A number of test cases failed with a similar pattern. All had `checkNegativeNumberExpression` as their compile task and PhaseMacroExpand::eliminate_macro_nodes at the top of the stack.

A quote from hs_err:

Current CompileTask:
C2: 33026 6290 b java.text.SimpleDateFormat::checkNegativeNumberExpression (127 bytes)

Stack: [0x0000fffe78400000,0x0000fffe78600000], sp=0x0000fffe785fb7f0, free space=2029k
Native frames: (J=compiled Java code, A=aot compiled Java code, j=interpreted, Vv=VM code, C=native code)
V [libjvm.so+0x976140] PhaseMacroExpand::eliminate_macro_nodes()+0x15c
V [libjvm.so+0x5549d0] Compile::Optimize()+0xf40
V [libjvm.so+0x555914] Compile::Compile(ciEnv*, ciMethod*, int, bool, bool, bool, bool, DirectiveSet*)+0xa14
V [libjvm.so+0x4a4be8] C2Compiler::compile_method(ciEnv*, ciMethod*, int, bool, DirectiveSet*)+0x108
V [libjvm.so+0x55cc08] CompileBroker::invoke_compiler_on_method(CompileTask*)+0x788
V [libjvm.so+0x55d628] CompileBroker::compiler_thread_loop()+0x438
V [libjvm.so+0xbdf53c] JavaThread::thread_main_inner()+0xe8
V [libjvm.so+0xbe5da8] Thread::call_run()+0x104
V [libjvm.so+0xa61068] thread_native_entry(Thread*)+0xf8
C [libpthread.so.0+0x7d40] start_thread+0xb4

It seems like the combination of '-Xcomp -XX:-TieredCompilation' raises the probability (all failed tests had it).