Custom archive format File Reading

Question

C#.NET 4.0

I'm having an interesting problem here with reading a custom file archive format. In C#, I wrote a program that creates an archive header (some overhead info about the archive as a whole, number of files, those kinds of things). It then takes an input file to be stored, reads and bytes, and then writes some overhead about the file (filename, type, size and such) and then the actual file data. I can also extract files from the archive through this program. To test it, I stored a png image and extracted it by reading the filesize from the overhead and then allocating an array of bytes of that size, pulled the filedata into that array, and then wrote it with a streamwriter. No big deal, worked fine. Now, we go to the C++ side...

C++

My C++ program needs to read the filedata in, determine the filetype, and then pass it off to the appropriate processing class. The processing classes were giving errors, which they shouldn't have. So I decided to write the filedata out fro the C++ program after reading it using fwrite(), and the resulting file appears to be damaged? In a nutshell, this is the code being used to read the file...

unsigned char * data = 0;
char temp = 0;
__int64 fileSize = 0;
fread(&fileSize, sizeof(__int64), 1, _fileHandle);
data = new unsigned char[fileSize];

for (__int64 i = 0; i < fileSize; i++)
{
    fread(&temp, 1, 1, _fileHandle);
    data[i] = temp;
}

(I'm at work right now, so I just wrote this from memory. However, I'm 99% positive it's accurate to my code at home. I'm also not concerned with non MS Standards at the moment, so please bear with the __int64.)

I haven't gone through all 300 something thousand bytes to determine if everything is consistent, but the first 20 or so bytes that I looked at appear to be correct. I don't exactly see why there is a problem. Is there something funny about fread()? I also to double check the file in the archive, removed all the archive overhead and saved just the image data to a new png image with notepad, which worked fine.

Should I be reading this differently? Is there something wrong with using fread() to read in this data?

Answer 1

Given that the first n bytes appear to be correct, did you by chance forget to open the file in binary mode ( "rb" )? If you didn't then it's helpfully converting any sequences of \\r\\n into \\n for you which would obviously not be what you want.

Since this question is tagged C++ did you consider using the canonical C++ approach of iostreams rather than the somewhat antiquated FILE* streams from C?