RFR (S) 8038633: crash in VM_Version::get_processor_features() on startup

[Vladimir Kozlov]

I did not want to touch that old code but it has to be consistent. Done:

src/share/vm/prims/jni.cpp	Mon Mar 31 13:06:21 2014 -0700
@@ -5208,7 +5208,7 @@
      }

  #ifndef PRODUCT
-  #ifndef TARGET_OS_FAMILY_windows
+  #ifndef CALL_TEST_FUNC_WITH_WRAPPER_IF_NEEDED
      #define CALL_TEST_FUNC_WITH_WRAPPER_IF_NEEDED(f) f()
    #endif

Thanks,
Vladimir

On 3/31/14 1:02 PM, Igor Ignatyev wrote:
> Vladimir,
>
> Can you please also replace it in share/vm/prims/jni.cpp:4005?
>
> Thanks,
> Igor
>
> On 03/31/2014 11:40 PM, Vladimir Kozlov wrote:
>> http://cr.openjdk.java.net/~kvn/8038633/webrev.01/
>>
>> I implemented Igor's suggestion and also used these changes to do
>> renaming from getPsrInfo_* to get_cpu_info_*.
>>
>> Thanks,
>> Vladimir
>>
>> On 3/29/14 9:16 AM, Vladimir Kozlov wrote:
>>> On 3/29/14 3:38 AM, Igor Ignatyev wrote:
>>>> Vladimir,
>>>>
>>>>> + #ifndef TARGET_OS_FAMILY_windows
>>>>> +   #define CALL_TEST_FUNC_WITH_WRAPPER_IF_NEEDED(f) f()
>>>>> + #endif
>>>> I think it's better to use '#ifndef
>>>> CALL_TEST_FUNC_WITH_WRAPPER_IF_NEEDED' instead of '#ifndef
>>>> TARGET_OS_FAMILY_windows', isn't it?
>>>
>>> You are right, it is more general. I will try it and do testing.
>>>
>>> Thanks,
>>> Vladimir
>>>
>>>>
>>>> Igor
>>>>
>>>> On 03/29/2014 05:28 AM, Vladimir Kozlov wrote:
>>>>> http://cr.openjdk.java.net/~kvn/8038633/webrev/
>>>>> https://bugs.openjdk.java.net/browse/JDK-8038633
>>>>>
>>>>> This is a bug tail of 8037226 changes.
>>>>>
>>>>> Windows need an exception wrapper around getPsrInfo_stub() call in
>>>>> order
>>>>> to properly handle SEGV for YMM registers test.
>>>>> The fix is based on code we already have in jni.cpp for
>>>>> test_error_handler().
>>>>>
>>>>> Tested on machines which show the problem.
>>>>>
>>>>> It will be backported into jdk8u20 which also have 8037226 changes.
>>>>>
>>>>> Thanks,
>>>>> Vladimir