Passing pointer of address

[Maurizio Cimadamore]

On 29/04/2024 13:28, Thiago Milczarek Sayão wrote:
> Hi Maurizio,
>
> My mistake. It works.
>
> I was playing around with it and my first version did not work. But 
> the problem was on another part of the code.
>
> elgConfig is an array, so eglCreateWindowSurface only works by passing 
> like below (I might be mistaken again).

I think that's correct. E.g. ELGConfig is defined as follows:


```
typedef void *EGLConfig;
```

So it's an alias for a void pointer. If your "eglConfig" is configured 
as "pointer to pointer" (which I think you need for the call you showed 
earlier), then taking the first pointer in this pointer-to-pointer is 
what you want.

I'm more suspicious about that "egWindow.address()" - the function seems 
to take a void* - and you probably have "eglWindow" which is a segment, 
so if you want to pass the segment "by reference" you just need to pass 
the segment directly. Perhaps you declared the function descriptor for 
this function as accepting a JAVA_LONG instead of ADDRESS?

Maurizio

>
> Thanks for replying.
>
> if (EGL.eglChooseConfig(eglDisplay, attributes, eglConfig,1, numConfigs) ==0) {
>      System.out.println("eglChooseConfig Failed");
>      return null;
> }
>
> MemorySegment eglSurface = EGL.eglCreateWindowSurface(eglDisplay, eglConfig.get(EGLConfig,0), eglWindow.address(), MemorySegment.NULL);
>
> if (eglSurface.equals(EGL_NO_SURFACE())) {
>      System.out.printf("eglCreateWindowSurface Failed %d%n",eglGetError());
>      return null;
> }
>
>
> Em seg., 29 de abr. de 2024 às 08:08, Maurizio Cimadamore 
> <maurizio.cimadamore at oracle.com> escreveu:
>
>     Hi Thiago,
>
>     So, you have a memory segment with a given address, called
>     "attributes".
>
>     The code you have is:
>
>     1. obtaining the addres from "attributes"
>     2. repackaging the address in a _new_ memory segment
>
>     Note that the segment you create in (2) is identical to the one
>     you started from. They have the same address. The only difference
>     is that the first segment has some meaningful spatial bounds (e.g.
>     it has  a size), the second segment is a zero-length memory
>     segment (so you can't dereference it).
>
>     But from the perspective of passing this segment "by reference" to
>     a native function, they are exactly the same - e.g. when passing a
>     segment by reference, the Linker will call
>     "MemorySegment::address()" and pass that to the underlying
>     function. Since the address of segments in (1) and (2) is indeed
>     the same, I don't think these segments should behave differently
>     at all - meaning your code can be simplified, and you can just use
>     "attributes".
>
>     This misconception might have originated from the fact that FFM
>     doesn't provide true "stack allocation". So when you have a
>     segment you already have some address. Meaning that if the
>     original C code said something like "&foo", and "foo" is a segment
>     in the FFM code, you can just pass "foo" as a pointer instead.
>
>     I hope this helps.
>
>     Maurizio
>
>     On 29/04/2024 11:45, Thiago Milczarek Sayão wrote:
>>     Hi,
>>
>>     I'm doing some wayland binding experiments with jextract and FFM.
>>     So far, so good.
>>
>>     But there's one thing that could be better, or I am doing it
>>     wrong (Please, excuse me if it's my mistake.).
>>
>>     The eglChooseConfig call expects the pointer of an address on the
>>     second argument.
>>     https://registry.khronos.org/EGL/sdk/docs/man/html/eglChooseConfig.xhtml
>>     <https://urldefense.com/v3/__https://registry.khronos.org/EGL/sdk/docs/man/html/eglChooseConfig.xhtml__;!!ACWV5N9M2RV99hQ!KNXTJkNJKuDXvkpPfP8ggUb2q4pK32qw-neOjb6Jem_gtbnVpR7KTYIhlbFAxzjIuITW6tZ9Zjjfcg4MNt-ybFdJ7zz2Mw$>
>>
>>     The below code works, but I feel it could be simpler. Maybe a
>>     call like attributes.addressPointer() ?
>>     MemorySegment eglDisplay = EGL.eglGetDisplay(display.getSegment());
>>     eglBindAPI(EGL_OPENGL_API());
>>
>>     if (eglDisplay.equals(EGL_NO_DISPLAY())) {
>>          System.out.println("NO DISPLAY");
>>          return null;
>>     }
>>
>>     if (eglInitialize(eglDisplay, MemorySegment.NULL, MemorySegment.NULL) ==0) {
>>          System.out.println("eglInitialize Failed");
>>          return null;
>>     };
>>
>>     int[] attribs = {
>>          EGL_RENDERABLE_TYPE(),
>>          EGL_OPENGL_BIT(),
>>          EGL_SURFACE_TYPE(),
>>          EGL_WINDOW_BIT(),
>>          EGL_RED_SIZE(),
>>          8,
>>          EGL_GREEN_SIZE(),
>>          8,
>>          EGL_BLUE_SIZE(),
>>          8,
>>          EGL_ALPHA_SIZE(),
>>          8,
>>          EGL_NONE()
>>     };
>>
>>     MemorySegment eglConfig = Arena.global().allocate(C_POINTER);
>>     MemorySegment numConfigs = Arena.global().allocate(EGLint);
>>     MemorySegment attributes = Arena.global().allocateFrom(EGLint, attribs);
>>
>>     if (EGL.eglChooseConfig(eglDisplay, MemorySegment.ofAddress(attributes.address()), eglConfig,1, numConfigs) ==0) {
>>          System.out.println("eglChooseConfig Failed");
>>          return null;
>>     }
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/panama-dev/attachments/20240429/eff47095/attachment-0001.htm>