在(gcc)go中的打包结构

huangapple go评论116阅读模式
英文:

Packed Structs in (gcc)go

问题

我有一些旧的C代码,其中使用了大量的紧凑结构。我正在研究使用Go作为这些代码的包装器,但是在传递或甚至编写这些结构的定义方面遇到了困难。

示例:

import "unsafe";

type AlignTest struct {
    c byte;
    y int16;
    z int16;
    q int32;
}

func main() {

    vr := new(AlignTest);

    fmt.Println(unsafe.Sizeof(*vr), "\n");

}

返回的结果是12,而不是我希望的1+2+2+4 = 9,这是一个紧凑/非对齐的结构。

我知道我可以只创建一个字节数组并手动解析,但这似乎非常脆弱和容易出错...

英文:

I have some old C code that makes somewhat heavy use of packed structures. I'm looking into using Go as a wrapper for this code, but am having difficulty finding a way to pass or even write definitions for these structures.

Example:

import "unsafe";

type AlignTest struct {
    c byte;
    y int16;
    z int16;
    q int32;
}

func main() {

    vr := new(AlignTest);

    fmt.Println(unsafe.Sizeof(*vr),  "\n");

}

Returns 12 rather than the 1+2+2+4 = 9 that I would want with a packed/unaligned struct.

I know that I could just create a byte array and do the parsing manually, but that seems very brittle and error prone...

答案1

得分: 6

你可以尝试这样做。

package main

import (
    "encoding/binary"
    "bytes"
    "fmt"
)

type Unpacked struct {
    C byte
    Y int16
    Z int16
    Q int32
}

type Packed struct {
    B [9]byte
}

func main() {
    var u Unpacked
    var p Packed
    var buf = bytes.NewBuffer(make([]byte, 0, len(p.B)))
    // Unpacked to Packed 
    u = Unpacked{1, 2, 3, 4}
    if err := binary.Write(buf, binary.BigEndian, &u); err != nil {
        fmt.Println(err)
    }
    if err := binary.Read(buf, binary.BigEndian, &p); err != nil {
        fmt.Println(err)
    }
    fmt.Println("u", u, "to", "p", p)
    // Packed to Unpacked
    p = Packed{[...]byte{1, 2, 3, 4, 5, 6, 7, 8, 9}}
    if err := binary.Write(buf, binary.BigEndian, &p); err != nil {
        fmt.Println(err)
    }
    if err := binary.Read(buf, binary.BigEndian, &u); err != nil {
        fmt.Println(err)
    }
    fmt.Println("p", p, "to", "u", u)
}

.

输出:
u {1 2 3 4} to p {[1 0 2 0 3 0 0 0 4]}
p {[1 2 3 4 5 6 7 8 9]} to u {1 515 1029 101124105}
英文:

You could try something like this.

package main

import (
	"encoding/binary"
	"bytes"
	"fmt"
)

type Unpacked struct {
	C byte
	Y int16
	Z int16
	Q int32
}

type Packed struct {
	B [9]byte
}

func main() {
	var u Unpacked
	var p Packed
	var buf = bytes.NewBuffer(make([]byte, 0, len(p.B)))
	// Unpacked to Packed 
	u = Unpacked{1, 2, 3, 4}
	if err := binary.Write(buf, binary.BigEndian, &u); err != nil {
		fmt.Println(err)
	}
	if err := binary.Read(buf, binary.BigEndian, &p); err != nil {
		fmt.Println(err)
	}
	fmt.Println("u", u, "to", "p", p)
	// Packed to Unpacked
	p = Packed{[...]byte{1, 2, 3, 4, 5, 6, 7, 8, 9}}
	if err := binary.Write(buf, binary.BigEndian, &p); err != nil {
		fmt.Println(err)
	}
	if err := binary.Read(buf, binary.BigEndian, &u); err != nil {
		fmt.Println(err)
	}
	fmt.Println("p", p, "to", "u", u)
}

.

Output:
u {1 2 3 4} to p {[1 0 2 0 3 0 0 0 4]}
p {[1 2 3 4 5 6 7 8 9]} to u {1 515 1029 101124105}

答案2

得分: 5

你可能需要重新考虑你的架构-尝试将二进制输入传递到C层,并使用现有的结构(你不会破坏你不改变的东西)。我假设结构体的打包看起来像这样:

#ifdef WINDOWS
#pragma pack(push)
#endif
#pragma pack(BYTEALIGNMENT) // 例如 "#pragma pack(1)" 或 "#pragma pack(8)"

//--- 一些打包的结构体

#ifdef WINDOWS
#pragma pack(pop)
#endif
#ifdef POSIX
#pragma pack()
#endif

所有底层或第三方库所做的就是将一些void或const char进行类型转换为这些结构体。所以如果可能的话,尝试将这些数据转发到C层(在那里你可以获得指针),并且不要暴露这些结构体。

英文:

You may want to rethink your architecture- try passing the binary input down to the C layer and use the existing structures (you won't break what you don't change). I'm assuming the structure packing looks something like this:

#ifdef WINDOWS
#pragma pack(push)
#endif
#pragma pack(BYTEALIGNMENT) // e.g. "#pragma pack(1)" or "#pragma pack(8)"

//--- Some packed structs

#ifdef WINDOWS
#pragma pack(pop)
#endif
#ifdef POSIX
#pragma pack()
#endif

All the underlying or 3rd Party libs are then doing is taking some void* or const char* and typecasting it to these. So if possible, try forwarding that data into a C layer (where you can get pointers) and don't expose the structures at all.

答案3

得分: 2

没有办法告诉gccgo编译紧凑结构。我能想到的最好的解决方案是手动添加填充:

type AlignTest struct {
    c byte
    _ [3]byte // 匿名填充
    y int16
    z int16
    q int32
}
英文:

There's no way to tell gccgo to compile packed structures. The best solution I can think of is to manually add padding:

type AlignTest struct {
    c byte
    _ [3]byte // anonymous padding
    y int16
    z int16
    q int32
}

答案4

得分: 1

这是工作的代码:

package main

import "unsafe"
import "fmt"

/*
struct test {
    char c;
    short i;
    short j;
    int k;
} __attribute__((packed));
*/
import "C"

type AlignTest struct {
    c byte
    y int16
    z int16
    q int32
}

func main() {

    vr := new(AlignTest)
    v := new(C.struct_test)

    fmt.Println(unsafe.Sizeof(*vr), "\n")
    fmt.Println(unsafe.Sizeof(*v), "\n")

}

输出:

12

9
英文:

This works:

package main

import "unsafe"
import "fmt"

/*
struct test {
    char c;
    short i;
    short j;
    int k;
} __attribute__((packed));
*/
import "C"

type AlignTest struct {
    c byte
    y int16
    z int16
    q int32
}

func main() {

    vr := new(AlignTest)
    v := new(C.struct_test)

    fmt.Println(unsafe.Sizeof(*vr), "\n")
    fmt.Println(unsafe.Sizeof(*v), "\n")

}

Output:

12

9

huangapple
  • 本文由 发表于 2011年7月13日 06:54:55
  • 转载请务必保留本文链接:https://go.coder-hub.com/6672205.html
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定