8354522: Clones of DecimalFormat cause interferences when used concurrently #24598

j3graham · 2025-04-11T17:31:08Z

The DigitList class used in DecimalFormat does not reset the data array in its clone method. This can cause interference when clones are used concurrently.

Progress

Change must be properly reviewed (1 review required, with at least 1 Reviewer)
Change must not contain extraneous whitespace
Commit message must refer to an issue

Issue

JDK-8354522: Clones of DecimalFormat cause interferences when used concurrently (Bug - P4)

Reviewers

Justin Lu (@justin-curtis-lu - Committer)
Naoto Sato (@naotoj - Reviewer)

Reviewing

Using git

Checkout this PR locally:
$ git fetch https://git.openjdk.org/jdk.git pull/24598/head:pull/24598
$ git checkout pull/24598

Update a local copy of the PR:
$ git checkout pull/24598
$ git pull https://git.openjdk.org/jdk.git pull/24598/head

Using Skara CLI tools

Checkout this PR locally:
$ git pr checkout 24598

View PR using the GUI difftool:
$ git pr show -t 24598

Using diff file

Download this PR as a diff file:
https://git.openjdk.org/jdk/pull/24598.diff

Using Webrev

Link to Webrev Comment

bridgekeeper · 2025-04-11T17:31:54Z

👋 Welcome back j3graham! A progress list of the required criteria for merging this PR into master will be added to the body of your pull request. There are additional pull request commands available for use with this pull request.

openjdk · 2025-04-11T17:32:46Z

@j3graham This change now passes all automated pre-integration checks.

ℹ️ This project also has non-automated pre-integration requirements. Please see the file CONTRIBUTING.md for details.

After integration, the commit message for the final commit will be:

8354522: Clones of DecimalFormat cause interferences when used concurrently

Reviewed-by: jlu, naoto

You can use pull request commands such as /summary, /contributor and /issue to adjust it as needed.

At the time when this comment was updated there had been 86 new commits pushed to the master branch:

e21387e: 8354365: Opensource few Modal and Full Screen related tests
4c99489: 8354285: Open source Swing tests Batch 3
e9c8986: 8354493: Opensource Several MultiScreen and Insets related tests
... and 83 more: https://git.openjdk.org/jdk/compare/1d7138fe021d64686ac59cddd1bc7c787c610aa9...master

As there are no conflicts, your changes will automatically be rebased on top of these commits when integrating. If you prefer to avoid this automatic rebasing, please check the documentation for the /integrate command for further details.

As you do not have Committer status in this project an existing Committer must agree to sponsor your change. Possible candidates are the reviewers of this PR (@justin-curtis-lu, @naotoj) but any other Committer may sponsor as well.

➡️ To flag this PR as ready for integration with the above commit message, type /integrate in a new comment. (Afterwards, your sponsor types /sponsor in a new comment to perform the integration).

openjdk · 2025-04-11T17:33:06Z

@j3graham The following label will be automatically applied to this pull request:

core-libs

When this pull request is ready to be reviewed, an "RFR" email will be sent to the corresponding mailing list. If you would like to change these labels, use the /label pull request command.

j3graham · 2025-04-11T17:33:08Z

Reproducing code example:

public class DecimalFormatTest {
    static AtomicInteger mismatchCount = new AtomicInteger(0);

    public static void main(String[] args) {
        DecimalFormat df = new DecimalFormat("#");
        String str = df.format(Math.PI); // initial use of formatter
        System.out.println(str);

        try (var ex = Executors.newThreadPerTaskExecutor(Thread.ofPlatform().factory())) {
            for (int i = 0; i < 50; i++) {
                // each thread gets its own clone of df
                DecimalFormat threadDf = (DecimalFormat) df.clone();
                ex.execute(makeTask(threadDf));
            }
        }

        System.out.println("mismatchCount = " + mismatchCount);
    }

    private static Runnable makeTask(DecimalFormat threadDf) {
        return () -> {
            for (int i = 0; i < 1000000; i++) {
                String dfString = threadDf.format(BigDecimal.valueOf(i));
                String str = String.valueOf(i);
                if (!str.equals(dfString)) {
                    System.err.println("mismatch: str = " + str + " dfString = " + dfString);
                    mismatchCount.incrementAndGet();
                }
            }
        };
    }
}

mlbridge · 2025-04-15T14:15:14Z

Webrevs

AlanBateman · 2025-04-15T14:19:36Z

Are you going to turn the reproducer into a test? I don't know if we have tests that for DecimalFormat cloning.

j3graham · 2025-04-15T14:56:10Z

I didn't notice any existing tests - I've added a version of the reproducer.

justin-curtis-lu

Thanks for the fix, it looks good. Left some minor comments. I will run tiers 1-3 with your change.

justin-curtis-lu · 2025-04-15T16:45:14Z

src/java.base/share/classes/java/text/DigitList.java

@@ -725,6 +725,7 @@ public Object clone() {
            char[] newDigits = new char[digits.length];
            System.arraycopy(digits, 0, newDigits, 0, digits.length);
            other.digits = newDigits;
+            other.data = null;


We don't have to copy the data array unlike the digits array (above) because it can just be reset during the next getDataChars call, but a comment as to why might be helpful.

justin-curtis-lu · 2025-04-15T16:48:32Z

test/jdk/java/text/Format/DecimalFormat/CloneTest.java

+                // each thread gets its own clone of df
+                DecimalFormat threadDf = (DecimalFormat) df.clone();
+                Runnable task = () -> {
+                    for (int j = 0; j < 1000000; j++) {


We should probably make this test less costly by decreasing some values, I don't the bug requires that many iterations to be exposed. (Without the fix and the break statement in the test, mismatchCount gets up into the tens of thousands on my machine.)

For the original reproducer I wanted to be really sure that it failed. I've reduced it to be more reasonable. With the current config it still gets 10-100 mismatches for me. I've limited the number of lines that get logged as well.

If a mismatch is found, I think we can stop there and report error

My thinking was that for a test pass, it would never hit the shortcut anyway, so it’s only an optimization for the failure case, which hopefully won’t happen anymore. Seeing the count of all collisions gives some reassurance that the test is still running for long enough to be effective.

This is a regression test. Detecting the regression and keeping doing the same test is unnecessary.

I’ve updated the test to exit early.

justin-curtis-lu · 2025-04-15T16:49:57Z

test/jdk/java/text/Format/DecimalFormat/CloneTest.java

+/*
+ * @test
+ * @bug 8354522
+ * @summary Check for cloning interference


It will probably be good to mention somewhere that this test/fix addresses the issue of the same data array reference being shared amongst DigitList clones.

I added more detail to the comment with the test method.

justin-curtis-lu · 2025-04-15T17:00:29Z

test/jdk/java/text/Format/DecimalFormat/CloneTest.java

+    public void testCloneIndependence() {
+        AtomicInteger mismatchCount = new AtomicInteger(0);
+        DecimalFormat df = new DecimalFormat("#");
+        String _ = df.format(Math.PI); // initial use of formatter


We should probably comment the importance of this line, as without it the test will pass without the fix. (It sets the data array to a non null value).

naotoj · 2025-04-15T21:23:20Z

test/jdk/java/text/Format/DecimalFormat/CloneTest.java

-                                }
+                            String str = String.valueOf(value);
+                            if (!str.equals(dfString)) {
+                                mismatchCount.getAndIncrement();


mismatchCount is no longer needed. Simply break after printing the error message would suffice.

It would still need something to indicate that it had failed, as well as a way to signal other threads that they should terminate early. Worth changing to an AtomicBoolean?

Ah, ok then. Thanks.

Ok to leave as is? Or is the AtomicBoolean more clear?

Ok to use it as a failure indicator (still I'd suggest breaking immediately after printing error)

naotoj · 2025-04-15T21:26:40Z

test/jdk/java/text/Format/DecimalFormat/CloneTest.java

+                            String str = String.valueOf(value);
+                            if (!str.equals(dfString)) {
+                                mismatchCount.getAndIncrement();
+                                System.err.println("mismatch: str = " + str + " dfString = " + dfString);
                            }
                        }
                    } catch (InterruptedException e) {


I'd suggest throwing the exception (or RuntimeException with it as the cause), not swallowing it silently in the test.

naotoj · 2025-04-15T23:43:15Z

I just wonder if we could do a white testing, ie, df.digitList.data != df.clone().digitList.data after df.format(Math.PI)

j3graham · 2025-04-16T15:56:20Z

I've added a reflection-based test to see how it looks. We could go with either test, or both.

naotoj

Thanks for adding the test. Let's keep them both.

naotoj · 2025-04-16T16:34:40Z

test/jdk/java/text/Format/DecimalFormat/CloneTest.java

+    static void setup() throws Exception {
+        DIGIT_LIST_FIELD = DecimalFormat.class.getDeclaredField("digitList");
+        DIGIT_LIST_FIELD.setAccessible(true);
+
+        DecimalFormat df = new DecimalFormat();
+        Object digitList = DIGIT_LIST_FIELD.get(df);
+
+        DIGIT_LIST_CLASS = digitList.getClass();
+    }


I'd add some comment saying that this is a whitebox testing, so it might fail if the internal implementation changes (quite unlikely though). And if accessing digitList/data fails, it should throw jtreg.SkippedException instead of a real exception in the actual @test methods.

justin-curtis-lu

Thanks for the changes, the new test looks good to me.

naotoj

Looks good to me. Thanks for fixing this.

j3graham · 2025-04-17T13:14:33Z

Thank you all for the reviews.

j3graham · 2025-04-17T13:15:21Z

/integrate

openjdk · 2025-04-17T13:16:29Z

@j3graham
Your change (at version 6c23e27) is now ready to be sponsored by a Committer.

naotoj · 2025-04-17T16:21:10Z

/sponsor

openjdk · 2025-04-17T16:22:05Z

Going to push as commit 04c32fc08a67eaf6d3f47a0f9ea3d620b7ec6a07.
Since your change was applied there have been 93 commits pushed to the master branch:

5660922: 8353070: Clean up and open source couple AWT Graphics related tests (Part 1)
e639cd6: 8353213: Open source several swing tests batch3
d1d81dd: 8353953: con/sun/jdi tests should be fixed to not always require includevirtualthreads=y
... and 90 more: https://git.openjdk.org/jdk/compare/1d7138fe021d64686ac59cddd1bc7c787c610aa9...master

Your commit was automatically rebased without conflicts.

openjdk · 2025-04-17T16:22:24Z

@naotoj @j3graham Pushed as commit 04c32fc.

💡 You may see a message that your pull request was closed with unmerged commits. This can be safely ignored.

reset "data" field in clone

Loading
Loading status checks…

7ef4cb6

openjdk bot added the core-libs label Apr 11, 2025

j3graham changed the title ~~Clones of DecimalFormat can share state~~ 8354522: Clones of DecimalFormat can share state Apr 15, 2025

j3graham changed the title ~~8354522: Clones of DecimalFormat can share state~~ 8354522: Clones of DecimalFormat cause interferences when used concurrently Apr 15, 2025

openjdk bot added the rfr label Apr 15, 2025

j3graham added 2 commits April 15, 2025 10:39

add test

Loading
Loading status checks…

1f39f7d

fix test summary

Loading
Loading status checks…

880b2d8

justin-curtis-lu reviewed Apr 15, 2025

View reviewed changes

j3graham added 2 commits April 15, 2025 14:32

Tune the test to shorten run-time, update comments

Loading
Loading status checks…

6133264

Terminate test early on failure

Loading
Loading status checks…

dc5e466

naotoj reviewed Apr 15, 2025

View reviewed changes

j3graham added 2 commits April 15, 2025 17:44

throw on InterruptedException

Loading
Loading status checks…

55b6183

break after mismatch

Loading
Loading status checks…

abda88e

reflective test

Loading
Loading status checks…

87e652b

naotoj reviewed Apr 16, 2025

View reviewed changes

j3graham added 2 commits April 16, 2025 13:33

use SkippedException if reflective access fails

Loading
Loading status checks…

08a32e1

fix catch block

Loading
Loading status checks…

6c23e27

justin-curtis-lu approved these changes Apr 16, 2025

View reviewed changes

naotoj approved these changes Apr 16, 2025

View reviewed changes

openjdk bot added the ready label Apr 16, 2025

openjdk bot added the sponsor label Apr 17, 2025

openjdk bot added the integrated label Apr 17, 2025

openjdk bot closed this Apr 17, 2025

openjdk bot removed ready rfr sponsor labels Apr 17, 2025

8354522: Clones of DecimalFormat cause interferences when used concurrently #24598

8354522: Clones of DecimalFormat cause interferences when used concurrently #24598

Conversation

j3graham commented Apr 11, 2025 • edited by openjdk bot Loading

Progress

Issue

Reviewers

Reviewing

bridgekeeper bot commented Apr 11, 2025

openjdk bot commented Apr 11, 2025 • edited Loading

openjdk bot commented Apr 11, 2025

j3graham commented Apr 11, 2025

mlbridge bot commented Apr 15, 2025 • edited Loading

Webrevs

AlanBateman commented Apr 15, 2025

j3graham commented Apr 15, 2025

justin-curtis-lu left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

j3graham Apr 15, 2025 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

j3graham Apr 15, 2025 • edited Loading

Choose a reason for hiding this comment

naotoj Apr 15, 2025 • edited Loading

Choose a reason for hiding this comment

naotoj Apr 15, 2025 • edited Loading

Choose a reason for hiding this comment

naotoj commented Apr 15, 2025

j3graham commented Apr 16, 2025

naotoj left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

justin-curtis-lu left a comment

Choose a reason for hiding this comment

naotoj left a comment

Choose a reason for hiding this comment

j3graham commented Apr 17, 2025 • edited Loading

j3graham commented Apr 17, 2025

openjdk bot commented Apr 17, 2025

naotoj commented Apr 17, 2025

openjdk bot commented Apr 17, 2025

openjdk bot commented Apr 17, 2025

j3graham commented Apr 11, 2025 •

edited by openjdk bot

Loading

openjdk bot commented Apr 11, 2025 •

edited

Loading

mlbridge bot commented Apr 15, 2025 •

edited

Loading

j3graham Apr 15, 2025 •

edited

Loading

j3graham Apr 15, 2025 •

edited

Loading

naotoj Apr 15, 2025 •

edited

Loading

naotoj Apr 15, 2025 •

edited

Loading

j3graham commented Apr 17, 2025 •

edited

Loading