RFR (S) 8042660: vm/mlvm/anonloader/stress/byteMutation failed with: assert(index >=0 && index < _length) failed: symbol index overflow

[harold seigel]

Thank you!

Harold

On 12/23/2015 1:17 PM, Karen Kinnear wrote:
> Harold,
>
> Many thanks for the fix. Looks good. Ship it!
>
> Karen
>
>> On Dec 21, 2015, at 8:57 PM, harold seigel <harold.seigel at oracle.com> wrote:
>>
>> Hi,
>>
>> Please review this updated webrev that includes the check for zero length suggested below.   The only change from the previous webrev is at line 625 of classFileParser.cpp.
>>
>> New webrev: http://cr.openjdk.java.net/~hseigel/bug_8042660.2/
>>
>> The new change was retested with the hotspot jtreg tests and the JCK vm tests.
>>
>> Thanks! Harold
>>
>> On 12/21/2015 4:00 PM, Karen Kinnear wrote:
>>> Harold,
>>>
>>> Looks good for the assertion you hit.
>>> Thank you for the fix and the extensive testing!
>>>
>>> And thank you for walking me privately through -
>>> spec reference to requiring at least one code point which means “” is not a valid name or signature
>>> and having the logic under the >= JAVA_7_VERSION since that is where we added verification that this
>>> was the right type of signature.
>>>
>>> And thank you for adding the test for the empty method name - if I understand correctly, that is testing
>>> existing code in verify_legal_method_name.
>>>
>>> I believe if you were to turn off verification you could
>>> run into the assertion under JVM_CONSTANT_Methodref “bad method name”. Perhaps instead
>>> of an assertion there you could check if name length == 0 or something so we do the init check
>>> properly but don’t assert/crash.
>>>
>>> thanks,
>>> Karen
>>>
>>>> On Dec 21, 2015, at 1:35 PM, harold seigel <harold.seigel at oracle.com> wrote:
>>>>
>>>> Hi,
>>>>
>>>> Please review this small change to fix bug 8042660 <https://bugs.openjdk.java.net/browse/JDK-8042660>.
>>>>
>>>> The code in ClassFileParser::parse_constant_pool() that handles CONSTANT_NameAndType and CONSTANT_*Ref assumes that the constant pool strings they reference are valid names and signatures because they are valid Utf8 strings.  But, "" is a valid Utf8 string but not a valid name or signature.  Assuming that the strings were of non-zero length caused crashes when test byteMutation specified zero length strings.  This fix adds checks for strings of zero length and throws ClassFormatError exceptions, preventing the JVM crashes.
>>>>
>>>> Open webrev: http://cr.openjdk.java.net/~hseigel/bug_8042660/
>>>>
>>>> JBS Bug: https://bugs.openjdk.java.net/browse/JDK-8042660
>>>>
>>>> The fix was tested with JCK Lang, VM, and API java_lang tests, the UTE quick and split verifier tests, and the hotspot, JDK vm, java/io, java/lang, and java/util JTreg tests, and the tests included in this RFR.
>>>>
>>>> Thanks, Harold