CODEWITHSUNDEEP
c++zlib
Krøllebølle
Krøllebølle

Reputation: 3018

zlib error -3 while decompressing archive: Incorrect data check

I am writing a C++ library that also decompresses zlib files. For all of the files, the last call to gzread() (or at least one of the last calls) gives error -3 (Z_DATA_ERROR) with message "incorrect data check". As I have not created the files myself I am not entirely sure what is wrong.

I found this answer and if I do

gzip -dc < myfile.gz > myfile.decomp
gzip: invalid compressed data--crc error

on the command line the contents of myfile.decomp seems to be correct. There is still the crc error printed in this case, however, which may or may not be the same problem. My code, pasted below, should be straightforward, but I am not sure how to get the same behavior in code as on the command line above.

How can I achieve the same behavior in code as on the command line?

std::vector<char> decompress(const std::string &path)
{
    gzFile inFileZ = gzopen(path.c_str(), "rb");
    if (inFileZ == NULL)
    {
        printf("Error: gzopen() failed for file %s.\n", path.c_str());
        return {};
    }

    constexpr size_t bufSize = 8192;
    char unzipBuffer[bufSize];
    int unzippedBytes = bufSize;

    std::vector<char> unzippedData;
    unzippedData.reserve(1048576); // 1 MiB is enough in most cases.

    while (unzippedBytes == bufSize)
    {
        unzippedBytes = gzread(inFileZ, unzipBuffer, bufSize);

        if (unzippedBytes == -1)
        {
            // Here the error is -3 / "incorrect data check" for (one of) the last block(s)
            // in the file. The bytes can be correctly decompressed, as demonstrated on the
            // command line, but how can this be achieved in code?
            int errnum;
            const char *err = gzerror(inFileZ, &errnum);
            printf(err, "%s\n");
            break;
        }

        if (unzippedBytes > 0)
        {
            unzippedData.insert(unzippedData.end(), unzipBuffer, unzipBuffer + unzippedBytes);
        }
    }

    gzclose(inFileZ);
    return unzippedData;
}

Upvotes: 0

Views: 1697

Answers (1)

Mark Adler
Mark Adler

Reputation: 112259

First off, the whole point of the CRC is to detect corrupted data. If the CRC is bad, then you should be going back to where this file came from and getting the data not corrupted. If the CRC is bad, discard the input and report an error.

You are not clear on the "behavior" you are trying to reproduce, but if you're trying to recover as much data as possible from a corrupted gzip file, then you will need to use zlib's inflate functions to decompress the file. int ret = inflateInit2(&strm, 31); will initialize the zlib stream to process a gzip file.

Upvotes: 2

Related Questions