Question

I have a function that iterates through a const char * and uses the character to add objects to an instance of std::map if it is one of series of recognized characters.

#define CHARSEQ const char*

void compile(CHARSEQ s) throw (BFCompilationError)
{
     std::cout << "@Receive call " << s << std::endl;

     for(int i = 0; s[i] != '\0'; i++)
     {
         if (std::string("<>-+.,[]").find_first_of(s[i]) == std::string::npos)
         {
             throw BFCompilationError("Unknown operator",*s,i);
         }
         std::cout << "@Compiling: " << s[i] << std::endl;
         std::cout << "@address s " << (void*)s << std::endl;
         std::cout << "@var s " << s << std::endl;
         controlstack.top().push_back(opmap[s[i]]);
     }
 }

The character sequence passed is "++++++++++."
For the first three iterations, the print statements display the expected values of ‘+’, ‘+’, and ‘+’, and the value of s continues to be "+++++++++++.". However, on the fourth iteration, s becomes mangled, producing bizarre values such as 'Ð', 'öê', 'cR ', 'œk' and many other character sequences. If the line that throws the exception is removed and the loop is allowed to continue, the value of s does not change after again.

Other functions have access to s but since this is not a multithreaded program I don’t see why that would matter. I am not so much confused about why s is changing but why it only changes on the fourth iteration.

I have searched SO and the only post that seems at all relevant is this one but it still doesn’t answer my question. (Research has been difficult because searching "const char* changing value" or similar terms just comes up with hundreds of posts about what part of is is const).

Lastly, I know I should probably be using std::string, which I will if no answers come forth, but I would still like to understand this behavior.

EDIT:

Here is the code that calls this function.

CHARSEQ text = load(s);

std::cout << "@Receive load " << text << std::endl;

try
{
    compile(text);
}
catch(BFCompilationError& err)
{
    std::cerr << "\nError in bf code: caught BFCompilationError @" << err.getIndex() << " in file " << s << ":\n";
    std::cerr << text << '\n';
    for(int i = 0; i < err.getIndex(); i++)
    {
        std::cerr << " ";
    }
    std::cerr << "^\n";
    std::cerr << err.what() << err.getProblemChar() << std::endl;
    return 1;
}

Where load is:

CHARSEQ load(CHARSEQ fname)
{
    std::ifstream infile (fname);
    std::string data(""), line;
    if (infile.is_open())
    {
        while(infile.good())
        {
            std::getline(infile,line);
            std::cout << "@loading: "<< line << '\n';
            data += line;
        }
        infile.close();
    }
    else
    {
        std::cerr << "Error: unable to open file: " << fname << std::endl;
    }

    return std::trim(data).c_str();
}

and the file fname is ++++++++++. spread such that there is one character per line.

EDIT 2:

Here is an example of console output:

@loading: +
@loading: +
@loading: +
@loading: +
@loading: + 
@loading: +
@loading: +
@loading: +
@loading: +
@loading: +
@loading: .
@Receive load ++++++++++.
@Receive call ++++++++++.
@Compiling: +
@address s 0x7513e4
@var s ++++++++++.
@Compiling: +
@address s 0x7513e4
@var s ++++++++++.
@Compiling: +
@address s 0x7513e4
@var s ++++++++++.
@Compiling: 
@address s 0x7513e4
@var s ßu

Error in bf code: caught BFCompilationError @4 in file bf_src/Hello.txt:
ßu
^
Unknown operatorß

Answer 1

Your load function is flawed. The const char* pointer returned by c_str() is valid only until the underlying std::string object exists. But data is a local variable in load and is cleared after return. Its buffer is not overwritten by zeroes but left as it were as free memory. Therefore printing out the value immediately after returning is likely to work but your program may put new values there and the value pointed by your pointer will change.

I suggest to use std::string as the return value of load as a workaround.