[foreign-abi] RFR 8228509: Add support for binding varargs functions on Windows

[Jorn Vernee]

Here is the updated webrev: 
http://cr.openjdk.java.net/~jvernee/panama/webrevs/8228509/webrev.01/

I removed the varargs version of CallingSequenceBuilder::addArgument, as 
well as the varargs binding computer.

I also went over the Windows CallingSequenceBuilderImpl and cut out a 
lot of stuff that wasn't actually needed.

Jorn

On 2019-07-25 17:32, Maurizio Cimadamore wrote:
> Looks very nice.
> 
> But I think we can make it simpler :-)
> 
> Note that in SysV and Aarch64 we don't care whether an argument is
> varargs or not. In Windows we do, but now the layout is expressive
> enough to encode the 'isVarargs' info itself.
> 
> So... I propose to change CallingSequenceBuilder::addArgument and to
> drop the boolean 'isVarargs' from there. Win64
> CallingSequenceBuilderImpl can easily check the layout annotation and
> act accordingly.
> 
> Other than that looks great - very nice test!
> 
> Maurizio
> 
> 
> On 25/07/2019 16:02, Jorn Vernee wrote:
>> Hi,
>> 
>> Please review the following patch which adds support for using varargs 
>> functions with float arguments on Windows (e.g. printf).
>> 
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8228509
>> Webrev: 
>> http://cr.openjdk.java.net/~jvernee/panama/webrevs/8228509/webrev.00/
>> 
>> This mainly consists of a way to mark varargs argument layouts as such 
>> through the use of an annotation, which is then detected when 
>> computing a CallingSequence.
>> 
>> I've also added a general purpose test for testing the passing of 
>> vararg arguments.
>> 
>> Thanks,
>> Jorn