The bzip2/lzma patches

["H. Peter Anvin" <hpa@zytor.com>]

Hi Alain,

I finally got the tree with your latest bzip2/LZMA code unburied.  My
apologies for having sat on it for so long (you know why.)

Anyway, I did run into a few things that I'd really like to get fixed.

First of all, there are more than just the x86 and ARM tree which is
using the old API, I count at least four more architectures.  I can't
apply the ARM tree anyway, and we clearly need a saner migration way.

Second, the ugly if...goto hacks in init/initramfs.c could be much
cleaner handled by iterating down a list of function pointers -- except
that gunzip() seems to have a different return value (in init/than the
others?!

Third, you're using the construct if (!foo() < 0 && message == NULL)

... which is wrong on three(!) accounts, one technical (! has higher
precedence than <) and two stylistic (write >= instead of !(.. < ..) and
write !message instead of message == NULL).

I think the solution to all of this is to introduce a new function
instead of gunzip(), and have the old gunzip() be a wrapper using the
old ABI.  That is much cleaner than relying on #ifdef NEW_CODE.

Finally, I know I suggested it, but I don't think the
<linux/decompress/*.h> header file are justified given that they only
contain a single function header.  Might as well put them all in a
single <linux/decompress.h> header.

	-hpa

-- 
H. Peter Anvin, Intel Open Source Technology Center
I work for Intel.  I don't speak on their behalf.