Repost: RFR (S) JDK-6311046: -Xcheck:jni should support checking of GetPrimitiveArrayCritical

[Daniel D. Daugherty]

On 6/5/14 9:56 AM, Zhengyu Gu wrote:
> src/share/vm/runtime/os.cpp
>> old line 743: if (memblock == NULL) {
>>     old line 744:   return malloc(size, memflags, (caller == 0 ? 
>> CALLER_PC : caller));
>>     new line 643: if (memblock == NULL) {
>>     new line 644:   return os::malloc(size, memflags, (caller == 0 ? 
>> CALLER_PC : caller));
>>         Looks like a case where NMT tracking was missing an
>>         allocation in non-ASSERT bits if os::realloc() was called
>>         with a NULL memblock parameter.
>>
>>         Nice catch!
>>
>>     old line 754: void* ptr = malloc(size, memflags, caller == 0 ? 
>> CALLER_PC : caller);
>>     new line 656: void* ptr = os::malloc(size, memflags, caller == 0 
>> ? CALLER_PC : caller);
>>     old line 764: free(memblock);
>>     new line 669: os::free(memblock);
>>         And again. You should run these by Zhengyu so he can
>>         sanity check that this wasn't intentional and make sure
>>         that NMT2 has the same fix.
>>
> It is not a problem, just not as clear as new ones. This part of code 
> is in os::realloc(), where os::malloc() takes precedence over 
> ::malloc(), besides ::malloc() does not have compatible function 
> signature.

Thanks for clearing up my mistaken mumbling about which
malloc() was being called. Not enough coffee on board
when I wrote that review comment... :-)

Dan


>
>
> -Zhengyu.