Talk about the bottom playing method of C ＃ method overloading

Recently in to see C++ Method overload , I was thinking. C# How to play the overloaded bottom layer in , A lot of friends should know C Overloading is not supported , For example, the following code will report an error .

#include <stdio.h>

int say() {
 return 1;
}
int say(int i) {
 return i;
}

int main()
{
 say(10);
 return 0;
}

From the error message , It said say Methods already exist , embarrassed ...

One ： Why? C I won't support it

To find the answer , Need to know a little bit of the underlying knowledge , That is, the compiler is compiling C Method will Function name Add as symbol to The symbol table in , This The symbol table Namely call To say Method bytecode   A carrier in the middle , Draw a picture, that's about it .

In short ,call Jump to the first The symbol table , And then again jmp To say Method , Here's the problem , The symbol table Is a kind of dictionary structure , Can't appear Symbol The same thing . by the way , stay windbg We can use x Command to search for these symbols ,

To prove my point , You can verify it at the assembly level , Modify the code as follows ：

#include <stdio.h>

int say(int i) {
 return i;
}

int main()
{
 say(10);
 return 0;
}

Next, let's look at the compilation .

--------------- say(10) -----------

00C41771  push        0Ah  
00C41773  call        _say (0C412ADh)  

---------------  The symbol table  -----------

00C412AD  jmp         say (0C417B0h)  

--------------- say body -----------

00C417B0  push        ebp  
00C417B1  mov         ebp,esp  
00C417B3  sub         esp,0C0h  
00C417B9  push        ebx  
00C417BA  push        esi  
00C417BB  push        edi  
00C417BC  mov         edi,ebp  
00C417BE  xor         ecx,ecx  
00C417C0  mov         eax,0CCCCCCCCh  
00C417C5  rep stos    dword ptr es:[edi]  
00C417C7  mov         ecx,offset [email protected] (0C4C008h)  
...

After knowing the principle , Let's see C++ How is the The symbol table To achieve a unique breakthrough .

Two ：C++ Symbol table breakthrough

For the convenience of narration , Let's start with a paragraph C++ Method overloaded code .

using namespace std;

class Person
{
public:
 void sayhello(int i) {
  cout << i << endl;
 }
 void sayhello(const char* c) {
  cout << c << endl;
 }
};

int main(int argc)
{
 Person person;

 person.sayhello(10);
 person.sayhello("hello world");
}

Logically speaking sayhello There are many. , There must be no breakthrough , With curiosity, let's take a look at its disassembly code .

----------     person.sayhello(10);  ----------------

003B2E5F  push        0Ah  
003B2E61  lea         ecx,[person]  
003B2E64  call        Person::sayhello (03B13A2h) 

------------  person.sayhello("hello world"); ----------------

003B2E69  push        offset string "hello world" (03B9C2Ch)  
003B2E6E  lea         ecx,[person]  
003B2E71  call        Person::sayhello (03B1302h)

From the perspective of assembly code , All of them Person::sayhello This symbol , Strangely, they belong to different addresses : 03B13A2h, 03B1302h, This is so strange , ha-ha , Dictionary class symbol table There must be no problem , The problem is Visual Studio 20222 The Disassembly window does some internal conversion during debugging , It has blinded our eyes ,

It's really annoying ！！！ Even the runtime assembly code is not thorough enough , Now how can we continue to dig ？ It can be used IDA Go and see the program Static disassembly code , The screenshot is as follows ：

You can clearly see from the comments on the code , original :

1. Person::sayhello(int) Turned into  [email protected]@@[email protected].

2. Person::sayhello(char const *) Turned into  [email protected]@@[email protected]

Here we finally find out , original C++　 To support method overloading , take Method name Recoded , This can really break through The symbol table The uniqueness limit of .

3、 ... and ：C# How to achieve breakthrough

We all know C# The bottom of the CLR By C++ Written , So the big probability game is the same , Next, the previous code ：

internal class Program
    {
        static void Main(string[] args)
        {
   // Make a deliberate repetition 
            Say(10);
            Say("hello world");

            Say(10);
            Say("hello world");
            Console.ReadLine();
        }

        static void Say(int i)
        {
            Console.WriteLine(i);
        }

        static void Say(string s)
        {
            Console.WriteLine(s);
        }
    }

because C# The method is by JIT Dynamically compiled at run time , And the first compilation method will jump to JIT Pile address of , So the breakpoint must be placed on the second call Say(10) To see the symbolic address of the method , The assembly code is as follows ：

----------- Say(10); -----------

00007FFB82134DFC  mov         ecx,0Ah  
00007FFB82134E01  call        Method stub for: ConsoleApp1.Program.Say(Int32) (07FFB81F6F118h)  
00007FFB82134E06  nop  

----------- Say("hello world");  -----------

00007FFB82134E07  mov         rcx,qword ptr [1A8C65E8h]  
00007FFB82134E0F  call        Method stub for: ConsoleApp1.Program.Say(System.String) (07FFB81F6F120h)  
00007FFB82134E14  nop

From the output information , Also two symbol table addresses , Then the symbol table addresses jmp To the last method body .

----------- Say(10); -----------
00007FFB82134E01  call        Method stub for: ConsoleApp1.Program.Say(Int32) (07FFB81F6F118h)  

-----------  The symbol table  -----------
00007FFB81F6F118  jmp         ConsoleApp1.Program.Say(Int32) (07FFB82134F10h)  

----------- Say body -----------

00007FFB82134F10  push        rbp  
00007FFB82134F11  push        rdi  
00007FFB82134F12  push        rsi  
00007FFB82134F13  sub         rsp,20h  
00007FFB82134F17  mov         rbp,rsp  
00007FFB82134F1A  mov         dword ptr [rbp+40h],ecx  
00007FFB82134F1D  cmp         dword ptr [7FFB82036B80h],0  
00007FFB82134F24  je          ConsoleApp1.Program.Say(Int32)+01Bh (07FFB82134F2Bh)  
00007FFB82134F26  call        00007FFBE1C2CC40

I don't know how to look at it yet JIT After the change of name Method name , You can leave a message if you know , But on the whole it's still C++ A set of .

OK, that's all for this article , I hope it helps you .