[Xcb] about xcb_image

[Vincent Torri]

On Thu, 21 Feb 2008, Barton C Massey wrote:

> In message <Pine.LNX.4.64.0802210909030.1308 at grozny.maths.univ-evry.fr> you wrote:
>> I'm trying to use the new xcb_image api in the efl (especially in evas),
>> and I have some difficulties to understand what values to pass to
>> xcb_create_native.
>>
>> The old call was:
>>
>> xcb_image_create(c, depth, XCB_IMAGE_FORMAT_Z_PIXMAP, 0, data, w, h, 32, 0);
>>
>> I tried something that I thought correct:
>>
>> xcb_image_create_native(c, w, h, XCB_IMAGE_FORMAT_Z_PIXMAP, depth, NULL, 0, data)
>>
>> Am I wrong ?
>
> The memory allocation (last three arguments) for
> xcb_image_create() is magic, to say the least.  What exactly
> is in "data", and what are you trying to do with it?

data is an argb (32 bpp) data of size (w,h), allocated by evas.

> Other than that potential problem, I see nothing obvious
> wrong.
>
>> same question for
>>
>> xcb_image_shm_create(c, depth, XCB_IMAGE_FORMAT_Z_PIXMAP, NULL, w, h);
>>
>> and
>>
>> xcb_image_create_native(c, w, h, XCB_IMAGE_FORMAT_Z_PIXMAP, depth, NULL, ~0, NULL);
>
> You might want 0 in the next-to-last argument here, if you
> want create_native() to allocate and set up the storage.
> Otherwise you'll need to do it yourself...

ok, I'll try

> My documentation for the memory stuff is *horrible*.  This
> comment in the source is slightly better:
>
>   * Ways this function can be called:
>   *   * with data: we fail if bytes isn't
>   *     large enough, else leave well enough alone.
>   *   * with base and !data: if bytes is zero, we
>   *     default; otherwise we fail if bytes isn't
>   *     large enough, else fill in data
>   *   * with !base and !data: we malloc storage
>   *     for the data, save that address as the base,
>   *     and fail if malloc does.
>   *
>   * When successful, we establish the invariant that data
>   * points at sufficient storage that may have been
>   * supplied, and base is set iff it should be
>   * auto-freed when the image is destroyed.
>   *
>   * Except as a special case when base = 0 && data == 0 &&
>   * bytes == ~0 we just return the image structure and let
>   * the caller deal with getting the allocation right.
>
> but I'm not sure it's exactly right :-9.  The C code is
> clearer, sadly:
>
>  if (!base && !data && bytes == ~0)
>      return image;
>  if (!base && data && bytes == 0)
>      bytes = image->size;
>  image->base = base;
>  image->data = data;
>  if (!image->data) {
>      if (image->base) {
>          image->data = image->base;
>      } else {
>          bytes = image->size;
>          image->base = malloc(bytes);
>          image->data = image->base;
>      }
>  }
>  if (!image->data || bytes < image->size) {
>      free(image);
>      return 0;
>  }
>  return image;

actually, i looked at the code to know how it works. But it's not 
completely clear for me. One thing that surprises me is that i do not need 
to call xcb_image_destroy in evas, now. I don't understand why. In 
addition, valgrind does not complain... Black magic, here.

Vincent