RFR: 8230398: Remove NULL checks before FREE_C_HEAP_ARRAY

[David Holmes]

Hi Leo,

On 2/09/2019 11:06 pm, Leo Korinth wrote:
> Hi!
> 
> This fixup removes NULL checks before FREE_C_HEAP_ARRAY. This is 
> unnecessary as FREE_C_HEAP_ARRAY does also do this NULL check. This is a 
> follow up to 8227168: Cleanup usage of NEW_C_HEAP_ARRAY.

I don't see such a NULL check except in debug builds:

#define FREE_C_HEAP_ARRAY(type, old) \
   FreeHeap((char*)(old))

void FreeHeap(void* p) {
   os::free(p);
}

void  os::free(void *memblock) {
   NOT_PRODUCT(inc_stat_counter(&num_frees, 1));
#ifdef ASSERT
   if (memblock == NULL) return;
   if ((intptr_t)memblock == (intptr_t)MallocCatchPtr) {
     log_warning(malloc, free)("os::free caught " PTR_FORMAT, 
p2i(memblock));
     breakpoint();
   }
   void* membase = MemTracker::record_free(memblock);
   verify_memory(membase);

   GuardedMemory guarded(membase);
   size_t size = guarded.get_user_size();
   inc_stat_counter(&free_bytes, size);
   membase = guarded.release_for_freeing();
   ::free(membase);
#else
   void* membase = MemTracker::record_free(memblock);
   ::free(membase);
#endif
}

Also this would now count NULL frees (which seems a bug in itself).

> deallocate_counters() in src/hotspot/os/windows/os_perf_windows.cpp does 
> (in addition to NULL the pointer) also set a counter to zero. Although 
> my change "looks" correct, I tried follow the usage of 
> deallocate_counters(), and could not find a call site. Should I create a 
> bug on this, or am I missing something here?

It looks like dead code. People often write clean symmetrical allocation 
and freeing code, but the VM doesn't actually free a lot of stuff at 
termination.

> Bug:
> https://bugs.openjdk.java.net/browse/JDK-8230398
> 
> Webrev:
> http://cr.openjdk.java.net/~lkorinth/8230398
> 
> Testing:
> mach5 tier1-3
> 
> Thanks,
> Leo