Question

我有一些unsafe C＃代码无法更改，并公开了这样的方法：

static unsafe void Foo(
    byte* a, int aLength,
    byte* b, int bLength,
    byte* c, int cLength,
    byte* d, int dLength,
    byte* e, int eLength);

我正在调用这样的方法：

static void Bar(
    byte[] a, int aOffset, int aLength,
    byte[] b, int bOffset, int bLength,
    byte[] c, int cOffset, int cLength,
    byte[] d, int dOffset, int dLength,
    byte[] e, int eOffset, int eLength)
{
    fixed (byte* a_ = &a[aOffset])
    fixed (byte* b_ = &b[bOffset])
    fixed (byte* c_ = &c[cOffset])
    fixed (byte* d_ = &d[dOffset])
    fixed (byte* e_ = &e[eOffset])
    {
        Foo(a_, aLength,
            b_, bLength,
            c_, cLength,
            d_, dLength,
            e_, eLength);
    }
}

（为简洁起见，省略了参数验证。）

除非其中一个字节数组的长度为零，否则这种方法很有效。在这种情况下，我得到一个IndexOutOfRangeException。

索引超出了数组的范围。

如何防止异常，最好不要编写大量样板代码而不从fixed转换为其他内容？

不安全的方法不读取或写入长度为零的参数。

Answer 1

您无法访问数组中为空的任何元素，因此请使用包含您可以访问的元素的虚拟数组替换任何空数组：

if (aLength == 0) a = new int[1];
if (bLength == 0) b = new int[1];
if (cLength == 0) c = new int[1];
if (dLength == 0) d = new int[1];
if (eLength == 0) e = new int[1];

Answer 2

是

static void Bar( ... )
{
  byte[] dummy = new byte[1];
  fixed (byte* a_ = a.Lenght>0 ? &a[aOffset] : &dummy[0])
  ...
}

太笨重了？

修改

仅出于历史原因离开上述地方，?运营商无法工作。

因此，您需要不使用fixed关键字，并执行类似
的操作
public unsafe struct UnsafePointerStruct { public GCHandle gch; public byte* addr; } static unsafe UnsafePointerStruct GetUnsafePointer(Array a, int ai) { UnsafePointerStruct ups=new UnsafePointerStruct(); ups.gch=GCHandle.Alloc(a,GCHandleType.Pinned); if (a.Length<=ai) ups.addr=(byte*)IntPtr.Zero; ups.addr=(byte*)Marshal.UnsafeAddrOfPinnedArrayElement(a,ai); return ups; } static unsafe void Bar( byte[] a, int aOffset, int aLength, byte[] b, int bOffset, int bLength, byte[] c, int cOffset, int cLength, byte[] d, int dOffset, int dLength, byte[] e, int eOffset, int eLength) { UnsafePointerStruct upsa=GetUnsafePointer(a,aOffset); UnsafePointerStruct upsb=GetUnsafePointer(b,bOffset); UnsafePointerStruct upsc=GetUnsafePointer(c,cOffset); UnsafePointerStruct upsd=GetUnsafePointer(d,dOffset); UnsafePointerStruct upse=GetUnsafePointer(e,eOffset); Foo(upsa.addr, aLength, upsb.addr, bLength, upsc.addr, cLength, upsd.addr, dLength, upse.addr, eLength); upsa.gch.Free(); upsb.gch.Free(); upsc.gch.Free(); upsd.gch.Free(); upse.gch.Free(); }

Answer 3

由于它不读取或写入长度为零的数组，因此可以更改

fixed (byte* a_ = &a[aOffset])
fixed (byte* b_ = &b[bOffset])
fixed (byte* c_ = &c[cOffset])
fixed (byte* d_ = &d[dOffset])
fixed (byte* e_ = &e[eOffset])

到

fixed (byte* a_ = (a.Length == 0 ? (byte*)IntPtr.Zero : &a[aOffset]))
fixed (byte* b_ = (b.Length == 0 ? (byte*)IntPtr.Zero : &b[aOffset]))
fixed (byte* c_ = (c.Length == 0 ? (byte*)IntPtr.Zero : &c[aOffset]))
fixed (byte* d_ = (d.Length == 0 ? (byte*)IntPtr.Zero : &d[aOffset]))
fixed (byte* e_ = (e.Length == 0 ? (byte*)IntPtr.Zero : &e[aOffset]))

但是如果你这样做然后尝试读取或写入其中一个空数组，你将得到一个nullpointer异常。

使用fixed关键字固定零长度数组

3 个答案: