Pointer/Scope API questions

[Maurizio Cimadamore]

On 08/01/2020 12:12, Ty Young wrote:
>
> On 1/8/20 5:53 AM, Maurizio Cimadamore wrote:
>>
>> On 08/01/2020 11:48, Ty Young wrote:
>>>
>>> On 1/8/20 5:43 AM, Maurizio Cimadamore wrote:
>>>>
>>>> On 08/01/2020 11:38, Ty Young wrote:
>>>>>
>>>>> On 1/8/20 5:09 AM, sundararajan.athijegannathan at oracle.com wrote:
>>>>>> [...]
>>>>>>>>>
>>>>>>>> Are you also adding `--add-modules 
>>>>>>>> jdk.incubator.foreign,jdk.incubator.jextract` when running? I 
>>>>>>>> was seeing a similar error when I forgot to do that.
>>>>>>>
>>>>>>>
>>>>>>> Ah, no. Works now.
>>>>>>>
>>>>>>>
>>>>>>> Is it supposed to throw an exception there though? The modules 
>>>>>>> are apart of the JDK and preview features are enabled so I don't 
>>>>>>> understand why the extra step of adding the modules at runtime 
>>>>>>> is being required here. Is it a bug?
>>>>>>
>>>>>> No. It is not a bug. You've to explicitly add "incubator" modules 
>>>>>> for compilation as well as for runtime.
>>>>>
>>>>>
>>>>> TIL.
>>>>>
>>>>>
>>>>> Just noticed though that while the Memory Access API works, the 
>>>>> jextract API fails because of javax.tools's JaveFileObject from 
>>>>> the java.compiler module.
>>>>>
>>>>>
>>>>> My compiler/runtime --add-modules argument looks like:
>>>>>
>>>>>
>>>>> --add-modules 
>>>>> jdk.incubator.foreign,jdk.incubator.jextract,java.compiler
>>>>>
>>>>>
>>>>> It is also required in module-info.java. Am I missing something here?
>>>>
>>>> Uhm - java.compiler should be available w/o extra flags... what 
>>>> code are you trying to run?
>>>
>>>
>>> Code:
>>>
>>>
>>> Path header = Path.of("/opt/cuda/targets/x86_64-linux/include/nvml.h");
>>>
>>> JextractTask task = JextractTask.newTask(true, header);
>>>
>>>  task.write(Path.of("./", args), new JavaFileObjectTest()); <---- ERROR
>>
>> Do you have a module info?
>
>
> Yep:
>
>
> module PanamaMemTest
> {
>     requires java.compiler;
>     requires jdk.incubator.foreign;
>     requires jdk.incubator.jextract;
>
>     exports org.goliath.panamamemtest;
> }

Ok - if you use a module info you have to say which modules you want - 
and so you do need a requires java.compiler.

So, your module info seems correct. Actually, I believe you only need 
'requires jdk.incubator.jextract' and that will also give you the memory 
access API (jextract "transitively" requires it).

Since everything seems almost correct here- how is the application ran 
(by Netbeans) ? Is there any command line anywhere which gives some 
hints as to whether netbeans is using classpath or modulepath when 
running the Main class?

>
>
> I'm using Netbeans 11.2, if that matters.
>
>
>> Or are you just running off the classpath?
>>
>> (btw, for now the jextract API doesn't really generate any bindings - 
>> but it can be used to 'parse' an header and inspect the structure of 
>> the declarations in it).
>
>
> I'm a bit confused here, your clang code that you linked is commented 
> saying that it was generated by jextract. Is it internal still then?

Yes, that jextract is still internal - we will open it up in the next 
few weeks (shouldn't take too long).

Maurizio

>
>
>>
>> Maurizio
>>
>>>
>>>
>>> Exception:
>>>
>>>
>>> Exception in thread "main" java.lang.NoClassDefFoundError: 
>>> javax/tools/JavaFileObject
>>>     at 
>>> jdk.incubator.jextract/jdk.internal.jextract.impl.Writer.resources(Writer.java:108)
>>>     at 
>>> jdk.incubator.jextract/jdk.internal.jextract.impl.Writer.writeAll(Writer.java:57)
>>>     at 
>>> jdk.incubator.jextract/jdk.internal.jextract.impl.JextractTaskImpl.write(JextractTaskImpl.java:65)
>>>     at 
>>> org.goliath.panamamemtest.PanamaMemTest.main(PanamaMemTest.java:37)
>>>
>>>
>>>>
>>>> Maurizio
>>>>
>>>>>
>>>>>
>>>>>>
>>>>>> -Sundar
>>>>>>