modprobe audit/check-list

Rob Landley rob at landley.net
Fri Nov 25 07:29:26 UTC 2005


On Thursday 24 November 2005 14:00, Bernhard Fischer wrote:
> >I was browsing gcc bugs and noticed one where some udel folks were
> >benchmarking gcc 2.95 against current, and current was 10 times worse,
>
> As you are talking on the busybox list, did you -- out of curiosity --
> compile busybox allyesconfig with gcc-2.95.x? If so, i'd be interrested
> in the patches which make the codebase compile with it while not
> punishing "recent" gcc's wrt variable sized arrays and other nifty stuff
> which helps in keeping things smallish :)

Booting a Red Hat 7.2 image with a 2.2 kernel and building busybox under it is 
on my todo list. :)

> >on this particular test.  Every time I browse gcc bugs I see some
> >report like that, where gcc 2.95 is the gold standard of performance,
> >and current does not measure up.
>
> hm, well, gcc-current needs about 380 MiB of RAM if you use that -combine
> patch of mine.

Ouch.  The current high water mark for gcc-4.0.2 compiling itself is somewhere 
south of 256 megs.  (Probably not far south of it, though.)

I think the -combine thing may have to be a config option.

> I only tested it on a machine with 1GiB of RAM and not 
> with 32Mib of RAM and no swap, so i do not know how good that idea was.

User Mode Linux and qemu make this sort of thing so much easier.

There's not a whole lot of development you can do in 32 megs anymore.  
Possibly with tcc...

> >Good thing 2.4 is actively maintained, and relatively stable, in an
> >honest sense of the word.
> >
> >That's something to be thankful for.  :-)
>
> Yes. There are constraints where current 2.6 simply doesn't cut,
> compared to 2.4. Most notably RAM usage int he network-stack addons,
> at least from my pov.

Does matt's -tiny tree address this?  (I haven't tried particularly complex 
iptables stuff, my longest set of iptables rules was about 30 lines.)

Rob
-- 
Steve Ballmer: Innovation!  Inigo Montoya: You keep using that word.
I do not think it means what you think it means.



More information about the busybox mailing list