review request (S) JSR292: SIGSEGV in JNIHandleBlock::release_block(JNIHandleBlock*, Thread*)+0x3c
christian.thalinger at oracle.com
Tue Feb 7 05:00:04 PST 2012
The fix looks good.
2975 // return address and rbp are already in place
2976 __ subptr(rsp, (framesize-4) << LogBytesPerInt); // prolog
Not related to your change, this is pretty useless because it always subtracts zero.
On Feb 7, 2012, at 11:09 AM, Roland Westrelin wrote:
> SIGSEGV happens in a method handle stub an is used to trigger a NPE but because, the method handle stubs are called from the interpreter (or through a c2i), when the thread executes the throw_NullPointerException_at_call_entry, the stack is not properly aligned. Aligning the stack before the call in throw_NullPointerException_at_call_entry fixes the problem.
More information about the hotspot-compiler-dev