It is unclear how to get a `CGImage` from a `*mut CGImage`

[ethanpailes]

As a disclaimer, I have a suspicion that what I want to do is possible with the API that already exists, but I've just been unable to figure out how to do it. If that turns out to be the case, I'll open up a documentation PR as a follow up.

I've been trying to create a CGImage from the *mut CGImage that the CGWindowListCreateImage routine returns, but it is not clear to me how this can be accomplished with the public API that core_graphics exposes. In particular, by grepping the sources of core_foundation itself, it seems like the CGImage::from_ptr method is what I want, but that method is attached to the ForeignType trait which is not publicly exposed.

To make this a little more concrete, here is some code that I would like to be able to write:

use core_graphics::image::CGImage;
use core_graphics::display::{
    CGWindowListCreateImage, kCGWindowListOptionIncludingWindow,
    kCGWindowImageBoundsIgnoreFraming, kCGWindowImageBestResolution, CGRectNull
};
use core_graphics::foreign_types::ForeignType;

fn main() {
    let img = unsafe {
        CGWindowListCreateImage(
            CGRectNull,
            kCGWindowListOptionIncludingWindow,
            0, // fake window id, in real life would be fetched with other CG calls
            kCGWindowImageBoundsIgnoreFraming | kCGWindowImageBestResolution,
        )
    };
    if img.is_null() {
        panic!("null ptr");
    }
    let _ = unsafe { CGImage::from_ptr(img) };
}

Unfortunately, this doesn't work:

$ cargo check
    Checking rs v0.1.0 (/Users/ethanpailes/repos/mucking/rs)
error[E0603]: crate `foreign_types` is private
  --> src/main.rs:6:20
   |
6  | use core_graphics::foreign_types::ForeignType;
   |                    ^^^^^^^^^^^^^ this crate is private
   |
note: the crate `foreign_types` is defined here
  --> /Users/ethanpailes/.cargo/registry/src/github.jpy.wang-1ecc6299db9ec823/core-graphics-0.19.0/src/lib.rs:20:1
   |
20 | extern crate foreign_types;
   | ^^^^^^^^^^^^^^^^^^^^^^^^^^^

error: aborting due to previous error

For more information about this error, try `rustc --explain E0603`.
error: could not compile `rs`.

To learn more, run the command again with --verbose.

My temporary work around for this issue is to apply the following patch to my fork of core-foundation-rs:

diff --git a/core-graphics/src/lib.rs b/core-graphics/src/lib.rs
index 6468af8..41cbccd 100644
--- a/core-graphics/src/lib.rs
+++ b/core-graphics/src/lib.rs
@@ -17,7 +17,7 @@ extern crate core_foundation;
 extern crate bitflags;
 
 #[macro_use]
-extern crate foreign_types;
+pub extern crate foreign_types;
 
 pub mod base;
 pub mod color;

This works fine for my use case, but it feels pretty hacky. I feel like I must be missing something if I have to punch such a big hole in the privacy boundary of core_graphics in order to get my code to compile.

Sorry in advance if there is a really obvious way to do this that I'm missing.

[jdm]

In this particular case it looks like you can use the CGDisplay::screenshot method and avoid the issue entirely.

I feel like you could also import the ForeignType trait directly from the foreign_types crate, rather than trying to import it via core_graphics, too.

[ethanpailes]

In this particular case it looks like you can use the CGDisplay::screenshot method and avoid the issue entirely.

Oh, that's exactly what I need. Thanks! Sorry I missed it!

I feel like you could also import the ForeignType trait directly from the foreign_types crate, rather than trying to import it via core_graphics, too.

I did try that by adding foreign-types = "0.5" to my Cargo.toml with no luck. Of course, now that I've already wasted your time I thought to check the version core_graphics actually uses and used foreign-types = "0.3" which worked like a charm.

It looks like my suspicion that I was being dumb was right on two different counts.

Would you be open to a documentation PR to try to help others from going down a similar wrong path?

[ethanpailes]

Alternatively, it might actually make sense to re-export the foreign_types crate (or at least some select traits). It is already an implicit public dependency if people can depend on it directly in order to work with types exported by core_graphics, and making it an explicit public dependency would both make it easier for downstream users to ensure they are using an in-sync version and make it more obvious to maintainers that bumping the foreign-types version is a breaking change.

It is probably not published for a reason, so I could be missing something here, but I think it is worth considering.

[arbaaz]

used foreign-types = "0.3" which worked like a charm.

It also worked for me, thanks.