compiler warnings om VS.NET 2003

[John-Philip Leonard Johansson]

Hi!

When compiling the latest stable release I get a few warnings on tinyxmlparser.cpp, I've checked against the latest CVS upload and their still there. I'll give the line numbers for the two versions like so:
[old] means the latest stable release
[new] means the latest commit in the cvs

All warnings are implicit type conversions that can be solved with static_cast, but some might have another way to avoid the type conversion entirely?

The warnings are:
..::[ WARNING 1 ]::..
warning C4244: 'argument' : conversion from 'std::basic_istream<_Elem,_Traits>::int_type' to 'char', possible loss of data
        with
        [
            _Elem=char,
            _Traits=std::char_traits<char>
        ]

Line:
*tag += in->get();
[old] 177
[new] 306

..::[ WARNING 2 ]::..
warning C4244: 'argument' : conversion from 'int' to 'char', possible loss of data

Line:
*tag += c;
[old] 190
[new] 320

..::[ WARNING 3-5 ]::..
warning C4244: '+=' : conversion from 'int' to 'char', possible loss of data

Lines:
*value += ( tolower( *(p+3) ) - 'a' + 10 );
[old] 241, 249 and 252
[new] gone?

..::[ WARNING 6 ]::..
warning C4267: 'initializing' : conversion from 'size_t' to 'int', possible loss of data

Line:
int tagIndex = tag->length();
[old] 397
[new] 561

..::[ WARNING 7 ]::..
warning C4267: 'initializing' : conversion from 'size_t' to 'int', possible loss of data

Line:
int tagIndex = tag->length();
[old] 638
[new] 810

..::[ WARNING 8-9 ]::..
warning C4244: 'argument' : conversion from 'int' to 'char', possible loss of data

Lines:
*tag += c;
[old] 653 and 669
[new] 825 and 841

..::[ WARNING 10-13 ]::..
warning C4244: 'argument' : conversion from 'int' to 'char', possible loss of data

Lines:
(*tag) += c;
[old] 878, 930, 1049 and 1079
[new] 1060, 1112, 1231 and 1261

[Lee Thomason]

Thanks! Fixed.

[B Sizer]

The first two warnings, and possibly the last two, are indicative of a slightly inappropriate use of get(). In particular, the chance of the stream returning EOF is being ignored, and therefore the EOF character - if encountered - would get appended onto the data before the reading loop terminates. (Assuming there is a separate termination condition.)

[Lee Thomason]

A very good point on the EOF issue! Definately worth looking in to.

I checked the STL docs:

"bool good ( ) const; Check if stream is good for i/o operations.  The function returns true if none of the stream's error flags are set."

I did a search on the "get" calls in the parser and looked for issues.

checks for good():
TiXmlBase::StreamTo
TiXmlBase::StreamWhiteSpace
TiXmlDocument::StreamIn
TiXmlUnknown::StreamIn
TiXmlComment::StreamIn
TiXmlText::StreamIn
TiXmlDeclaration::StreamIn

needed a fixup:
TiXmlElement::StreamIn

With the fix - and assuming I understand good() - I think it's in pretty good shape. Do you agree on good() protecting against the EOF?

lee

[B Sizer]

Bear in mind that 'The function returns true if none of the stream's error flags are set'. And the only thing that can set those flags is a failed I/O operation, which in these cases is going to be a get() I believe. I heard that this is the case in C and C++ for the sake of efficiency (no hidden peek() calls to check the next operation will work).

Easiest to show this with an example:
#include <fstream>
#include <iostream>

using namespace std;

int main(int argc, char *argv[])
{
  fstream file("test.txt");

  while (file.good())
  {
    int c = file.get();
    cout << "Read character with value: " << c << endl;
  }
   
  return 0;
}

Create test.txt, fill it with a word or two, then watch the output. Look for the -1 at the end.

[Lee Thomason]

Very good point. I'll change the parsing code to check for the eof().

lee