Borland C++ crazy Access violation error

[Matej Tomcik]

I have a project in Borland C++ and few days ago a strange error started to appear. Everywhere I use std::strcmp, I get an exception "Access violation at address XXXXXXX. Write of address 0000000" during runtime. Usually I need to compare data retrieved from MySQL result using my SQL wrapper.
Example:

CMySQLResult * res = CConnection::GetSingletonPtr()->RunQuery(CSQLBuilder::GetSingletonPtr()->GetTemplate("objednavky/prijate/profily/nahlad"));
for (int i=0;i<res->RowsCount();i++)
{
  TListItem * itm = list->Items->Add();
  itm->Caption = (*res)[i][3];
  itm->SubItems->Add(COMMAREPL((*res)[i][4],".",","));
  itm->SubItems->Add(COMMAREPL((*res)[i][5],".",",")+AnsiString(std::strcmp((*res)[i][24],"1") == 0 ? " m" : " ks"));
  // ...
}
delete res;
 

I'm sure that (*res)[i][j] always returns valid value, even if the result doesn't contain such an index or it cointains NULL value, returned value will always be not-null. But strcmp always fails...

EDIT: When running another part of the code I got this exceptions: "Privileged instruction" or "Control has no parent", all raised when strcmp was called.

[weaknessforcats]

This code:

CMySQLResult * res = CConnection::GetSingletonPtr  etc...

says res is a pointer to a MySQLResult.

whereas this code:

itm->Caption = (*res)[i][3]; 

says that res is a pointer to and array element but I don't see the array declared. Therefore y are pointing to indeterminate memory. Tha is, memory that is not yours and therefore, your access violation. It does not appear strcmp is your culprit.

[Matej Tomcik]

No, you got it wrong.
CMySQLResult contains operator [] which returns an array of data holders, and the second [] operator then returns actual value obtained from MySQL. The code I posted is working.

Please, read the description, when I say the error always comes from strcmp, it just comes from strcmp. And as I posted earlier, it is not just access violation, I would handle that kind of exception as I always do, but it raises other strange exception, again, only at the point when application calls strcmp.

[weaknessforcats]

It's hard to know that CMySQLResult is a class. The posted code says it's a pointer. But that aside, you are saying then that
(*res)[i][24] is a const char* that always points to a valid C-string?

std::strcmp((*res)[i][24],"1") 

One thing I would do is follow this call using a debugger.

[Matej Tomcik]

Yes, CMySQLResult is a class, (*res) dereferences the object.
And yes, it always points to a valid string (it's always char * string, no matter if the indexes are correct or not, at least it points to "\0" const string)

I would not post here if I knew the reason why is this happening. It never happened before. For example, I wrote this code:

if (std::strcmp("ab","ab") == 0)
    ShowMessage("Equals");
 

Even this results in an access violation error, is there anything wrong with these 2 lines of code?

[weaknessforcats]

Your code compiles and executes OK using Visual Studio.NET 2008.

I just changed ShowMessage to cout.

So your error may be not in the code but in some project setting. Like CMySQLResult points to data that is not part of your process's address space. Maybe you have a multi-process program and are not sharing memory.

Have you been able to cout << (*res)[i][24]?

[Matej Tomcik]

Unfortunately, I use Borland C++ for this project. As I mentioned, it started to happen only few days ago. CMySQLResult works with libmysql.dll, you may be right now but I don't quite understand why, it used to work as expected, why so sudden...
I think it has something to do with VCL.
I avoid using strcmp now, I just take the first parameter, cast it to AnsiString and directly compare to the second param and it works.

[weaknessforcats]

Then you need to insert a srcmp call as an individual line that uses the same call arguments as the one in the code now. Then set a breakpoint and step in using your debugger.

The cout suggestion is a uickie debug but you should be using your debugger now.