This is an automated email from the ASF dual-hosted git repository.
hansva pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/hop.git
The following commit(s) were added to refs/heads/main by this push:
new e991658dc3 Force driver load to avoid deadlocks, fixes #5698 (#7829)
e991658dc3 is described below
commit e991658dc37a11087998f6eb97602f13ccd4afc3
Author: Hans Van Akelyen <[email protected]>
AuthorDate: Fri Aug 7 13:59:46 2026 +0200
Force driver load to avoid deadlocks, fixes #5698 (#7829)
---
.../src/main/java/org/apache/hop/core/database/Database.java | 12 ++++++++++++
1 file changed, 12 insertions(+)
diff --git a/core/src/main/java/org/apache/hop/core/database/Database.java
b/core/src/main/java/org/apache/hop/core/database/Database.java
index 0a5d73babd..b64ddc6b05 100644
--- a/core/src/main/java/org/apache/hop/core/database/Database.java
+++ b/core/src/main/java/org/apache/hop/core/database/Database.java
@@ -436,6 +436,18 @@ public class Database implements IVariables,
ILoggingObject, AutoCloseable {
try {
synchronized (DriverManager.class) {
+ // Force the JDBC 4 driver ServiceLoader scan to run here, while we
hold the lock.
+ //
+ // DriverManager only performs that scan once, lazily, the first time
a connection is
+ // opened. It instantiates every java.sql.Driver on the classpath, not
just the one we
+ // need. If that happens in DriverManager.getConnection() below, it
runs unsynchronized
+ // with the driver class loading of another transform's init thread:
both threads then
+ // initialize the same driver class hierarchy from opposite ends and
deadlock on the
+ // class initialization monitors, hanging the pipeline. Doing it here
makes the scan
+ // happen exactly once, single threaded, before any driver class can
be initialized by
+ // another route.
+ DriverManager.getDrivers();
+
ClassLoader classLoader =
PluginRegistry.getInstance().getClassLoader(plugin);
Class<?> driverClass = classLoader.loadClass(classname);
