Issue with decompressing data with original size 73080 bytes

[ragkousism]

I have a gzip data array in C which has original size 73080 and compressed 34703. Based on your examples I wrote the following code to decompress the file

#define U32 uint32_t
#define U16 uint16_t
#define U8 uint8_t
#define OUT_CHUNK_SIZE 1

int gzip_uncompress(U8 *dest,
                    U32 *destLen,
                    const U8 *source,
                    U32 sourceLen)
{
  U32 dlen = 0;
  U32 gz_len = 0;
  TINF_DATA decomp;
  int res = 0;
  /* Decompress */
  uzlib_init();

  /* Length of gz file or buffer in our case */
  gz_len = sourceLen;

  /* /\* Take the decompressed length from the last bytes *\/ */
  /* dlen = (buffer[len - 1] << 24) | (buffer[len - 2] << 16) | (buffer[len - 3] << 8) | buffer[len - 4]; */
  dlen =            source[gz_len - 1];
  dlen = 256*dlen + source[gz_len - 2];
  dlen = 256*dlen + source[gz_len - 3];
  dlen = 256*dlen + source[gz_len - 4];

  *destLen = dlen;
  /* Out length must always be +1 for the decompression to work properly */
  dlen++;
  dest = calloc(dlen, sizeof(U8));
  if (dest == NULL)
    {
      printf("Error allocating memory\n");
      return -1;
    }

  uzlib_uncompress_init(&decomp, NULL, 0);

  decomp.source = source;
  decomp.source_limit = source + gz_len - 4;

  res = uzlib_gzip_parse_header(&decomp);
  if (res != TINF_OK)
    {
      printf("Error parsing header: %d\n", res);
      return -1;
    }

  decomp.destStart = decomp.dest = dest;
  /* Decompress byte by byte */
  /* decomp.destSize = 1; */

  while (dlen)
    {
      unsigned int chunk_len = dlen < OUT_CHUNK_SIZE ? dlen : OUT_CHUNK_SIZE;
      decomp.destSize = chunk_len;
      res = uzlib_uncompress_chksum(&decomp);
      dlen -= chunk_len;
      if (res != TINF_OK)
        {
          break;
        }
    }

  if (res != TINF_DONE)
    {
      printf("Error during decompression: %d\n", res);
      return -res;
    }
}

int main() 
{
...
gzip_uncompress (dest_buf,
                       &dest_buf_size,
                       source_buf,
                       source_buf_size);
...
}

The problem is that one KB before the end, uzlib_uncompress exits with TINF_DONE and the crc check obviously fails. I wrote a test with exactly the same code on a smaller file and it woks without an issue. Is there some upper limit on the files the library can decompress? Any ideas what I am doing wrong?

[ragkousism]

Forgot to add the gzip in question

file.gz

[pfalcon]

An obvious test is running this file thru the included "tgunzip" util. With the current master, 35e9c23, the result is the file of size 73080 with md5sum of 331c99fc2898a4f360a790d45989084d - the same as when decompressing with gunzip.

Apparently, your code has some bug. You would need to compare it with tgunzip's code and see what's wrong.

[ragkousism]

Hello @pfalcon yes I did try with tgunzip and it worked. The problem was not the code above. The code is correct. The issue is that before calling gzip_uncompress() the buffer I was writing the file, was 20 bytes sorter and as a result I was passing an invalid size to the library. The strange thing is that it didn't segfault on normal operation but only if I ran it with valgrind. Anyhow the case has nothing to do with the library. I am sorry for the inconvenience.