[9] RFR: 8177136: Caller sensitive method System.getLogger should specify what happens if there is no caller on the stack.

[Daniel Fuchs]

Hi Jason,

I'm not sure JDK-8145302 is the same problem: I double checked
that I could call Logger.getLogger("foo") and Logger.info("foo")
from JNI with no issue (works both with jdk8 122 and jdk9/dev).

So I'm quite puzzled about what could have caused JDK-8145302.
If Logger.getLogger() is called from a context where
there is no java frame on the stack - which I'm told might
happen with a JNI attached thread - then you will get a NPE
from Logger.getLogger. But because there's a stack trace
displayed in JDK-8145302 this does not look to be such a case.
For lack of information JDK-8145302 has been marked as incomplete.

The issue at hand (8177136) just addresses the case of
System.getLogger (which is a new API): I believe it's important
we specify what happens in the rare corner case where
this method would be called from a context where there is
no java frame on the stack.

If we get more information on JDK-8145302 then we might try
to fix that on an update (or not).

best regards,

-- daniel

On 22/03/2017 19:53, Jason Mehrens wrote:
> Hi Daniel,
>
> For my own understanding, would it be possible and yet still secure to walk up the call stack to substitute a caller in the case of method being defined with the 'native' keyword assuming it is in the same class?
> For instance, if you look at the stacktrace in JDK-8145302 there is a native method calling 'getLogger' with a signature defined in class called MWMCR.
> Then are 2 frames from the MWMCR class and 2 frames from what appears to be anonymous inner classes of MWMCR.
>
> If all we care about is the calling class it seems like there are enough frames to choose from it just that they are not between 'getLogger' and the native method.
>
> Jason
>
> ________________________________________
> From: core-libs-dev <core-libs-dev-bounces at openjdk.java.net> on behalf of Daniel Fuchs <daniel.fuchs at oracle.com>
> Sent: Wednesday, March 22, 2017 1:29 PM
> To: Mandy Chung
> Cc: core-libs-dev
> Subject: Re: [9] RFR: 8177136: Caller sensitive method System.getLogger should specify what happens if there is no caller on the stack.
>
> On 22/03/2017 18:20, Mandy Chung wrote:
>>
>>> On Mar 22, 2017, at 10:34 AM, Daniel Fuchs <daniel.fuchs at oracle.com> wrote:
>>>
>>> http://cr.openjdk.java.net/~dfuchs/webrev_8177136/webrev.04
>>
>> Looks okay.
>>
>> "an {@code IllegalCallerException} is thrown.”
>>    - It might read better if you drop “an” and add a comma instead
>>
>> No need to have a new webrev.
>
> Thanks Mandy,
>
> I updated the webrev in place. I will now proceed with all
> the paperwork to get that fix in 9 ;-)
>
> http://cr.openjdk.java.net/~dfuchs/webrev_8177136/webrev.04
>
> -- daniel
>
>>
>> Thanks
>> Mandy
>>
>