static inline and jextract

[Duncan Gittins]

If you are experimenting with generating all symbols before deciding on
which --include options to use, then you can workaround the missing static
symbols issue (which happens frequently on Windows) by making jextract use
one symbol per file:

     set _JAVA_OPTIONS=-Djextract.constants.per.class=0

Obviously the number of source files increases a lot but you won't see
"java.lang.UnsatisfiedLinkError: unresolved symbol: xxx" unless you
inadvertently use one of the those static import symbols.

Kind regards

Duncan Gittins

On Mon, 5 Sept 2022 at 12:30, Manuel Bleichenbacher <
manuel.bleichenbacher at gmail.com> wrote:

> Is there a reason to treat inline static functions different from
> function-like macros?
>
> Inline static functions are never exported. So if code is generated, it
> will always fail. It would make much more sense to ignore them and not
> generate code.
>
> If code is generated, workarounds must be used. In particular, the
> function and data structures must be explicitly specified. It is no longer
> possible to use jextract for entire header files if just a single inline
> static function is present.
>
>
> On Mon, Sep 5, 2022 at 4:42 AM Sundararajan Athijegannathan <
> sundararajan.athijegannathan at oracle.com> wrote:
>
>> Hi,
>>
>> Only symbols exported from shared objects/DLLs/dylibs can be called (for
>> eg. marked declspec(dllexport) on Windows, listed in mapfile on Unix
>> platforms and so on). Using platform specific tools such as "nm" (Linux),
>> "otool" (macOS) etc, you can check if that symbol is exported from the
>> library.
>>
>> Also, you need to load that native library using System.load/loadLibrary
>> API before creating native method handles.
>>
>> Hope this helps,
>> -Sundar
>> ------------------------------
>> *From:* jextract-dev <jextract-dev-retn at openjdk.org> on behalf of
>> dreamlike_ocean lei <dreamlike.vertx at gmail.com>
>> *Sent:* 04 September 2022 20:00
>> *To:* jextract-dev at openjdk.org <jextract-dev at openjdk.org>
>> *Subject:* static inline and jextract
>>
>> English is not my first language
>>
>> I used jextract to generate a java-bound version of liburing.h.
>> I found an unreasonable situation。 such as:
>> `static inline void io_uring_prep_readv`
>>
>> It is not an exported function，But jextract generates code like this
>>  public static void io_uring_prep_readv(Addressable var0, int var1,
>> Addressable var2, int var3, long var4)
>> when I call ` io_uring_prep_readv` in java, It will throw Exception in
>> thread "main" java.lang.UnsatisfiedLinkError: unresolved symbol:
>> io_uring_prep_readv
>>  Isn't this unreasonable and inconvenient？Shouldn't only exported
>> functions be generated？
>> ————————Chinese version—————————————
>> 我利用jextract 生成liburing.h的java绑定
>> 但是我发现了一个很奇怪的情况，像`static inline void
>> io_uring_prep_readv`这种函数并不是导出的函数，但是jextract 确实生成了这个的java代码。
>>  public static void io_uring_prep_readv(Addressable var0, int var1,
>> Addressable var2, int var3, long var4)
>> 当我调用用java去调用io_uring_prep_readv时，它会抛出一个Exception in thread "main"
>> java.lang.UnsatisfiedLinkError: unresolved symbol: io_uring_prep_readv异常
>> 这确实不太合理也不太方便吧，难道不应该只生成导出的函数吗？
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/jextract-dev/attachments/20220905/64c428aa/attachment-0001.htm>