is specific to some JVM version?

alex@localhost:~/projects/duck/samples/duckdb-httpfs-java-debug-issue$ git rev-parse HEAD
555cdb54ef96a1d7b173b87737aefaa6cd9bd6e5

alex@localhost:~/projects/duck/samples/duckdb-httpfs-java-debug-issue$ git status
On branch master
Your branch is up to date with 'origin/master'.
nothing to commit, working tree clean

alex@localhost:~/projects/duck/samples/duckdb-httpfs-java-debug-issue$ $JAVA_HOME/bin/java -version
openjdk version "24.0.1" 2025-04-15
OpenJDK Runtime Environment Corretto-24.0.1.9.1 (build 24.0.1+9-FR)
OpenJDK 64-Bit Server VM Corretto-24.0.1.9.1 (build 24.0.1+9-FR, mixed mode, sharing)

alex@localhost:~/projects/duck/samples/duckdb-httpfs-java-debug-issue$ mvn clean package
WARNING: A restricted method in java.lang.System has been called
WARNING: java.lang.System::load has been called by org.fusesource.jansi.internal.JansiLoader in an unnamed module (file:/usr/share/maven/lib/jansi-2.4.1.jar)
WARNING: Use --enable-native-access=ALL-UNNAMED to avoid a warning for callers in this module
WARNING: Restricted methods will be blocked in a future release unless native access is enabled
WARNING: A terminally deprecated method in sun.misc.Unsafe has been called
WARNING: sun.misc.Unsafe::objectFieldOffset has been called by com.google.common.util.concurrent.AbstractFuture$UnsafeAtomicHelper (file:/usr/share/maven/lib/guava-33.2.1-jre.jar)
WARNING: Please consider reporting this to the maintainers of class com.google.common.util.concurrent.AbstractFuture$UnsafeAtomicHelper
WARNING: sun.misc.Unsafe::objectFieldOffset will be removed in a future release
[INFO] Scanning for projects...
[INFO] 
[INFO] ---------------< com.juja.example:DuckHttpfsDebugIssue >----------------
[INFO] Building DuckHttpfsDebugIssue 1.0-SNAPSHOT
[INFO]   from pom.xml
[INFO] --------------------------------[ jar ]---------------------------------
[INFO] 
[INFO] --- clean:3.2.0:clean (default-clean) @ DuckHttpfsDebugIssue ---
[INFO] 
[INFO] --- resources:3.3.1:resources (default-resources) @ DuckHttpfsDebugIssue ---
[INFO] skip non existing resourceDirectory /home/alex/projects/duck/samples/duckdb-httpfs-java-debug-issue/src/main/resources
[INFO] 
[INFO] --- compiler:3.13.0:compile (default-compile) @ DuckHttpfsDebugIssue ---
[INFO] Recompiling the module because of changed source code.
[INFO] Compiling 1 source file with javac [debug target 22] to target/classes
[WARNING] location of system modules is not set in conjunction with -source 22
  not setting the location of system modules may lead to class files that cannot run on JDK 22
    --release 22 is recommended instead of -source 22 -target 22 because it sets the location of system modules automatically
[INFO] 
[INFO] --- resources:3.3.1:testResources (default-testResources) @ DuckHttpfsDebugIssue ---
[INFO] skip non existing resourceDirectory /home/alex/projects/duck/samples/duckdb-httpfs-java-debug-issue/src/test/resources
[INFO] 
[INFO] --- compiler:3.13.0:testCompile (default-testCompile) @ DuckHttpfsDebugIssue ---
[INFO] No sources to compile
[INFO] 
[INFO] --- surefire:3.2.5:test (default-test) @ DuckHttpfsDebugIssue ---
[INFO] No tests to run.
[INFO] 
[INFO] --- jar:3.4.1:jar (default-jar) @ DuckHttpfsDebugIssue ---
[INFO] Building jar: /home/alex/projects/duck/samples/duckdb-httpfs-java-debug-issue/target/DuckHttpfsDebugIssue-1.0-SNAPSHOT.jar
[INFO] ------------------------------------------------------------------------
[INFO] BUILD SUCCESS
[INFO] ------------------------------------------------------------------------
[INFO] Total time:  2.302 s
[INFO] Finished at: 2026-03-06T12:49:03Z
[INFO] ------------------------------------------------------------------------

alex@localhost:~/projects/duck/samples/duckdb-httpfs-java-debug-issue$ $JAVA_HOME/bin/java -agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=*:5005 -cp ./target/DuckHttpfsDebugIssue-1.0-SNAPSHOT.jar:/home/alex/.m2/repository/org/duckdb/duckdb_jdbc/1.4.4.0/duckdb_jdbc-1.4.4.0.jar:/home/alex/.m2/repository/net/bytebuddy/byte-buddy-agent/1.18.7/byte-buddy-agent-1.18.7.jar DuckHttpfsJavaDebugIssue
Listening for transport dt_socket at address: 5005
WARNING: A restricted method in java.lang.System has been called
WARNING: java.lang.System::load has been called by org.duckdb.DuckDBNative in an unnamed module (file:/home/alex/.m2/repository/org/duckdb/duckdb_jdbc/1.4.4.0/duckdb_jdbc-1.4.4.0.jar)
WARNING: Use --enable-native-access=ALL-UNNAMED to avoid a warning for callers in this module
WARNING: Restricted methods will be blocked in a future release unless native access is enabled
42
WARNING: A Java agent has been loaded dynamically (/home/alex/.m2/repository/net/bytebuddy/byte-buddy-agent/1.18.7/byte-buddy-agent-1.18.7.jar)
WARNING: If a serviceability tool is in use, please run with -XX:+EnableDynamicAgentLoading to hide this warning
WARNING: If a serviceability tool is not in use, please run with -Djdk.instrument.traceUsage for more information
WARNING: Dynamic loading of agents will be disallowed by default in a future release

[juja0]

Thanks for checking. I was originally using eclipse-temurin 22. I just tried again with corretto 24 and the issue still happens.

I was also executing the program via an IDE (intellij) and suspected that that might be playing a role but i was able to rule that out as well by running the exact command you've used above in a terminal and the issue still happens.

Could it be related to the OS/arch it's running on ? I'm trying this on Macos Sequoia (aarch64).

[staticlibs]

@juja0

Thanks for the details! I can see the following on macOS:

alexkasko@Mac duckdb-httpfs-java-debug-issue % $JAVA_HOME/bin/java -version 
openjdk version "25.0.2" 2026-01-20 LTS
OpenJDK Runtime Environment Corretto-25.0.2.10.1 (build 25.0.2+10-LTS)
OpenJDK 64-Bit Server VM Corretto-25.0.2.10.1 (build 25.0.2+10-LTS, mixed mode, sharing)

alexkasko@Mac duckdb-httpfs-java-debug-issue % file $JAVA_HOME/bin/java 
/Volumes/data/apps/amazon-corretto-25.jdk/Contents/Home/bin/java: Mach-O 64-bit executable arm64

alexkasko@Mac duckdb-httpfs-java-debug-issue % $JAVA_HOME/bin/java -cp ./target/DuckHttpfsDebugIssue-1.0-SNAPSHOT.jar:/Users/alexkasko/.m2/repository/org/duckdb/duckdb_jdbc/1.4.4.0/duckdb_jdbc-1.4.4.0.jar:/Users/alexkasko/.m2/repository/net/bytebuddy/byte-buddy-agent/1.18.7/byte-buddy-agent-1.18.7.jar DuckHttpfsJavaDebugIssue

Note: Program needs to be run in debug mode to reproduce issue

WARNING: A restricted method in java.lang.System has been called
WARNING: java.lang.System::load has been called by org.duckdb.DuckDBNative in an unnamed module (file:/Users/alexkasko/.m2/repository/org/duckdb/duckdb_jdbc/1.4.4.0/duckdb_jdbc-1.4.4.0.jar)
WARNING: Use --enable-native-access=ALL-UNNAMED to avoid a warning for callers in this module
WARNING: Restricted methods will be blocked in a future release unless native access is enabled

42
Exception in thread "main" java.lang.IllegalStateException: Could not self-attach to current VM using external process - set a property net.bytebuddy.agent.attacher.dump to dump the process output to a file at the specified location
	at net.bytebuddy.agent.ByteBuddyAgent.installExternal(ByteBuddyAgent.java:685)
	at net.bytebuddy.agent.ByteBuddyAgent.install(ByteBuddyAgent.java:612)
	at net.bytebuddy.agent.ByteBuddyAgent.install(ByteBuddyAgent.java:592)
	at net.bytebuddy.agent.ByteBuddyAgent.install(ByteBuddyAgent.java:544)
	at net.bytebuddy.agent.ByteBuddyAgent.install(ByteBuddyAgent.java:521)
	at DuckHttpfsJavaDebugIssue.main(DuckHttpfsJavaDebugIssue.java:36)

alexkasko@Mac duckdb-httpfs-java-debug-issue % $JAVA_HOME/bin/java '-agentlib:jdwp=transport=dt_socket,server=y,suspend=n,address=*:5005' -cp ./target/DuckHttpfsDebugIssue-1.0-SNAPSHOT.jar:/Users/alexkasko/.m2/repository/org/duckdb/duckdb_jdbc/1.4.4.0/duckdb_jdbc-1.4.4.0.jar:/Users/alexkasko/.m2/repository/net/bytebuddy/byte-buddy-agent/1.18.7/byte-buddy-agent-1.18.7.jar DuckHttpfsJavaDebugIssue
Listening for transport dt_socket at address: 5005
WARNING: A restricted method in java.lang.System has been called
WARNING: java.lang.System::load has been called by org.duckdb.DuckDBNative in an unnamed module (file:/Users/alexkasko/.m2/repository/org/duckdb/duckdb_jdbc/1.4.4.0/duckdb_jdbc-1.4.4.0.jar)
WARNING: Use --enable-native-access=ALL-UNNAMED to avoid a warning for callers in this module
WARNING: Restricted methods will be blocked in a future release unless native access is enabled

42
Exception in thread "main" java.lang.IllegalStateException: Could not self-attach to current VM using external process - set a property net.bytebuddy.agent.attacher.dump to dump the process output to a file at the specified location
	at net.bytebuddy.agent.ByteBuddyAgent.installExternal(ByteBuddyAgent.java:685)
	at net.bytebuddy.agent.ByteBuddyAgent.install(ByteBuddyAgent.java:612)
	at net.bytebuddy.agent.ByteBuddyAgent.install(ByteBuddyAgent.java:592)
	at net.bytebuddy.agent.ByteBuddyAgent.install(ByteBuddyAgent.java:544)
	at net.bytebuddy.agent.ByteBuddyAgent.install(ByteBuddyAgent.java:521)
	at DuckHttpfsJavaDebugIssue.main(DuckHttpfsJavaDebugIssue.java:36)

AFAIU the behaviour is the same with or without JDWP agent. I wonder is this the same symptoms you are observing?

[juja0]

Yes. This is the error I'm getting. But it's weird that it's happening without the debug agent for you. For me it works fine when I'm not debugging. I'm starting to think that the fact that this happens only after loading httpfs is probably a red herring and that the issue is probably somewhere else and the extension load is somehow triggering it.

Looking closer at what bytebuddyagent.install() is doing, I can see that it spawns a second JVM and runs Attacher.class which then tries to attach to the first JVM and in this case it's not able to. I believe it's trying to connect to the first JVM via a unix socket that's supposed to have been created by the first JVM but somehow when this issue happens, either the unix socket is not created or the connectivity has some issues. To test this, I started the first JVM with -XX:+StartAttachListener and that seems to solve the issue even in debug mode for me.

This leads me to believe that loading httpfs extension is somehow preventing the JVM's attach listener from starting on-demand. That's probably why when the attach listener is forced to start early it works fine. Could this somehow be related to the libraries that httpfs links to ? something like openssl ?

[staticlibs]

I'm starting to think that the fact that this happens only after loading httpfs is probably a red herring and that the issue is probably somewhere else and the extension load is somehow triggering it.

I can only guess that some of the linked dependencies is doing something on shared lib load that JVM does not like later. It seems that there is a few circumstances when ByteBuddy cannot attach without -XX:+StartAttachListener (raphw/byte-buddy#953), it looks like loading httpfs hits one of these cases.

[juja0]

Thanks for the link. That sent me down a rabbit hole of potential causes for this issue and I was able to rule out most of them. But I was able to confirm the previous theory that it's related to unix socket connectivity issues.

When running without the debug agent, the first jvm creates a unix socket and the second one is able to connect to it properly. When paused in debug mode (on the second jvm, not the first) I can see that the unix socket exists.

But when running the first jvm in debug mode, the socket does not get created. So it appears that the httpfs extension is somehow preventing the jvm from creating the socket. It looks like the jvm is supposed to create the socket when it receives a sigquit signal. I can see that the second jvm is sending the sigquit but the first jvm just ignores it.

Is it possible that duckdb or one of its subsystems is registering a signal handler and not delegating to the default one ?

[staticlibs]

Is it possible that duckdb or one of its subsystems is registering a signal handler and not delegating to the default one ?

I think some dependency can mess with posix signals, for example cURL surely is doing something with them.

[juja0]

Hmm. How should we proceed with this ? This is not really a blocker for me. Just wanted to highlight the issue in case it's a symptom of some other deeper issue. If you think this issue is not relevant, I'm happy to close it.

[staticlibs]

First, thanks for highlighting the issue and researching the -XX:+StartAttachListener workaround, this is surely useful! While this is not blocking for ByteBuddy, similar problem can happen with other tools/runtimes. So the usage of signals in dependencies needs to be investigated (and signals hijacking to be disabled where possible). I don't have cycles to dig this right now, so lets keep this issue open for the future.

[juja0]

Appears to be fixed in 1.5.0.0