[PATCH 1/2] inflate_codes(): Add fix from upstream gzip to prevent false CRC error

[Rob Landley]

On Wednesday 17 February 2010 18:54:29 Denys Vlasenko wrote:
> > I'd say the kernel's version is probably the best one for us to use.  I
> > can take a stab at porting it over if Denys is interested...
>
> Yes, by all means! :)

Looking at it now.

Resisting urge to write a simpler version...  (Do we really need 
deflate_stored?  I mean honestly...)

Still resisting...  (Large #if 0 blocks?  I know they're keeping it as close 
to upstream as possible but really...)

This cannot end well...  (The flush_pending() infrastructure already exists in 
bunzip, put_byte() is a truly unnecessary macro, it's got asserts in it, 
they're using a 16-bit endianness adjust macro to wrote 0xffff, the code was 
written to support 16 bit platforms and has #ifdefs for it all over the 
place...)

Ahem.  Possibly I can clear this weekend to work on this...

Rob
-- 
Latency is more important than throughput. It's that simple. - Linus Torvalds