Maplesoft Open Maple Java Programming Interface : Bug Report 2

Subject

Native crash (SIGSEGV) in libmaple.dylib when evaluating symbolic sum with indexed coefficients (B[n]) — macOS Apple Silicon — Maple 2024.2

Hello Maplesoft Support team,

I’m reporting a native crash (SIGSEGV) inside libmaple.dylib that occurs when evaluating a symbolic sum containing indexed symbolic coefficients (for example B[n]) via the Maple Java Engine API (jopenmaple). The crash aborts the JVM and produces hs_err_pid34602.log. This is available: email me to receive it.. Rewriting the indexed symbol as a function (for example B(n)) avoids the crash, which suggests a native bug in Maple’s handling of indexed symbolic objects during summation/conversion/evaluation.

Please find environment, exact commands, a minimal reproducer, hs_err_pid34602.log (email), and suggested tests and actions below.

Environment

* Maple: Maple 2024.2 (build used in our tests) Build ID #18723373

* OS: macOS (Apple Silicon / arm64) Sequoia 15.7.1 32 GB
    * uname output from the failing machine: Darwin Mac-Mini.local 24.6.0 Darwin Kernel Version 24.6.0: Mon Aug 11 21:16:30 PDT 2025; root:xnu-11417.140.69.701.11~1/RELEASE_ARM64_T8132 arm64  

JDK: openjdk 21.0.8 2025-07-15 LTS
OpenJDK Runtime Environment Temurin-21.0.8+9 (build 21.0.8+9-LTS)
OpenJDK 64-Bit Server VM Temurin-21.0.8+9 (build 21.0.8+9-LTS, mixed mode, sharing)

Maple Java jars in use:
    * /Library/Frameworks/Maple.framework/Versions/2024/java/jopenmaple.jar
    * /Library/Frameworks/Maple.framework/Versions/2024/java/externalcall.jar

Native libraries loaded from:
    * /Library/Frameworks/Maple.framework/Versions/2024/bin.APPLE_ARM64_MACOS

Exact java command used to run MinimalSumTest2:

java -Xcheck:jni --enable-native-access=ALL-UNNAMED -Xrs \  
  -cp .:/Library/Frameworks/Maple.framework/Versions/2024/java/jopenmaple.jar:/Library/Frameworks/Maple.framework/Versions/2024/java/externalcall.jar \  
  -Djava.library.path=/Library/Frameworks/Maple.framework/Versions/2024/bin.APPLE_ARM64_MACOS \  
  MinimalSumTest2  

Note: Adding -Xss100M did not make any apparent difference.

Minimal Java reproducer (attach as MinimalSumTest2.java):

// MinimalSumTest2.java

public class MinimalSumTest2
{
    public static void main(String[] args)
    {
        String[] mapleArgs = new String[] { "java" }; // some versions require this
        com.maplesoft.openmaple.Engine engine = null;
        try
        {
            System.out.println("Creating Engine...");
            engine = new com.maplesoft.openmaple.Engine(
                mapleArgs,
                new com.maplesoft.openmaple.EngineCallBacksDefault(),
                null,
                null
            );

            // assign with ":" to suppress printing (should avoid large marshalled result)
            String assignSuppressed = "u_test := sum(B[n]*sin(n*Pi*x/L)*exp(-alpha*(n*Pi/L)^2*t), n=1..100):";
            System.out.println("Evaluating (suppressed): " + assignSuppressed);
            engine.evaluate(assignSuppressed);
            System.out.println("Assigned (suppressed) OK");

            System.out.println("Done.");
        } catch (Throwable t)
        {
            System.err.println("Unexpected throwable (may be native crash):");
            t.printStackTrace();
            // If a SIGSEGV occurs, JVM may abort before this runs; look for hs_err_pid*.log
        } finally
        {
            // no stop() method available
        }
    }
}

Observed Behavior

* JVM aborts with SIGSEGV in native code (libmaple.dylib). The hs_err log shows frames in:
    * libmaple.dylib : errorJmp(_ThreadLocalData*) / extIsNumeric
    * libjopenmaple.jnilib : newAlgebraic / Java_com_maplesoft_openmaple_Engine_evaluate

* Using B(n) (function form) instead of B[n] or substituting numeric coefficients avoids the crash.

Attached crash evidence

* hs_err_pid34602.log is available: email me to receive it.
* Please use the full hs_err file for diagnostics; it contains native frames, thread state and loaded libraries.

Suggested quick tests to run (please include pass/fail for each):

1. Replace indexed form with function form:
    * B := n -> b^n: u := sum(B(n)*sin(...), n=1..100): — does this crash?
2. Reduce N (small upper bound):
    * sum(B[n]*sin(...), n=1..10): — crash or not?
3. Use a table in place of indexed name: Btab := table():  
4. for k from 1 to 100 do Btab[k] := b || k od:  
5. sum(Btab[n]*sin(...), n=1..100):  
6. Pre-substitute a simple expression for B[n] before summing:
    * sum(subs(B[n]=1/n, B[n]*sin(...)), n=1..100):
7. Try same reproducer on a non-Apple-Silicon mac or Linux (if available) and with JDK 17.

Suggested cause (based on hs_err frames):

* Crash appears to occur in native routines that check numeric-ness / build algebraic objects (functions such as extIsNumeric and newAlgebraic). This indicates a bug in Maple’s native handling/conversion of indexed symbolic objects in sum/convert/evaluate code paths.

Temporary workarounds
* Use function form: B := n -> ... and B(n) in sums rather than B[n].
* Substitute definitions for B[n] before summation (client-side or via subs) so that the sum does not carry an indexed symbolic name into the conversion step.
* Build sums client-side (Java/Clojure) or use explicit truncated numeric sums when possible.

Files to attach (recommended)
* hs_err_pid*.log (full file)
* MinimalSumTest.java (the Java reproducer above)
* README_cmd.txt (contains exact javac and java commands used)
* java_version.txt (output of java --version)
* maplesoft_jars_list.txt (paths to jopenmaple.jar and externalcall.jar used)

=========================================
Suggested actions for Maplesoft engineers
=========================================

* Email me to receive the error log file produced for the crash.
* Reproduce on macOS Apple Silicon using the provided reproducer and command line.
* Inspect handling of indexed name objects inside summation/conversion code paths — particularly code interacting with extIsNumeric, newAlgebraic, and conversion routines invoked by Engine.evaluate.
* Check whether this regression is present on other platforms (x86 mac, Linux) and JDK versions.
* Provide either a patch that prevents this conversion path from dereferencing invalid memory, or return a safe Maple-level error for unsupported symbolic forms.

Thank-you.