D'oh! I blindly clicked on that links of course :)
> Am 04.12.2017 um 17:05 schrieb Attila Szegedi <szege...@gmail.com>:
>
> I know, it’s subtle; that’s why I decided to describe it in detail :-)
>
>> On 2017. Dec 4., at 16:36, Hannes Wallnöfer <hannes.wallnoe...@oracle.com>
>> wrote:
>>
>> Took me some time to spot the difference in behaviour (adaptation of method
>> type in Binder)!
>>
>> +1
>>
>> Hannes
>>
>>> Am 04.12.2017 um 13:40 schrieb Attila Szegedi <szege...@gmail.com>:
>>>
>>> Please review JDK-8192970 "Element getters/setters with fixed key fail to
>>> link properly" at <http://cr.openjdk.java.net/~attila/8192970/webrev.jdk>
>>> for <https://bugs.openjdk.java.net/browse/JDK-8192970>
>>>
>>> Note that this is unrelated to the previous refactoring (8191878), it is
>>> present even before that and also in JDK9, I’ll have to separately produce
>>> a backport fix for it. I discovered this while trying to expand the test
>>> coverage (and good thing that I did, too…)
>>>
>>> The gist of the problem is that convertArgToNumber as it was before will
>>> fail with IndexOutOfBoundsException invoking parameterType(1) on method
>>> type for fixed-key getters, as they only have one parameter at index 0.
>>>
>>> The fix is to move the logic into the Binder and use the methodType it
>>> manages, not the call site method type. This works because Binder always
>>> works with a method type as if the getter/setter were variable-key (it adds
>>> back the parameter type in its constructor if the operation is for a fixed
>>> key).
>>>
>>> Thanks,
>>> Attila.
>>
>
