[foreign-abi] RFR 8237357: Add verification code for binding recipes

[Jorn Vernee]

Here is the update: 
http://cr.openjdk.java.net/~jvernee/panama/webrevs/8237357/webrev.01-inc/

Moving the verification to the interpreter is a possibility as well, but 
I'll also note that then there's less of a guarantee that the 
verification is done, since a function might be linked, but never executed.

Jorn

On 29/01/2020 12:19, Maurizio Cimadamore wrote:
> Flag turned on by default is fine.
>
> I guess what I wanted to say is that, if you plan to have this enabled 
> all the time, ther eare ways to fold the sanity check together with 
> the interpretation (so that you do a single pass). But I guess we can 
> also decide later if we want to turn this on/off by default, and 
> restructure the code accordingly.
>
> Maurizio
>
> On 29/01/2020 11:06, Jorn Vernee wrote:
>> You mean to speed up linking by skipping verification? Possibly, but 
>> I'd say let's always have it on for now, since it might catch a case 
>> not covered by the tests.
>>
>> I could add the flag and have it default to 'true'. What do you think?
>>
>> Jorn
>>
>> On 28/01/2020 17:00, Maurizio Cimadamore wrote:
>>> Looks good - question, maybe it's worth putting verification behind 
>>> a flag and enable it in our downcall/upcall stress tests?
>>>
>>> Maurizio
>>>
>>> On 28/01/2020 15:17, Jorn Vernee wrote:
>>>> Hi,
>>>>
>>>> Please review the following patch that adds verification code for 
>>>> binding recipes.
>>>>
>>>> Bug: https://bugs.openjdk.java.net/browse/JDK-8237357
>>>> Webrev: 
>>>> http://cr.openjdk.java.net/~jvernee/panama/webrevs/8237357/webrev.00/
>>>>
>>>> Tested on Windows and Linux (WSL).
>>>>
>>>> I'll hold off on pushing until the repo has been fixed.
>>>>
>>>> Thanks,
>>>> Jorn
>>>>