RFR(XXS): 8176505: Wrong assertion 'should be an array copy/clone' in arraycopynode.cpp

Volker Simonis volker.simonis at gmail.com
Fri Mar 10 13:58:20 UTC 2017


first of all I think this isn't really a high prio bug because the
assertion is wrong and there's no impact for product builds. On the
other hand I think a lot of test running on debug builds may fail
because of this issue so probably better to fix this in 9 as well (I
actually wonder that you haven't seen this issue until now :)


Here are the gory details: the following assertion in
arraycopynode.cpp is wrong because src_type can be a plain Object:

const TypeAryPtr* ary_src = src_type->isa_aryptr();
assert(ary_src != NULL, "should be an array copy/clone");

This can be easily demonstrated with the following trivial test program:

public class ArraySrcType {
  public static boolean crash(Object src) {
    String[] dst = new String[1];
    System.arraycopy(src, 0, dst, 0, 1);
    return dst[0] == null;
  public static void main(String[] args) {
    String[] sa = new String[1];
    for (int i = 0; i < 20_000; i++) {

Running it with a slow- or fastdebug build results in the following crash:

# To suppress the following error report, specify this argument
# after -XX: or in .hotspotrc: SuppressErrorAt=/arraycopynode.cpp:228
# A fatal error has been detected by the Java Runtime Environment:
# Internal Error
pid=34335, tid=34410
# assert(ary_src != __null) failed: should be an array copy/clone

Notice that the assertion is unnecessary anyway, because some lines
later int he code we explicitely check for ary_src being NULL and bail
out if that 's the case:

    if (ary_src == NULL || ary_src->klass() == NULL ||
        ary_dest == NULL || ary_dest->klass() == NULL) {
      // We don't know if arguments are arrays
      return false;

Unfortunately, I still need a sponsor :(

Thank you and best regards,

More information about the hotspot-compiler-dev mailing list