Pointers...

[hne]

Hello gurus,
I have a code I inherited that is loaded with pointers. Here is my problem. The application is reading a stream file to get the data. There is a delimiter 'xle' or 'xlc' or 'eof' that let the app know when it is at the end of a block. The app uses a 'get(char*, n, c)' function to get the data. However, when the app tries to determine the 'strlen' or 'strcmp' the pointer. I get an Access Violation'. Example below. Please help. Thx.

char IVectrfile::getField(char **saveaddr = NULL)
{
long curloc = tellg();
int n = 0;
bool done = false;
char c;
char *dat;
while ( !done )
{
c = ifstream::get();
if((c == '\x1c') || (c == '\x1d') || (c == '\x1e') || (c == EOF))
done = true;
n++; // count will include the terminator
}
if (n > 1) // data other than the terminator
{
seekg(curloc);
dat = new char[n];
//copy the field to the memory up to the trmator
get(dat, n, c); // end the field with \0
ifstream::get(); // get rid of the terminator
// put the address into saveaddr
*saveaddr = dat;
}
}
...
...
char ** flds = 0;
if (pTf->WantThisData(flds))
{
// if 'all passengers' add it to the pnr ssrlist
// else find the passenger(s) and add it to them
findAndAddData(flds);
}
clear(flds);
...
...
int WantThisData(char ** ssrInfo)
{
/***************************************************
The application fails on the 'if' statement below
***************************************************/
// esrDataNo is a 'typedef enum' value is 6
if ((strlen(ssrInfo[esrDataNo]) > 0)
&& (strcmp(fltno, ssrInfo[esrDataNo])))
return 0;

return 1;
}

 

[ArkM]

Of course, it was not the best idea to inherit IVectfile from std::ifstream, but one of possible problem causes is here:

   char ** flds = 0;
   if (pTf->WantThisData(flds))
...
int WantThisData(char ** ssrInfo)
{
   /***************************************************
    The application fails on the 'if' statement below
   ***************************************************/
   // esrDataNo is a 'typedef enum' value is 6
   if ((strlen(ssrInfo[esrDataNo]) > 0)
      && (strcmp(fltno, ssrInfo[esrDataNo])))

It's obviously the program tries to get a string address via null pointer...
I don't know why WantThisData called as a member function (via pTf pointer) but no class prefix in its definition presented. May be you extract the definition from a class declaration...

Regrettably, it's not the only defect of this snippet. For example:
1. It's a wrong way to test C++ istream eof condition:

char c;
...
c = ifstream::get(); // what's a dirty trick...
if(... || (c == EOF))

Never mix C and C++ stream stuff! The EOF is the C library macros with int type value (as usually -1). Now the program compares char and int values. If a C++ compiler treats char type as unsigned (it's conforming with C++ Standard implementation), c is NEVER equal to EOF!

In actual fact istream::get() returns char_traits<char>::int_type type value and the eof condidition signalled with char_traits<char>::eof() value. That's why better use get(c) member function: it returns istream& and it's easy to detect not only eof but fail consition too (see below).

Moreover: no any special EOF char in the stream. You can't skip or ungetc this phantom after eof condition.

The right and portable method (one of) in that case:

if (!ifstream::get(c))
{
   // eof or i/o error... break the loop...
}
else // now chech for other real delimiters...
...

Now Coup de Grace:
After end-of-file you (or your predecessor) can't do any stream ops until clear the stream. So seekg and the last gets do nothing in that case - no data in dat buf at all!
You must clear the stream state with clear() member function.

Apropos, why ifstream::get() but simply seekg(x) and get(x,y,z) used in this unprepossessing ifstream bastard?

I hope, you can make proper corrections of this strange code...