SharpDevelop Community

I've created test folder with few empty files. size of each file = 0 bytes, names variant: russian and latin, while zipping files to archive i found a problem with crc cheching: ZipEntry.Name = "новый файл.txt", crc = 3809153438. Such crc is more than max acceptable value for Int32, so '(int)crc' throwed an exception and zip process failed.

I wasn't able to fix this problem with simple changing int32->uint32 because of trouble in reading crc

Sorry, size of those problem file was 6 bytes, text file filled with few latin symbols

I think you can solve this problem in such way:
save CRC32 as uint (you've got some WriteLeUint functions inside ZipLib),
read them in such way:

uint crc2 = this.IntAsUint(inputBuffer.ReadLeInt());

where
private uint IntAsUint(int i)
{
    return (uint)((0x80000000 & i) | (i & 0x7FFFFFFF));
}

this function converts int32->uint32 without using .NET convetions. That's something like C++: UINT ui = *((UINT*)&i))

You have to compile SharpZipLib with integer overflow checking disabled.

I had the same problem in the ZipInputStream.cs and ZipOutputStream.cs CRC's. 

In the ZipInputStream public ZipEntry GetNextEntry() I added

if ((flags & 8) == 0) {    entry.Crc = crc2 & 0xFFFFFFFFL;
   entry.Size = size & 0xFFFFFFFFL;
   entry.CompressedSize = csize & 0xFFFFFFFFL;
   int checkVal = (crc2 >> 24);
   if (checkVal < 0) checkVal = byte.MaxValue + 1 + checkVal; // this line
   entry.CryptoCheckValue = Convert.ToByte(checkVal); ;


and in the ZipOutputStream.cs public void CloseEntry()  I added

// Patch the header if possible
if (patchEntryHeader == true) {
    patchEntryHeader = false;
    long curPos = baseOutputStream.Position;
    baseOutputStream.Seek(crcPatchPos, SeekOrigin.Begin);
    int crcInteger = (int) (curEntry.Crc & int.MaxValue); // this line
    WriteLeInt(crcInteger);



The trick here is to force the CRC into the integer type that is required by 2s complement logic or using a logical AND (&) to lop off the extra bits. 

I hope this helps.