ADDITIONAL SYSTEM INFORMATION :
I've tested it on Linux 6.0 AMD64 and Mac ARM64

A DESCRIPTION OF THE PROBLEM :
While developing a tokenizer and character detection library based on the Vector API, i came across a strange segfault. I nailed it down to the testcase below. If your maschine differs you may increase one the loop counters to trigger the bug. I can trigger it always on two recent laptops.

STEPS TO FOLLOW TO REPRODUCE THE PROBLEM :
If you run the test case it crashes after a certain round with a SIGSEGV if you change the line:

for (int j = 0; j < lit_len; j++) {

into
for (int j = 0; j < 5; j++) {

it will work.

EXPECTED VERSUS ACTUAL BEHAVIOR :
EXPECTED -
I should work always despite the variable.
ACTUAL -
SIGSEGV and jvm crash

---------- BEGIN SOURCE ----------
package org.knownhosts.libfindchars;

import java.net.URISyntaxException;

import org.junit.jupiter.api.Assertions;
import org.junit.jupiter.api.BeforeEach;
import org.junit.jupiter.api.Test;

import jdk.incubator.vector.ByteVector;
import jdk.incubator.vector.Vector;
import jdk.incubator.vector.VectorOperators;

public class TestVectorCrash {

    int lit_len;
    
    @BeforeEach
    void setup() {
     lit_len=5;
    }
    
    
class TestOP {
public Vector<Byte> find(){

var reduced = ByteVector.SPECIES_PREFERRED.broadcast(0L);
int rounds = 1_000;

for (int i = 0; i < rounds; i++) {
var buf = ByteVector.SPECIES_PREFERRED.broadcast(0L);

for (int j = 0; j < lit_len; j++) {
         var vOnlyLiteral = ByteVector.SPECIES_PREFERRED.broadcast(-64L);
         reduced = reduced.add(buf, vOnlyLiteral.compare(VectorOperators.EQ, 0));
}

}
return reduced;
}
}

@Test
public void crashes() throws URISyntaxException, Exception {

var op = new TestOP();
for (int i = 0; i < 2_000; i++) {
System.out.println(i);
var reduced = op.find();
Assertions.assertNotNull(reduced);
}

}

}

---------- END SOURCE ----------

CUSTOMER SUBMITTED WORKAROUND :
non exists to my knowledge

FREQUENCY : always