RFR: 8049632: JDK 1.8.0 b132 :Linux x64 : Crash in ClassFileParser::copy_localvariable_table(..)

[serguei.spitsyn at oracle.com]

On 1/29/15 11:26 AM, Coleen Phillimore wrote:
>
> On 1/28/15, 7:33 PM, serguei.spitsyn at oracle.com wrote:
>> Coleen,
>>
>> The fix looks good.
>> It is great, you added new tests for it, thanks!
>
> Thank you for the code review, Serguei!
>
>>
>> I have one question about the tests though.
>> It seems there are no duplicated LVT/LVTT in the tests *DuplicateLVT.cod,
>> **DuplicateLVTT.cod* and *NotFoundLVTT.cod* has LVTT.
>
> In DuplicateLVT.cod, the LVT entry starting with '5' was duplicated.
>
>             Attr(#39) { // LocalVariableTable
>               [] { // LocalVariableTable
>                 0 279 44 45 0;
>                 2 277 46 47 1;
>                 5 274 48 49 2;
>                 5 274 48 49 2;
>                 8 271 50 51 3;
>                 13 266 52 53 4;
>                 17 262 54 55 6;
>                 21 258 56 57 7;
>                 26 253 58 59 8;
>                 30 249 60 61 10;
>                 39 240 62 63 11;
>               }
>
> DuplicateLVTT.cod has:
>
>             Attr(#64) { // LocalVariableTypeTable
>               [] { // LocalVariableTypeTable
>                 39 240 62 65 11;
>                 39 240 62 65 11;
>               }
>
> NotFoundLVTT.cod is more interesting because it has to refer to LVT, 
> so I changed the entry from 39 to 38:
>
>             Attr(#64) { // LocalVariableTypeTable
>               [] { // LocalVariableTypeTable
>                 38 240 62 65 11;
>               }
>

Ok, I see now.

Thanks!
Serguei

> Thanks,
> Coleen
>
>>
>> Do I miss anything?
>>
>> Thanks,
>> Serguei
>>
>>
>> On 1/28/15 8:11 AM, Coleen Phillimore wrote:
>>> Summary: Use resource allocated hashtable for local variable table 
>>> checking
>>>
>>> I was never able to reproduce this after considerable time trying. I 
>>> also never found anything wrong with the code that would cause an 
>>> extra or incorrect os::free call.  This change checks LVT entries 
>>> using the existing ResourceHashtable rather than a ad-hoc extra 
>>> hashtable definition in classFileParser.cpp.  So this change removes 
>>> the os::free call.  I think whatever caused the original bug will 
>>> crash somewhere else now, and hopefully be easier to find.
>>>
>>> Tested with jck9, jck1.4 runThese tests, nsk and jtreg tests, and 
>>> added test.
>>>
>>> open webrev at http://cr.openjdk.java.net/~coleenp/8049632/
>>> bug link https://bugs.openjdk.java.net/browse/JDK-8049632
>>>
>>> Thanks,
>>> Coleen
>>>
>>
>