[Zlib-devel] pigz 1.6 available for testing

[Greg Roelofs]

Glenn wrote:

> At 11:44 AM 3/30/2007 -0700, Mark Adler wrote:
>>
>>By popular demand, pigz 1.6 is available for testing with zlib format  
>>compression (using the -z or --zlib option) with a .zz suffix:
>>
>>     http://zlib.net/pigz16.c.gz

> Appears to work on FreeBSD.

Seems to work on Linux, too, but I noticed the following oddity when
comparing against one of my own raw-zlib blobs from a while back:

% diff red-green-swap.iCCP.zlib.hex red-green-swap.iCCP.zlib.dezlib.pigz-zlib.hex
1c1
< 00000  78 9c 63 60 60 92 71 74  71 72 65 12 60 60 c8 cd  x.c``.qtqre.``..
---
> 00000  78 5e 62 60 60 92 71 74  71 72 65 12 60 60 c8 cd  x^b``.qtqre.``..
            ^^^^^
17c17,18
< 00100  e0 6f 61 60 d8 76 18 00  b7 8c 4b 64              .oa`.v....Kd
---
> 00100  e0 6f 61 60 d8 76 18 00  00 00 ff ff 03 00 b7 8c  .oa`.v..........
> 00110  4b 64                    ^^^^^^^^^^^^^^^^^        Kd

That is, the initial flag bits changed somewhat (didn't check the spec to
decode, sorry--window size?), and there are six extra bytes just before
the checksum.

The pigz-decoded version of my original zlib-compressed file matches the
original uncompressed file, so I'm mainly curious about what the extra
bytes add (other than bulk ;-) ).  The rest of the compressed bytes are
absolutely identical, so the six bytes seem completely extraneous.

Greg

P.S.  Thanks very much either way--I never really wrote my own decoder, so
      I was always hand-converting my zlib files to gzip files, which was
      nasty and brutish...  This is _much_ nicer.