Incomplete bindings generated when using the --include parameter

[f2bo]

I'm generating mappings for a library where a few of its functions have parameters with types defined in headers external to the library, for example, timespec defined in <time.h>. It's my understanding that I need to traverse this header and since I do not want to bring in every definition in these headers but just generate bindings for the few types used by the library, I use the --include parameter to specify them.

I originally assumed that its behavior would mirror --exclude and that I just need to specify the names to include.

--traverse
C:/Program Files (x86)/Windows Kits/10/Include/10.0.22000.0/ucrt/time.h
C:/Program Files (x86)/Windows Kits/10/Include/10.0.22000.0/ucrt/corecrt_wtime.h

--include
timespec
tm

However, when used in this manner, while it does generate bindings for the specified names, these definitions are empty.

public partial struct timespec
{
}

After some trial and error, I found that specifying the name of every member that needs to be included does work (wildcards are not supported). For example,

--include
timespec
    timespec::tv_sec
    timespec::tv_nsec
tm
    tm::tm_sec
    tm::tm_min
    tm::tm_hour
    tm::tm_mday
    tm::tm_mon
    tm::tm_year
    tm::tm_wday
    tm::tm_yday
    tm::tm_isdst

This seems quite inconvenient and error prone, but lets say that it needs to be this way. However, I'm still having trouble with the definition for sockaddr_in in <winsock.h>.

struct sockaddr_in {
    short   sin_family;
    u_short sin_port;
    struct  in_addr sin_addr;
    char    sin_zero[8];
};

This also requires a definition for in_addr in <inaddr.h>.

typedef struct in_addr {
    union {
        struct { UCHAR s_b1,s_b2,s_b3,s_b4; } S_un_b;
        struct { USHORT s_w1,s_w2; } S_un_w;
        ULONG S_addr;
    } S_un;
#define s_addr  S_un.S_addr /* can be used for most tcp & ip code */
#define s_host  S_un.S_un_b.s_b2    // host on imp
#define s_net   S_un.S_un_b.s_b1    // network
#define s_imp   S_un.S_un_w.s_w2    // imp
#define s_impno S_un.S_un_b.s_b4    // imp #
#define s_lh    S_un.S_un_b.s_b3    // logical host
} IN_ADDR, *PIN_ADDR, FAR *LPIN_ADDR;

I've tried all sorts of combinations but while I can produce a definition for sockaddr_in, I'm unable to generate a complete binding for in_addr.

internal partial struct in_addr
{
    [NativeTypeName("__AnonymousRecord_inaddr_L26_C9")]
    public _S_un_e__Union S_un;
}

internal unsafe partial struct sockaddr_in
{
    public short sin_family;

    [NativeTypeName("u_short")]
    public ushort sin_port;

    [NativeTypeName("struct in_addr")]
    public in_addr sin_addr;

    [NativeTypeName("char[8]")]
    public fixed sbyte sin_zero[8];
}

For completeness, this is the binding generated for `in_addr` when I remove the --include parameter and bindings are generated for every member.

internal partial struct in_addr
{
    [NativeTypeName("__AnonymousRecord_inaddr_L26_C9")]
    public _S_un_e__Union S_un;

    [StructLayout(LayoutKind.Explicit)]
    internal partial struct _S_un_e__Union
    {
        [FieldOffset(0)]
        [NativeTypeName("__AnonymousRecord_inaddr_L27_C17")]
        public _S_un_b_e__Struct S_un_b;

        [FieldOffset(0)]
        [NativeTypeName("__AnonymousRecord_inaddr_L28_C17")]
        public _S_un_w_e__Struct S_un_w;

        [FieldOffset(0)]
        [NativeTypeName("ULONG")]
        public uint S_addr;

        internal partial struct _S_un_b_e__Struct
        {
            [NativeTypeName("UCHAR")]
            public byte s_b1;

            [NativeTypeName("UCHAR")]
            public byte s_b2;

            [NativeTypeName("UCHAR")]
            public byte s_b3;

            [NativeTypeName("UCHAR")]
            public byte s_b4;
        }

        internal partial struct _S_un_w_e__Struct
        {
            public ushort s_w1;

            public ushort s_w2;
        }
    }
}

Any suggestions on how this should be handled? Thanks.

[tannergooding]

timespec::* should work, but I also could have sword that just timespec should be sufficient.

I will take a look.

[f2bo]

timespec::* should work, but I also could have sword that just timespec should be sufficient.

Neither worked for me. I just tested again to make sure though I cannot exclude some other mistake as I've only started using the tool recently. Thanks for taking a look.

[tannergooding]

Sorry for the delay, have been busy with the .NET 8 release.

I've got a PR up that ensures --include makes it implicit that semantic children are also included.

[f2bo]

I ran a quick test and now it behaves as I had originally expected. Thank you!