[foreign] RFR 8212560 : jextract should generate a static forwarder class

[Sundararajan Athijegannathan]

On option: do we need an explicit option? given that one file extra and 
that too only when -l is specified. (too many options already? ;) )

On naming: we've derived automatic names elsewhere. I think we could 
revisit all naming options by another round - perhaps even java 
convention for header interface name.

On subclassing: Nice suggestion. I made it as AsmCodeFactoryExt class 
and extended it from AsmCodeFactory. visit methods now return Boolean to 
tell if the particular tree was handled or not. This is needed so that 
subclass can avoid the same tree if super class avoided it (for eg. 
function-like macros, repeated definitions..)

Updated: http://cr.openjdk.java.net/~sundar/8212560/webrev.01/index.html

Thanks,
-Sundar

On 17/10/18, 3:25 PM, Maurizio Cimadamore wrote:
> I like it! Few questions/comments:
>
> * should it be enabled/disabled with explicit option
> * should the name of the statics class be customizable
> * I like the code organization - have you thought of pushing it 
> further and make StaticForwarderGenerator a _subclass_ of 
> AsmCodeFactory - each visitor could maybe delegate to super.visitXyz() 
> first and then do its own bit? Then when you setup the pipeline, 
> depending on what info is available on the context (e.g. presence of 
> -l) you can decide whether to use a 'bare' ASMCodeFactory or the 
> 'embellished' one. That should remove all the 'if staticFwd != null 
> ...' sections from AsmCodeFactory.
>
> Maurizio
>
>
> On 17/10/2018 06:35, Sundararajan Athijegannathan wrote:
>> Please review.
>>
>> Bug: https://bugs.openjdk.java.net/browse/JDK-8212560
>> Webrev: http://cr.openjdk.java.net/~sundar/8212560/webrev.00/index.html
>>
>> Thanks,
>> -Sundar