AssignmentExpression assert nullptr Error about medusa HOT 8 CLOSED

I am preventing the abortion with this code:

    Expression::SPType operand0 = rInsn.GetOperand(0);
          Expression::SPType operand1 = rInsn.GetOperand(1);

          if (operand1 == nullptr) {
              Log::Write("operand 1 is null");
              return false;
          }

          auto pExpr1 = /* Semantic: if zf.id == int1(1): op0.val = op1.val */
          Expr::MakeIfElseCond(
            ConditionExpression::CondEq,
            Expr::MakeId(X86_FlZf, &m_CpuInfo),
            Expr::MakeConst(1, 0x1),
            Expr::MakeAssign(operand0, operand1),
            nullptr);
          AllExpr.push_back(pExpr1);
          rInsn.SetSemantic(AllExpr);

in the following file: src/arch/x86/x86_opcode.cpp

from medusa.

It may looks radical, but this behavior is desired. Just attach a debugger and look in the call stack to find the buggy instruction.
I can take a look on this bug if you can share your executable. :)

from medusa.

Well, how could I best share this exe with you?

from medusa.

4shared.com is ok I guess.

from medusa.

I really think that the code of the file has something wrong. At least if I look at this: http://pedump.me/a772ef7449783ee4edc60dd0b006ffe6/#disasm

from medusa.

Thanks for sharing, the buggy instruction was cmpxchg8b, the fix is 1d61c8e

from medusa.

Hey I've just tested your fix. It works :)
Nice job.
How do you test these things? Just debugging the application or using something different?

from medusa.

Thanks, it's actually pretty easy to debug these issues:

• start qMedusa with a debugger,
• wait for the assert to make the application crash,
• the debugger catches this exception/signal and stops,
• display the callstack
• find the faulty instruction
• ???
• profit :)

from medusa.