Importing miniaudio.h library results in dependency loop error

[kavika13]

Zig Version

0.12.0-dev.3403+b5cef9e8b

Steps to Reproduce and Observed Behavior

1. zig init

2. git clone https://github.com/mackron/miniaudio

3. edit build.zig, and add these lines just after the const exe = ... declaration:

    exe.addIncludePath(.{ .path = "miniaudio" });
    exe.linkLibC();

4. edit src/main.zig, and delete all the code but the const std = @import("std"); line.

5. add the following additional code code:

const miniaudio = @cImport({
    @cDefine("MINIAUDIO_IMPLEMENTATION", {});
    @cInclude("miniaudio.h");
});

fn data_callback(_: *ma_device, _: *void, _: *const void, _: u32) void {}

pub fn main() !void {
    var config: miniaudio.ma_device_config = miniaudio.ma_device_config_init(miniaudio.ma_device_type_playback);
    config.playback.format = miniaudio.ma_format_f32;
    config.playback.channels = 2;
    config.sampleRate = 48000;
    config.dataCallback = data_callback;
}

6. Run zig build

Behavior:

<srcroot>\zig-cache\o\3ce1300da0f0bde9e958a78552702100\cimport.zig:118:5: error: dependency loop detected
pub const ma_device_data_proc = ?*const fn ([*c]ma_device, ?*anyopaque, ?*const anyopaque, ma_uint32) callconv(.C) void;

Expected Behavior

I expected the program to compile successfully, or at least get further.
There may be other errors to fix in the zig code, but the error in the C import would go away.

[kavika13]

I stripped the miniaudio.h header down as much as I could, and here's zig and C code that replicate the problem about as minimally as I could manage:

const std = @import("std");
const miniaudio = @cImport({
    @cInclude("repro.h");
});

pub fn main() !void {
    miniaudio.ma_device_config_init();
}

typedef struct ma_device ma_device;

typedef void (* ma_device_data_proc)(struct ma_device* pDevice);

struct ma_device_config
{
    ma_device_data_proc dataCallback;
};

struct ma_device
{
    ma_device_data_proc onData;
};

void ma_device_config_init(void) {
    struct ma_device_config val;
}

The generated zig code from cimport.zig looks likes this:

pub const ma_device_data_proc = ?*const fn ([*c]struct_ma_device) callconv(.C) void;
pub const struct_ma_device = extern struct {
    onData: ma_device_data_proc = @import("std").mem.zeroes(ma_device_data_proc),
};
pub const ma_device = struct_ma_device;
pub const struct_ma_device_config = extern struct {
    dataCallback: ma_device_data_proc = @import("std").mem.zeroes(ma_device_data_proc),
};
pub export fn ma_device_config_init() void {
    var val: struct_ma_device_config = undefined;
    _ = &val;
}

[xdBronch]

i think this looks like #12325

[odecaux]

I've had a similar issue using miniaudio in the past, but it got fixed by #17692, according to @kcbanner

[sphaerophoria]

If it's helpful, I've got this to work by just patching miniaudio to replace ma_device* function arguments with void* function arguments.

--- miniaudio.h 2024-05-09 15:09:33.966673423 -0700
+++ miniaudio.h_        2024-05-09 15:03:58.688300510 -0700
@@ -6785,7 +6785,7 @@
 due to things like an incoming phone call. Currently this is only implemented on iOS. None of the
 Android backends will report this notification.
 */
-typedef void (* ma_device_notification_proc)(const ma_device_notification* pNotification);
+typedef void (* ma_device_notification_proc)(const void* pNotification);


 /*
@@ -6828,7 +6828,7 @@

 The proper way to stop the device is to call `ma_device_stop()` from a different thread, normally the main application thread.
 */
-typedef void (* ma_device_data_proc)(ma_device* pDevice, void* pOutput, const void* pInput, ma_uint32 frameCount);
+typedef void (* ma_device_data_proc)(void* pDevice, void* pOutput, const void* pInput, ma_uint32 frameCount);



@@ -6852,7 +6852,7 @@
 -------
 Do not restart or uninitialize the device from the callback.
 */
-typedef void (* ma_stop_proc)(ma_device* pDevice);  /* DEPRECATED. Use ma_device_notification_proc instead. */
+typedef void (* ma_stop_proc)(void* pDevice);  /* DEPRECATED. Use ma_device_notification_proc instead. */

 typedef enum
 {
@@ -16187,7 +16187,7 @@
 static ma_result ma_mutex_init__posix(ma_mutex* pMutex)
 {
     int result;
-
+
     if (pMutex == NULL) {
         return MA_INVALID_ARGS;
     }

[kavika13]

I've dug in a bit more.
I triggered this bug with code generated by TranslateC, but it appears to me the TranslateC step itself is succeeding, and the code in the cache looks just fine. I don't think this bug should necessarily be tagged TranslateC.

This is definitely just a special case of #12325

I've had a similar issue using miniaudio in the past, but it got fixed by #17692, according to @kcbanner

I've verified, and this bug still persists. It's not fixed by that other fix.