Request for reviews (L): 6840775: Multiple JVM crashes seen with 1.6.0_10 through early access of 1.6.0_14

Vladimir Kozlov Vladimir.Kozlov at Sun.COM
Mon Jun 29 15:29:15 PDT 2009


Tom Rodriguez wrote:
> The change in slow_arraycopy_C seems like overkill but wouldn't a 
> No_Safepoint_Verifier be more appropriate?
> 

Tom,

I will remove the check in slow_arraycopy_C.

After I replaced my code in slow_arraycopy_C with No_Safepoint_Verifier
I hit assert when the callee tries to throw allocatable exception:

   // Check is all offsets and lengths are non negative
   if (src_pos < 0 || dst_pos < 0 || length < 0) {
     THROW(vmSymbols::java_lang_ArrayIndexOutOfBoundsException());
   }

I also can't move No_Safepoint_Verifier into objArrayKlass::copy_array()
since when an element's class does not match the array's class it
will also throw exception:

   THROW(vmSymbols::java_lang_ArrayStoreException());

The copy_array() (and do_copy()) code should be GC safe since it will not copy
after an exception.

Thanks,
Vladimir





More information about the hotspot-compiler-dev mailing list