RFR (XS) JDK-8051002: Incorrectly merged share/vm/classfile/classFileParser.cpp was pushed to 8u20

Lois Foltan lois.foltan at oracle.com
Thu Aug 28 10:50:46 UTC 2014

On 8/28/2014 6:42 AM, David Holmes wrote:
> Hi Lois,
> On 23/08/2014 4:14 AM, Lois Foltan wrote:
>> Hi Everyone,
>> Please review the following fix targeted specifically for 8u.
>> Webrev:
>>      http://cr.openjdk.java.net/~lfoltan/bug_jdk8051002/
>> Bug: Incorrectly merged share/vm/classfile/classFileParser.cpp was
>> pushed to 8u20
>>      https://bugs.openjdk.java.net/browse/JDK-8051002
>> Summary of fix:
>> Incorrect JDK 8u merge of classfile/classFileParser.cpp resulted in a
>> duplicate call to check the attribute_byte_length within
>> ClassFileParser::parse_classfile_bootstrap_methods_attribute(). The
>> first call did not reflect the fixed condition check and resulted in the
>> failure of
>> test/runtime/classFileParserBug/TestEmptyBootstrapMethodsAttr.java. This
>> change fixes the condition check and removes the duplicate call to
>> guarantee_property.
> From what I recall of this it was the second call site that was the 
> correct one, so the first (with the wrong check) is what should have 
> been removed.

Hi David,

Thanks for pointing that out.  Actually, there is a lot of history here, 
but why I think the 8u merges went awry is because Harold and I both 
changed the same area of code.  I changed the second call site in my 
backport of 8u to fix the condition.  Harold, in a different fix, made a 
change to move the guarantee_property call to the first call site.  So, 
we do want the location of the call to be the first call site but of 
course with the correct condition.  This is what this fix does and this 
matches the code in JDK 9.  Hope that makes sense.


> David
> -----
>> Testing:
>>      vm.quick.testlist, Hotspot's full jtreg tests
>> Thanks,
>> Lois

More information about the hotspot-runtime-dev mailing list