[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: reduced d-i demo3



Erik Andersen wrote:
> BTW, on the subject of library reduction...  I was wondering if we might want
> to use some of H.J Lu's work on the glibc pic libs to help reduce the morass of
> libc dependancies so that when we get around to doing library reduction it will
> be even more sucessful.  I'm hopefull that this will help.
> 
>     http://sourceforge.net/projects/sglibc/

So the idea (quick scan of the mailing list) is to set compile-time
options to disable stuff and stub out functions? The function stuffing
seems unnecessary since we have library reduction, I guess the rest
could be useful. Library rediction cna indeed only go so far -- this
libc is library reduced to run only "hello, world":

-rw-rw-r--    1 joey     joey       348300 Jan 13 13:20 libc-2.2.1.so

Even though mklibs.sh only pulled in the symbols __libc_start_main
and printf, I'm sure there is still bloat in there. This libc can run
only the null program (main(){}):

-rw-rw-r--    1 joey     joey       348364 Jan 13 13:20 libc-2.2.1.so

Running strings on it shows all kinds of crud is really still in there,
from networking to pthreads to printf to 64 bit file operations.

So library reduction has its limits. Limits we won't really hit so
badly, since we have relatively a much more complete system..

I still haven't really ruled out uclibc BTW, but we seem to be doing fine
on size as-is.

Also BTW, I cannot do a cvs checkout of the thing from sourceforge. Auth
failure.

-- 
see shy jo



Reply to: