Limit console output in CharsetsTest to prevent OOM in resultsSum.pl by annaibm · Pull Request #7000 · adoptium/aqa-tests

annaibm · 2026-04-09T21:18:03Z

Without this fix, MBCS_Tests_charsets_0 -> CharsetsTest prints all encoding/decoding errors to stdout, which gets captured into TestTargetResult and TAP files. With 33M+ errors on AIX ppc64, this causes resultsSum.pl to OOM when building the TAP file.

Cap console error output to MAX_CONSOLE_ERRORS_PER_CHARSET (10) per charset
Write all errors to charset_errors.txt for full debug details
Add verbose flag (EXTRA_OPTIONS=-Dcharsets.verbose=true`) for debug runs - prints all errors as default to console but should not be used in CI.

related: https://github.ibm.com/runtimes/automation/issues/921

Without this fix, CharsetsTest prints all encoding/decoding errors to stdout, which gets captured into TestTargetResult and TAP files. With 33M+ errors on AIX ppc64, this causes resultsSum.pl to OOM when building the TAP file. - Cap console error output to MAX_CONSOLE_ERRORS_PER_CHARSET (10) per charset - Write all errors to charset_errors.txt for full debug details - Add verbose flag (-Dcharsets.verbose=true) for debug runs - prints all errors to console but should not be used in CI related: https://github.ibm.com/runtimes/automation/issues/921 Signed-off-by: Anna Babu Palathingal <anna.bp@ibm.com>

annaibm · 2026-04-09T21:23:20Z

Tested:

Grinder #59529: without verbose, no OOM ✅
- charset_errors.txt archived in functional_test_output.tar.gz → Artifactory
Grinder #59531: with EXTRA_OPTIONS: -Dcharsets.verbose=true, OOM reproduced as expected ❌
verbose mode should not be used in CI runs — use only in personal Grinder runs for debugging

Starting Mics Charsets Test (rerbose sode enabled,

Copilot

Pull request overview

Reduces stdout volume from CharsetsTest to prevent downstream TAP/result aggregation (resultsSum.pl) from running out of memory when extremely large numbers of charset mismatches occur (e.g., on AIX ppc64).

Changes:

Caps per-charset console error output (unless -Dcharsets.verbose=true is enabled).
Writes mismatch details to charset_errors.txt under test.output.dir (or . by default).
Adds summary counters for tested/skipped charsets and total errors.

Show a summary per file

File	Description
functional/MBCS_Tests/charsets/src/CharsetsTest.java	Limits console logging of mismatches, adds file logging + verbose mode, and prints a run summary to reduce TAP/result size.

Copilot's findings

Tip

Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comments suppressed due to low confidence (1)

functional/MBCS_Tests/charsets/src/CharsetsTest.java:314

The console message says further errors are "written to charset_errors.txt" even when errorFile could not be created (errorFile == null). Guard this message on errorFile != null or adjust the wording so it’s not misleading.

                    System.out.println(":"+sb0.toString()+"<>"+sb1.toString());
                } else if (charset_err_cnt == MAX_CONSOLE_ERRORS_PER_CHARSET + 1) {
                    System.out.println("  ... further errors for " + ce.charset().name() + " written to charset_errors.txt");
                }

Files reviewed: 1/1 changed files
Comments generated: 5

Copilot · 2026-04-10T04:20:36Z

+    static int tested_cnt = 0;
+    static int skipped_cnt = 0;
+    static boolean verbose = Boolean.getBoolean("charsets.verbose");
+    static int MAX_CONSOLE_ERRORS_PER_CHARSET = 10;


MAX_CONSOLE_ERRORS_PER_CHARSET is a constant but is declared as a mutable static int. To match the existing constant style in this class (e.g., sbcs_space/dbcs_space/mix_space), make it static final (and ideally private) so it can’t be modified at runtime.

Suggested change

static int MAX_CONSOLE_ERRORS_PER_CHARSET = 10;

private static final int MAX_CONSOLE_ERRORS_PER_CHARSET = 10;

Copilot · 2026-04-10T04:20:36Z

                    err_cnt++;
+                    charset_err_cnt++;
+
+                    // Always write to error file
+                    if (errorFile != null) {
+                        errorFile.print(cd.charset().name()+":");
+                        for(byte b : ba) errorFile.printf("\\x%02X",(int)b&0xFF);
+                        errorFile.println(":"+sb0.toString()+"<>"+sb1.toString());
+                    }
+
+                    // Print to console based on verbose flag or error limit
+                    if (verbose || charset_err_cnt <= MAX_CONSOLE_ERRORS_PER_CHARSET) {
+                        if (charset_err_cnt == 1) {
+                            System.out.println("\n=== Decoding Errors in " + cd.charset().name() + " ===");
+                        }
+                        System.out.print("  ");
+                        for(byte b : ba) System.out.printf("\\x%02X",(int)b&0xFF);
+                        System.out.println(":"+sb0.toString()+"<>"+sb1.toString());
+                    } else if (charset_err_cnt == MAX_CONSOLE_ERRORS_PER_CHARSET + 1) {
+                        System.out.println("  ... further errors for " + cd.charset().name() + " written to charset_errors.txt");


charset_err_cnt is shared between decoding and encoding. If a charset has decoding mismatches first, the later encoding mismatches won’t print the "=== Encoding Errors ===" header (and may be suppressed earlier than intended) because charset_err_cnt is already > 1. Consider tracking separate counters (or header flags) for decode vs encode, or resetting before the encoder phase.

Copilot · 2026-04-10T04:20:36Z

+                        for(byte b : ba) System.out.printf("\\x%02X",(int)b&0xFF);
+                        System.out.println(":"+sb0.toString()+"<>"+sb1.toString());
+                    } else if (charset_err_cnt == MAX_CONSOLE_ERRORS_PER_CHARSET + 1) {
+                        System.out.println("  ... further errors for " + cd.charset().name() + " written to charset_errors.txt");


The console message says further errors are "written to charset_errors.txt" even when errorFile could not be created (errorFile == null). Guard this message on errorFile != null or adjust the wording so it’s not misleading.

This issue also appears on line 311 of the same file.

Suggested change

System.out.println(" ... further errors for " + cd.charset().name() + " written to charset_errors.txt");

if (errorFile != null) {

System.out.println(" ... further errors for " + cd.charset().name() + " written to charset_errors.txt");

} else {

System.out.println(" ... further errors for " + cd.charset().name() + " not shown");

}

Copilot · 2026-04-10T04:20:36Z

+        System.out.println("Errors found: " + err_cnt);
+        if (err_cnt > 0) {
+            System.out.println("Full error details written to: charset_errors.txt");
+        }


If charset_errors.txt cannot be created, the summary still prints "Full error details written to: charset_errors.txt". This should be conditional on errorFile being successfully opened (or print the earlier warning again / include the actual path).

Copilot · 2026-04-10T04:20:36Z

+                    // Always write to error file
+                    if (errorFile != null) {
+                        errorFile.print(cd.charset().name()+":");
+                        for(byte b : ba) errorFile.printf("\\x%02X",(int)b&0xFF);
+                        errorFile.println(":"+sb0.toString()+"<>"+sb1.toString());
+                    }


This writes every mismatch to charset_errors.txt unconditionally. With the reported 33M+ errors, this can create a multi-GB file and significantly slow the test or exhaust disk in CI. Consider capping file output by default (similar to console), and only enabling full file output under an explicit debug flag.

…rset -Dcharsets.verbose=true: Uncapped console output (local debugging only) -Dcharsets.debug.file=true: Uncapped file output, limited console (CI-safe debugging) - Update on review comments from copilot

karianna · 2026-04-11T22:40:51Z

-                        System.out.println("  ... further errors for " + cd.charset().name() + " written to charset_errors.txt");
+                    } else if (charset_decode_err_cnt == MAX_CONSOLE_ERRORS_PER_CHARSET + 1) {
+                        if (errorFile != null) {
+                            System.out.println("  ... further decoding errors for " + cd.charset().name() + " written to charset_errors.txt");


Do we want a proper logger as opposed to s.o.p?

As this is a test file, System.out.println is the approach used throughout the existing test suite. Used it this way toward keeping it consistent with the rest of the test file.

We can look at other improvements to these tests in separate PRs.

annaibm · 2026-04-13T19:57:22Z

TESTED RESULTS

Grinders: No flags (default)
https://hyc-runtimes-jenkins.swg-devops.com/job/Grinder/59546/ (no verbose) ⚠️
Console: shows first 10 errors per charset only
File (charset_errors.txt): writes up to 1000 errors per charset Artifactory
Good for normal CI runs — quiet output, limited file size
EXTRA_OPTIONS: -Dcharsets.verbose=true
https://hyc-runtimes-jenkins.swg-devops.com/job/Grinder/59547/ ❌ OOM
Console: shows ALL errors per charset (can be huge)
File: same as default (1000 per charset)
Only use locally when debugging a specific charset
EXTRA_OPTIONS: -Dcharsets.debug.file=true
https://hyc-runtimes-jenkins.swg-devops.com/job/Grinder/59548/ ⚠️
Console: same as default (first 10 only)
File: writes ALL errors with no cap Artifactory
CI-safe — doesn't affect console output, but file can get very large

smlambert · 2026-04-17T10:59:06Z

-                        System.out.println("  ... further errors for " + cd.charset().name() + " written to charset_errors.txt");
+                    } else if (charset_decode_err_cnt == MAX_CONSOLE_ERRORS_PER_CHARSET + 1) {
+                        if (errorFile != null) {
+                            System.out.println("  ... further decoding errors for " + cd.charset().name() + " written to charset_errors.txt");


We can look at other improvements to these tests in separate PRs.

karianna requested review from Copilot and smlambert April 10, 2026 04:17

Copilot started reviewing on behalf of karianna April 10, 2026 04:18 View session

Copilot AI reviewed Apr 10, 2026

View reviewed changes

Console limited to 10 errors/charset, file limited to 1000 errors/cha…

8854fd0

…rset -Dcharsets.verbose=true: Uncapped console output (local debugging only) -Dcharsets.debug.file=true: Uncapped file output, limited console (CI-safe debugging) - Update on review comments from copilot

karianna reviewed Apr 11, 2026

View reviewed changes

smlambert requested a review from pshipton April 13, 2026 17:41

pshipton approved these changes Apr 13, 2026

View reviewed changes

annaibm requested a review from karianna April 16, 2026 17:39

smlambert approved these changes Apr 17, 2026

View reviewed changes

smlambert merged commit 3b2a0eb into adoptium:master Apr 17, 2026
2 checks passed

	static int MAX_CONSOLE_ERRORS_PER_CHARSET = 10;
	private static final int MAX_CONSOLE_ERRORS_PER_CHARSET = 10;

Uh oh!

Conversation

annaibm commented Apr 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

annaibm commented Apr 9, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Copilot's findings

Uh oh!

Copilot AI Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Apr 10, 2026

Choose a reason for hiding this comment

Uh oh!

karianna Apr 11, 2026

Choose a reason for hiding this comment

Uh oh!

annaibm Apr 13, 2026

Choose a reason for hiding this comment

Uh oh!

smlambert Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

annaibm commented Apr 13, 2026

Uh oh!

smlambert Apr 17, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

annaibm commented Apr 9, 2026 •

edited

Loading