Issue about Globalvariable names

[seviezhou]

I use bin2llvmir to translate an ELF file and I get Globalvariables look like:

@global_var_6dbd10.1702 = global [2 x i8] c"-\00" 
@global_var_6dc940.2028 = global [2 x i8] c" \00"  
@global_var_6dc942.2029 = global [2 x i8] c" \00"

They have numbers in the terminal, it is because these Globalvariables are create in same names more than once. After looking into bin2llvmir/utils/ir_modifier.cpp file, I think I can add a single line of code and solve this problem. In function IrModifier::getGlobalVariable , and add a new piece of code after line 687:

		auto* conv = IrModifier::convertConstantToType(ngv, gv->getType());
		if (conv != gv)
		{
			gv->replaceAllUsesWith(conv);
		}
		gv->eraseFromParent();
		gv = ngv;

+		ngv->setName(name);
        }

Then we can get:

@global_var_6dbd10.1702 = global [2 x i8] c"-\00" 
@global_var_6dc940.2028 = global [2 x i8] c" \00"  
@global_var_6dc942.2029 = global [2 x i8] c" \00"

[s3rvac]

Hi! The current and new outputs look identical (see the two red rectangles below). Could you please verify that the new output is correct?

Screenshot:

[seviezhou]

My mistake, I wrongly paste the original output, the new is here:

@global_var_6dbd10 = global [2 x i8] c"-\00"
@global_var_6dc940 = global [2 x i8] c" \00"
@global_var_6dc942 = global [2 x i8] c" \00"

[s3rvac]

Thank you. @PeterMatula, what do you think about this?

[xkubov]

Hi, I've tried the change in ir_modifier.cpp suggested by @seviezhou and all regression tests will pass. Furthermore, I've looked into decompilation output of some regression tests and decompilation seems to be improved because now in some cases the C code output includes comments about the address of a global variable.

For example you can see the code taken from decompilation of bashbot sample in retdec-regression-tests:

before modification:

  // --------------------- Global Variables ---------------------
  
  int32_t g1 = 0; // ebp
  int32_t g2 = 0; // ebx
  int32_t g3 = 0; // edi         
  int32_t g4 = 0; // esi
  int32_t g5 = 0; // esp
  int32_t g6 = 0x200072;
  char (*g7)[20] = "188.209.52.143:8080";
  int32_t g8 = -1;
  uint32_t g9 = 0x587c4;
  int32_t g10 = 4095;
  char * g11;
  int32_t g12 = 0;
  int32_t g13 = 0;
  char * g14;
  int32_t g15 = 0;
  int32_t g16 = 0;
  int32_t g17 = 0;
  char * g18;         
  char g19 = 0;
  char g20 = 0;
  char g21 = 0;
  char g22 = 0;
  char g23 = 0;
  int32_t g24 = 0;
  int32_t g25 = 0;
  int32_t g26 = 0;
  int32_t g27;
  int32_t g28;

after modification:

 // --------------------- Global Variables ---------------------
  
  int32_t g1 = 0; // eax
  int32_t g2 = 0; // ebp
  int32_t g3 = 0; // ebx
  int32_t g4 = 0; // edi
  int32_t g5 = 0; // esi
  int32_t g6 = 0; // esp
  int32_t g7 = -1; // 0x804f1e0
  int32_t g8 = 0x200072; // 0x804f853
  int32_t g9 = -1; // 0x8051058
  bool g10 = true; // 0x8051060
  int32_t g11 = 0; // 0x8051064
  char (*g12)[20] = "188.209.52.143:8080"; // 0x8051260
  int32_t g13 = -1; // 0x8051264
  uint32_t g14 = 0x587c4; // 0x8051364
  int32_t g15 = 4095; // 0x8051368
  char g16 = 0; // 0x8051380
  int32_t g17 = 0; // 0x8051384
  char * g18; // 0x80513a0
  int32_t g19 = 0; // 0x80513a8
  int32_t g20 = 0; // 0x80513ac
  char * g21; // 0x80513b0
  int32_t g22 = 0; // 0x80513c0
  int32_t g23 = 0; // 0x80513c4
  int32_t g24 = 0; // 0x80513c8
  char * g25; // 0x80553c0
  char g26 = 0; // 0x80553c5
  char g27 = 0; // 0x80553c6
  char g28 = 0; // 0x80553c7
  char g29 = 0; // 0x80553c8
  char g30 = 0; // 0x80553c9
  int32_t g31 = 0; // 0x80553cc
  int32_t g32 = 0; // 0x80553d0
  int32_t g33 = 0; // 0x80553d4
  int32_t g34;
  int32_t g35;

[PeterMatula]

Looks ok. Will you open a Pull Request, or should we do the change? No problem either way.

[PeterMatula]

Fixed. Regression test added in avast/retdec-regression-tests@7523176.