RFR: JDK-8051990 - Uninitialised memory in jdk/src/share/native/java/lang/fdlibm/src/k_standard.c

[joe darcy]

Hello,

There is an upstream of sorts for fdlibm, but the code changes extremely 
rarely and I'm doubtful is these sorts of changes would be accepted.

I (still) plan to port fdlibm from C to Java these release (this time 
for sure!), but haven't done so yet.

Cheers,

-Joe

On 2/18/2015 6:05 AM, Mark Sheppard wrote:
>
> Hi Alan,
>
> as I understand it,
> it is a case that the switch statement doesn't have an explicit 
> default case, and as such it is seen that there is the potential
> for exc to return garbage as it is a local (stack) variable.
> code exc.return is not explicitly initialized in that scenario
>
> I have followed the comments 8011989 and set this to wnf
>
> but the fix below seemed the correct thing to do  ... it is a coding 
> style idiom
> as to whether switch statements should have an explicit default case.
>
> regards
> Mark
>
> On 18/02/2015 13:47, Alan Bateman wrote:
>> On 18/02/2015 13:29, Mark Sheppard wrote:
>>> Hi
>>>    please review the small change
>>> http://cr.openjdk.java.net/~msheppar/8051990/webrev/
>>>
>>> to address the parfait issue in
>>> https://bugs.openjdk.java.net/browse/JDK-8051990
>>>
>> Is this a false positive or is there really something calling 
>> __kernel_standard with an invalid type? If there is then setting 
>> errno would be good.
>>
>> The other thing that I wonder about is modifying fdlibm in OpenJDK, 
>> should fixes be sent upstream? I remember at one point there was a 
>> proposal to replace fdlibm with pure java code but I don't think this 
>> has bubbled up to the top of anyone's list yet.
>>
>> -Alan.
>