1

我正在使用带有 JNA 框架的 Java 应用程序中的本机 c++ dll。我有一个函数调用的问题。也许我没有正确分配内存？我没有意见，下一步该尝试什么。我没有得到更多信息来帮助我摆脱文档和论坛线程。我希望你能给我一个提示，会很好。

我想调用一个本机函数（在示例中FooInterface）。该函数将结构体TNativeFoo作为输入和输出参数。该结构TNativeFoo包含一个Double**/Double[][]数据。该字段可以看作是一个多维数组，但第一个维度的长度只有 1。所以它更像是一个指向大小为 的双精度数组的指针Items。

StringArray结构（char**/ ）中还有一个String[]带有 length 的StringsCount，这是无关紧要的，因为它没有被使用。我提到它是因为我不确定该错误是否与此有关。下面是对本机 dll 和结构的定义。

 public interface Foodll extends StdCallLibrary {

    Foodll INSTANCE = (Foodll) Native.loadLibrary("foodll.dll", Foodll.class);

    public static class TNativeFoo extends com.sun.jna.Structure {

    public TNativeFoo (){
            super();
            setAlignType(ALIGN_NONE);
        }
        public TNativeFoo(com.sun.jna.Pointer pointer,int offset) {
            super();
            setAlignType(ALIGN_NONE);
            useMemory(pointer,offset);
            read();
        }
        public TNativeFoo(TNativeFoo struct) {
        this(struct.getPointer(),0);
    }

        public static class ByReference extends TNativeFoo implements com.sun.jna.Structure.ByReference {
            ByReference() {}
            ByReference(TNativeFoo struct){super(struct.getPointer(),0);}
        }
        public static class ByValue extends TNativeFoo implements com.sun.jna.Structure.ByValue {
            ByValue() {}
            ByValue(TNativeFoo struct){super(struct.getPointer(),0);}
        }

        public PointerByReference Data;
        public NativeLong Items;
        public PointerByReference IrrelevantStringArray;
        public NativeLong StringsCounts = new NativeLong(0);
    }

    NativeLong FooInterface(TNativeFoo input, TNativeFoo output);
}

对于函数的调用，我尝试在本机堆上分配内存并将数据写入其中。在像示例中那样将数据写入本机堆之后，我可以像在示例中那样使用替代方案 2 读取它（我在没有调用本机函数的情况下尝试了这个，并且直接从inputFoo. 像下面的示例中调用本机函数会引发致命异常。

调用：

public class FooInvocationClass
{
    public static FooInvocationMethod(double[] fooData)
    {

                Foodll foodllJnaLib = Foodll.INSTANCE;

                Foodll.TNativeFoo outputFoo = new Foodll.TNativeFoo();
                Foodll.TNativeFoo inputFoo = new Foodll.TNativeFoo();

        //Writing input data to the native heap
                Memory dataPointer = new Memory (fooData.length * Double.SIZE);
                dataPointer.write(0, fooData, 0, fooData.length);

                inputFoo.Data = new PointerByReference();
                inputFoo.Data.setValue(dataPointer);

                outputProfile.Data = new PointerByReference();
                inputFoo.Items = outputFoo.Items = new NativeLong(fooData.length);

        //Setting some irrelevant StringArray Parameters
                inputFoo.IrrelevantStringArray = outputFoo.IrrelevantStringArray = new PointerByReference();
                inputFoo.StringsCounts = outputFoo.StringsCounts = new NativeLong(0);


        //Invocation
        foodllJnaLib.FooInterface(inputFoo, outputFoo);

        //Reading Output
                Pointer outputFooDataPointer = outputFoo.Data.getValue();

        //Reading Output alternative  1
                Double[] outDataAlt1 = outputFooDataPointer.getDoubleArray(0, outputFoo.items);

        //Reading Output alternative  2
                Double[] outDataAlt2 = new Double[outputFoo.items];
                for (int x = 0; x < outputFoo.items; x += 1)
                {
                    outDataAlt2[x] = outputFooDataPointer.getDouble(x * 8);
                }
        }
}

例外：

# A fatal error has been detected by the Java Runtime Environment:
#
#  Internal Error (0x0), pid=5904, tid=220
#
# JRE version: 7.0_04-b22
# Java VM: Java HotSpot(TM) Client VM (23.0-b21 mixed mode, sharing windows-x86 )
# Problematic frame:
# C  [KERNELBASE.dll+0xb9bc]  RaiseException+0x58
#
# Failed to write core dump. Minidumps are not enabled by default on client versions of Windows
#
# An error report file with more information is saved as:
# E:\Programme\apache-tomcat-7.0.27\bin\hs_err_pid5904.log
#

我还可以想象 32 位 JVM 和在 64 位系统上运行的本机代码存在问题。我在我的 64 位系统上安装了一个 32 位 JVM，因为我遇到了 JNA“没有找到本机 dll”的问题。

编辑：

我发现我犯了一个错误，我查看了错误的结构定义，它看起来一样，但不是。所以我必须改变我的问题。

但我也提出我的问题是错误的，我尝试了你们提出的建议。他们帮助我理解了这个话题，我在发现正确的结构定义后尝试了他们。谢谢你。

但是异常保持不变。也许我在代码中有不止一个错误。所以我的问题又来了，希望现在是正确的，并且比以前有更多的信息。不得不 Foo 一些事情（关于公司），但语义保持不变。

我真的没有 h 文件或源代码。我加入这个项目很晚，我只是在做兼职。其他人正在用 C++/CLI 对本机调用进行编程，并且成功了。所以我附上了下面的代码。c++/CLI 代码下方是更改后的 Java 代码。

c++/CLI:

typedef void (__stdcall *TFooInterface) (
   LONGLONG *APrgId,
   void *input,
   void *output,
   double value,
   HANDLE AAppHandle,
   HANDLE AProgessBar,
   char *AText, 
   char **ReturnText
   );

//-APrgID is a ID, which tells the Method, what type of processing it should do.
//-The void Pointers both take a TNativeFOO Structure. One is for the input, the other one is for the output.
//-The double-value is a Parameter for the algorithm.
//-The Handle Parameters are for a programm, the dll was created for. They give status about the processing or something like that. We now use the dll in an other context, so they aren't needed anymore.
//-char *AText amd. char **ReturnText are, also not needed anymore. They are for exception handling as far as I know. We handle the Exception in the new Context more general.

typedef struct {
   double         *Data;
   long           Items;
   unsigned char  *String;
   long           StringCounts;
   double         Value1, Value2;
   char           *FileName;
   char           *DataObject;
   char           *Comment;
   bool           Valid;
} TNativeFoo;

//-*Data is a double Array holding the data to process in TFooInterface
//-Items is the lenght of Data
//-*String and StringCounts are not needed anymore, because the dll is used in another way. I can't tell what it is for.
//-Value1, Value2 are Parameters used in the algorithm
//-FileName is for storing the result. Isn't used.
//-DataObject, Comment is also never used. I can't tell what it is for.
//-Valid is a success-flag of the algorithm.

void FooClass::FooIt(double[] FooData, HMODULE hFooLib)
{

    TNativeFoo Input, Output; //Declaration of the Structures   
    LONGLONG PrgID;
    LONGLONG ID;
    double value;

    double[] OutputData = new double[FooData.length];
    ID =-35547318716283305;

    //setting up the Structures 

    memset(&Input, 0, sizeof(TNativeFoo));
    memset(&Output, 0, sizeof(TNativeFoo));

    Output.Data = new double[FooData.length];
    Input.Data =  &(FooData[0]);
    Input.Items = Output.Items = FooData.length;
    Input.Value1 = Output.Value1 = 1.3;
    Input.Value2 = Output.Value2 = 2.3;
    Input.Valid = Output.Valid = true;

    FooInterface = (TFooInterface) GetProcAddress(hFooLib, "FooInterface");
    FooInterface(&ID, &Input, &Output, value, 0, 0, "", NULL);

    for (long i=0; i<Output.Items; i++)
    {
        OutputData[i]= Output.Data[i];
    }

    ...

}

Java 定义：

public interface Foodll extends StdCallLibrary {

    Foodll INSTANCE = (Foodll) Native.loadLibrary("foodll.dll", Foodll.class);

    //Structure
    public static class TNativeFoo extends com.sun.jna.Structure {

        public TNativeFoo(){
            setAlignType(Structure.ALIGN_NONE);
        }
        public TNativeFoo(com.sun.jna.Pointer pointer, int offset) {
            super(pointer.share(offset));
            setAlignType(Structure.ALIGN_NONE); //Tested all align-types. I was told this is the one used by the dlls
            read();
        }
        public TNativeFoo(TNativeProfile struct) {
            super(struct.getPointer());
            setAlignType(Structure.ALIGN_NONE); //Tested all align-types. I was told this is the one used by the dlls
            read();
        }
        public static class ByReference extends TNativeFoo implements com.sun.jna.Structure.ByReference {
            ByReference() {}
            ByReferenceTNativeFoo struct){super(struct.getPointer(),0);}
        }
       public static class ByValue extends TNativeFoo implements com.sun.jna.Structure.ByValue {
            ByValue() {}
            ByValueTNativeFoo struct){super(struct.getPointer(),0);}
        }
        public Pointer Data;
        public double[] getData() {
            if (this.Data == null) return null;
            return this.Data.getDoubleArray(0, Items.intValue());
        }
        public void setData(double[] data) {
            if (this.Data == null) {
                this.Data = new Memory(data.length * 8);
            }
            this.Data.write(0, data, 0, data.length);
        }
        public NativeLong Items;
        public String String;
        public NativeLong StringCounts = new NativeLong(0);
        public double Value1;
        public double Value2;
        public String FileName;
        public String DataObject;
        public String Comment;
        public boolean Valid;
    }

    NativeLong FooInterface(DoubleByReference prgid, TNativeFoo input, TNativeFoo output, double value, Pointer aAppHandle, Pointer aProgessBar, String AText, String[] ReturnText);
}

Java 调用：

public class FooInvocationClass
{
        public static FooInvocationMethod(double[] fooData, ID)
        {
                Foodll foodllJnaLib = Foodll.INSTANCE;

                Foodll.TNativeFoo outputFoo = new Foodll.TNativeFoo();
                Foodll.TNativeFoo inputFoo = new Foodll.TNativeFoo();

                inputFoo.setData(fooData);
                outputFoo.setData(new double[fooData.length]);
                inputFoo.Items = outputFoo.Items = new NativeLong(fooData.length);

                inputFoo.String = outputFoo.String = "";
                inputFoo.StringCounts = outputFoo.StringCounts = new NativeLong(0);
                inputFoo.Value1 = outputFoo.Value1 = 0.1;
                inputFoo.Value2 = outputFoo.Value2 = 0.3;

                Double ID =-35547318716283305;
                double value = 0.025;
                impDriveFiltJnaLib.ProfileFilterInterface(new DoubleByReference(ID), input, output, value, null, null, null, null);

                Double[] outDataAlt1 = outputFoo.getDate();
        }
}

编辑 2

在我解决了 dll 中的文件错误问题后，异常发生了变化。必须再次复制 dll。这解决了无法加载 dll 的问题。

#
# A fatal error has been detected by the Java Runtime Environment:
#
#  EXCEPTION_ACCESS_VIOLATION (0xc0000005) at pc=0x027047b8, pid=800, tid=7584
#
# JRE version: 7.0_04-b22
# Java VM: Java HotSpot(TM) Client VM (23.0-b21 mixed mode, sharing windows-x86 )
# Problematic frame:
# C  [FooDll.dll+0x47b8]  FooInterface+0x2288
#
# Failed to write core dump. Minidumps are not enabled by default on client versions of Windows
#
# An error report file with more information is saved as:
# E:\Programme\apache-tomcat-7.0.27\bin\hs_err_pid800.log
#
# If you would like to submit a bug report, please visit:
#   http://bugreport.sun.com/bugreport/crash.jsp
# The crash happened outside the Java Virtual Machine in native code.
# See problematic frame for where to report the bug.
#
4

2 回答 2

2

我有两个建议：-

  1. 您应该尝试使用 jnaerator 为给定的 c/c++ 标头生成 java 代码。（这将是一个很好的开始）。

  2. 根据环境，您的本机 double/float 大小将不同于 Java double/float。即本机双精度可能等于Java 浮点数。所以你也需要检查一下。

因此，您需要使用不同的结构组合对自己进行一些分析，以获得原生想要的东西。如果你不通过exact，那么它将导致崩溃。

也请发布您的答案。

于 2012-05-21T16:05:25.813 回答
1

为什么您的结构不使用对齐方式？这是您崩溃的最可能原因。验证 Java 中的字段偏移量（Structure.toString() 会告诉您）与本机代码中的字段偏移量相同。

PointerByReference旨在传递给被调用者将指针的值写入调用者传递的地址的函数。虽然您的用法在技术上是正确的，但它实际上混淆了您的实际意图。

如果您的结构包含任何类型的指针，您应该从 的类型映射开始Pointer，然后根据需要调整类型映射。 PointerByReference永远不适合作为结构字段。

   public static class TNativeFoo extends com.sun.jna.Structure {

       public TNativeFoo (){ }
       public TNativeFoo(com.sun.jna.Pointer pointer, int offset) {
            super(pointer.share(offset));
            read();
        }
        public TNativeFoo(TNativeFoo struct) {
            super(struct.getPointer());
            read();
        }
        // NOTE: use Java conventions for field names, please
        private Pointer buffer;
        public Pointer data = new Memory(Pointer.SIZE);
        public NativeLong items;
        public Pointer irrelevantStringArray = new Memory(Pointer.SIZE);
        public NativeLong stringsCounts = new NativeLong(0);

        public double[] getData() { 
            Pointer p = data.getPointer(0);
            if (p == null) return null;
            return p.getDoubleArray(items.intValue());  
        }
        public void setData(double[] data) {
            Pointer p = this.data.getPointer(0);
            if (p == null) {
               p = buffer = new Memory(data.length * 8);    
               this.data.setPointer(0, buffer);
            }
            p.write(0, data, 0, data.length);
        }
    }

    NativeLong FooInterface(TNativeFoo input, TNativeFoo output);
}

请注意， Double.SIZE 是双精度位的大小，而不是字节，因此您分配的大小是实际需要的大小的 8 倍。

于 2012-05-21T15:53:10.113 回答