diff --git a/megawarc b/megawarc
index 0340bd6..d9d6449 100755
--- a/megawarc
+++ b/megawarc
@@ -65,6 +65,7 @@ import gzip
 import json
 import os.path
 import re
+import subprocess
 import sys
 import tarfile
 import zlib
@@ -85,10 +86,11 @@ def copy_to_stream(stream, input_filename, offset, size):
     while to_read > 0:
       buf_size = min(to_read, 4096)
       buf = f.read(buf_size)
-      if len(buf) < buf_size:
-        raise Exception("End of file: %d bytes expected, but %d bytes read." % (buf_size, len(buf)))
+      l = len(buf)
+      if l < buf_size:
+        raise Exception("End of file: %d bytes expected, but %d bytes read." % (buf_size, l))
       stream.write(buf)
-      to_read -= len(buf)
+      to_read -= l
 
   stream.flush()
 
@@ -103,6 +105,8 @@ class RangeFile(object):
 
     self._current_rel_offset = 0
 
+    self.seek(0)
+
   def tell(self):
     return self._current_rel_offset
 
@@ -127,22 +131,69 @@ class RangeFile(object):
       raise Exception("Expected to read %d but received %d." % (size, len(buf)))
     return buf
 
+# copies while reading
+class CopyReader(object):
+  def __init__(self, in_stream, out_stream):
+    self._in_stream = in_stream
+    self._out_stream = out_stream
+    self._last_read = 0
+
+  def tell(self):
+    return self._in_stream.tell()
+
+  def seek(self, pos, whence=os.SEEK_SET):
+    self._in_stream.seek(pos, whence)
+
+  def read(self, size):
+    pos = self.tell()
+    if self._last_read < pos:
+      raise Exception("Last read: %d   Current pos: %d" % (self._last_read, pos))
+    buf = self._in_stream.read(size)
+    read_before = self._last_read - pos
+    if read_before == 0:
+      new_read = buf
+    else:
+      new_read = buf[read_before:]
+    l = len(new_read)
+    if l > 0:
+      self._last_read += l
+      self._out_stream.write(new_read)
+    return buf
+
 
 # check for gzip errors
-def test_gz(filename, offset, size, verbose=False):
+def test_gz(filename, offset, size, verbose=False, copy_to_file=None):
   with open(filename, "r") as f_stream:
     f = RangeFile(f_stream, offset, size)
+    if copy_to_file:
+      f = CopyReader(f, copy_to_file)
+      start_pos = copy_to_file.tell()
     try:
-      gz = gzip.GzipFile(fileobj=f, mode="rb")
+      gz = subprocess.Popen(["gunzip", "-t", "-q"],
+                            shell=False,
+                            stdin=subprocess.PIPE,
+                            stdout=subprocess.PIPE,
+                            stderr=subprocess.PIPE)
       while True:
-        buf = gz.read(4096)
-        if len(buf) == 0:
+        buf = f.read(4096)
+        size -= len(buf)
+        if len(buf) > 0:
+          gz.stdin.write(buf)
+        else:
           break
-    except (IOError, ValueError, zlib.error) as e:
+      gz.stdin.close()
+      gz.stdout.close()
+      gz.stderr.close()
+      ret = gz.wait()
+      if ret != 0:
+        raise IOError("Could not decompress warc.gz. gunzip returned %d." % ret)
+    except (IOError, OSError) as e:
       if verbose:
         print >>sys.stderr, e
+      if copy_to_file:
+        copy_to_file.truncate(start_pos)
+        copy_to_file.seek(start_pos)
       return False
-
   return True
 
 
@@ -187,21 +238,20 @@ class MegawarcBuilder(object):
     # decide what to do with this entry
     valid_warc_gz = False
     if entry.isfile() and re.search(r"\.warc\.gz", entry.name):
+      # this is a .warc.gz
       if self.verbose:
         print >>sys.stderr, "Checking %s" % entry.name
-      valid_warc_gz = test_gz(self.input_filename, entry.offset_data, entry.size, self.verbose)
-      if not valid_warc_gz:
-        if self.verbose:
-          print >>sys.stderr, "Invalid gzip %s" % entry.name
+      # add to megawarc while copying to the megawarc.warc.gz
+      warc_offset = warc_out.tell()
+      valid_warc_gz = test_gz(self.input_filename, entry.offset_data, entry.size,
+                              copy_to_file=warc_out, verbose=self.verbose)
 
     # save in megawarc or in tar
     d_target = OrderedDict()
     if valid_warc_gz:
       # a warc file.gz, add to megawarc
-      warc_offset = warc_out.tell()
       if self.verbose:
-        print >>sys.stderr, "Copying %s to warc" % entry.name
-      copy_to_stream(warc_out, self.input_filename, entry.offset_data, entry.size)
+        print >>sys.stderr, "Copied %s to warc" % entry.name
 
       d_target["container"] = "warc"
       d_target["offset"] = warc_offset
@@ -284,10 +334,11 @@ class MegawarcPacker(object):
     entry.offset = self.tar_pos
 
     # calculate position of tar entry
-    block_size = (len(tar_header) +    # header
+    tar_header_l = len(tar_header)
+    block_size = (tar_header_l +    # header
                   entry.size +         # data
                   (tarfile.BLOCKSIZE - entry.size) % tarfile.BLOCKSIZE)
-    data_offset = entry.offset + len(tar_header)
+    data_offset = entry.offset + tar_header_l
     next_offset = entry.offset + block_size
 
     # move to next position in imaginary tar
@@ -303,19 +354,16 @@ class MegawarcPacker(object):
     if re.search(r"\.warc\.gz", filename):
       if self.verbose:
         print >>sys.stderr, "Checking %s" % filename
-      valid_warc_gz = test_gz(filename, 0, entry.size, self.verbose)
-      if not valid_warc_gz:
-        if self.verbose:
-          print >>sys.stderr, "Invalid gzip %s" % filename
+      warc_offset = warc_out.tell()
+      valid_warc_gz = test_gz(filename, 0, entry.size,
+                              copy_to_file=warc_out, verbose=self.verbose)
 
     # save in megawarc or in tar
     d_target = OrderedDict()
     if valid_warc_gz:
       # a warc file.gz, add to megawarc
-      warc_offset = warc_out.tell()
       if self.verbose:
-        print >>sys.stderr, "Copying %s to warc" % filename
-      copy_to_stream(warc_out, filename, 0, entry.size)
+        print >>sys.stderr, "Copied %s to warc" % filename
 
       d_target["container"] = "warc"
       d_target["offset"] = warc_offset