cvs commit: src/sys/boot/i386/loader Makefile

[John Baldwin]

On Wednesday 01 June 2005 10:48 am, Maxim Sobolev wrote:
> David O'Brien wrote:
> > On Wed, Jun 01, 2005 at 12:28:08AM +0300, Maxim Sobolev wrote:
> >>David E. O'Brien wrote:
> >>>obrien      2005-05-31 21:16:51 UTC
> >>>
> >>> FreeBSD src repository
> >>>
> >>> Modified files:
> >>>   sys/boot/i386/loader Makefile
> >>> Log:
> >>> Put bzip2 support on equal footing with gzip support.
> >>> Enable bzip2 support by default, set LOADER_NO_BZIP2_SUPPORT to disable
> >>> it.
> >>
> >>This only increases the size of loader without any good reason. Please
> >>stop making nonsense commits.
> >
> > And having gzip support on by default only increases the loader without
> > any good reason.  We provide a feature rich environment.  The loader
> > isn't at the brink of its size limit like boot0 is.  I put bzip2 on par
>
> No, it is, since currently it's still limited to 640KB of memory both
> for data and for text on i386 and possibly on amd64. For that reason you
> can't decompress bzip2-compressed filesystems compressed with block size
> higher than 100k, which is the smalles block size possible with bzip2.
> Still it provides noticeably better compression  than libz.
>
> Decompression requires about 100k + ( 2.5 x block size ), that is 350k.
> Given that loader's text is currently 200-250k and the fact that it
> needs some heap for other functionality even 20-30k can really matter.
> Actually I am not sure if you can do bzip2 decompression w/o turning
> forth and gzip support off on i386 now.

Hmm, given this, does anyone actually use bzip2fs in the loader now?  I guess 
it might be ok for non-i386, but perhaps i386 should not even try to use it?

-- 
John Baldwin <jhb at FreeBSD.org>  <><  http://www.FreeBSD.org/~jhb/
"Power Users Use the Power to Serve"  =  http://www.FreeBSD.org