Re: tellg modifying the state of an input file stream

[Ivan Kolev <ikolev at abac dot com>]

Paolo Carlini wrote:
> The correct output, which I actually get on x86-linux (and all the other 
> arches I have available), would be:
>
> c = T, pos: 1
> c = e, pos: 2
> c = s, pos: 3
> c = t, pos: 4
> c =
> , pos: 5
>
> if you are really seeing something different, then it's something MinGW 
> specific and I would suggest reporting it to the maintainers of that 
> project, to begin with, because if you are *really* seeing the first 
> output above, then something is *badly* broken for that target and can't 
> be a bug in the generic code (nobody would use our library for serious 
> projects ;)

Hi Paolo,

There is one possibility for the problem to be in generic code and that it hasn't been spotted until now - it seems the problem is related to the \r\n line endings on Windows (you mention x86 *linux*, and it seems most GCC users are on Unix systems). A few more tests showed the following dependency: the skipping of the offset after each read is related to the number of newlines remaining till the end of the file. E.g., if I change line <1> in the program to the following
	ofs << "Test" << endl << "Test" << endl;
the result is

c = T, pos: 3
c = t, pos: 6
c = T, pos: 8
c = s, pos: 10
c =
, pos: 12

Unfortunately I cannot debug the example myself because I'm not very experienced with GCC and GDB and I cannot make the CodeBlocks IDE step into libstdc++'s code.

Thanks,
Ivan