Обсуждение: BUG #10847: Connection.setSavepoint()/releaseSavepoint() is not thread-safe

The following bug has been logged on the website:

Bug reference:      10847
Logged by:          Christian Schlichtherle
Email address:      christian@schlichtherle.de
PostgreSQL version: 9.3.4
Operating system:   OS X 10.9.3
Description:

I am not sure if this is a bug or a feature. One might argue that you
shouldn't share a connection between threads. Anyway, following is a
standalone test to reproduce the issue in PostgreSQL 9.3.4 using the JDBC
driver version 9.3-1101-jdbc41:

<pre><code>
package cpssd.db;

import org.junit.Test;

import java.sql.Connection;
import java.sql.DriverManager;
import java.sql.SQLException;
import java.sql.Savepoint;
import java.util.concurrent.CountDownLatch;

/** @author Christian Schlichtherle */
public class Ticket251IT {

    private static final String CONNECTION_STRING =
"jdbc:postgresql:postgres";
    private static final int NUM_THREADS = 2;

    @Test public void foo() throws SQLException, InterruptedException {
        try (Connection c = DriverManager.getConnection(CONNECTION_STRING))
{
            c.setAutoCommit(false);
            final Runnable task = new Runnable() {
                final CountDownLatch startSignal = new
CountDownLatch(NUM_THREADS);

                @Override public void run() {
                    try {
                        startSignal.countDown();
                        startSignal.await();
                        // FIXME: This idiom doesn't work on a shared
connection!
                        Savepoint sp = c.setSavepoint();
                        try {
                            // Insert transaction script here...
                        } finally {
                            c.releaseSavepoint(sp);
                        }
                    } catch (SQLException | InterruptedException e) {
                        e.printStackTrace();
                    }
                }
            };
            final Thread[] threads = new Thread[NUM_THREADS];
            for (int i = 0; i < threads.length; i++)
                (threads[i] = new Thread(task)).start();
            for (Thread thread : threads)
                thread.join();
        },
    }
}
</code></pre>

When run, this code frequently outputs the following stack trace:

<pre><code>
org.postgresql.util.PSQLException: ERROR: no such savepoint
    at
org.postgresql.core.v3.QueryExecutorImpl.receiveErrorResponse(QueryExecutorImpl.java:2161)
    at
org.postgresql.core.v3.QueryExecutorImpl.processResults(QueryExecutorImpl.java:1890)
    at
org.postgresql.core.v3.QueryExecutorImpl.execute(QueryExecutorImpl.java:255)
    at
org.postgresql.jdbc2.AbstractJdbc2Statement.execute(AbstractJdbc2Statement.java:559)
    at
org.postgresql.jdbc2.AbstractJdbc2Statement.executeWithFlags(AbstractJdbc2Statement.java:403)
    at
org.postgresql.jdbc2.AbstractJdbc2Connection.execSQLUpdate(AbstractJdbc2Connection.java:376)
    at
org.postgresql.jdbc3.AbstractJdbc3Connection.releaseSavepoint(AbstractJdbc3Connection.java:192)
    at cpssd.db.Ticket251IT$1.run(Ticket251IT.java:32)
    at java.lang.Thread.run(Thread.java:745)
</code></pre>

I think you need to report this to the JDBC mailing list at
pgsql-jdbc@postgresql.org

http://www.postgresql.org/list/pgsql-jdbc/

The Postgres libpq interface is most definitely not threadsafe but I
don't see any indication in that stack trace that it's using libpq. It
seems to be pure Java there.

On 7/3/2014 11:54 AM, Greg Stark wrote:
> I think you need to report this to the JDBC mailing list at
> pgsql-jdbc@postgresql.org
>
> http://www.postgresql.org/list/pgsql-jdbc/
>
> The Postgres libpq interface is most definitely not threadsafe but I
> don't see any indication in that stack trace that it's using libpq. It
> seems to be pure Java there.

yes, the jdbc driver is pure java.  you still should avoid using a jdbc
sql connection from more than one thread on the same transaction.  its
fine to reuse a connection between transactions as long as only one
thread is using it at a time.



--
john r pierce                                      37N 122W
somewhere on the middle of the left coast