Alan Stevens wrote:

>A spec question from iSeries -
>
>JVMTI Several of the Methods functions cannot return any data if the input
>jmethodID identifies a native method. The spec does not indicate what
>return code should be returned in this case. The methods are:

Below is what the RI does (which is, arguably, correct) --

> Get Line Number Table

JVMTI_ERROR_ABSENT_INFORMATION

> Get Method Location

Start of -1 (even though it also has JVMTI_ERROR_ABSENT_INFORMATION)

> Get Local Variable Table


JVMTI_ERROR_ABSENT_INFORMATION

> Get Bytecodes

Zero length array (I think, hopefully not the stub)

At the very minimum the returns for native methods need to be clearly specified. And maybe some of this
should change. That there is an absent information return on GetMethodLocation seems goofy esp given
none on GetBytecodes. I will open a bug. Do you or the iSeries team have a feeling about what these
should be?


>Is the VM expected to return JVMTI_ERROR_INVALID_METHODID (23) ?

Invalid method ID is only when it is a bad method ID (eg NULL). That probably should be clarified too.

###@###.### 2004-03-04

The full set of JVMTI functions whose behavior is unspecified when passed a native method:
GetLineNumberTable
GetLocalVariableTable
GetMaxLocals
GetArgumentSize
GetMethodLocation
GetBytecodes
###@###.### 2004-03-12