networking chapter

ashes cendres at videotron.ca
Fri Jan 2 08:57:03 PST 2004


On January 2, 2004 11:29 am, Archaic wrote:
> On Fri, Jan 02, 2004 at 10:07:16AM -0500, ashes wrote:
> > I assume hlfs is importing ssl...
>
> So would I, but  I'm most curious about running auditers and such on the
> toolchain so we can at least have that with which to start building and
> testing stuff. Are you planning on having et_dyn and PaX in the first
> book, or is that still for later? Any word about the testsuite errors?

To start with, the hcc script is very bloated. It does not restore gcc spec 
file to exactly what they were originally, and hcc -l is neither pretty nor 
helpfull. This script is now written for bash, I'm looking at perl tutorials 
to convert hcc to perl, and license it as public domain. Perl has a smaller 
footprint then bash, and I think its more suitable and secure; someone who 
knows more perl then I do should take a look into this. It should also be 
possible for hcc to enable ssp without et_dyn, and vise versa. This can 
directly affect toolchain errors. Gentoo already has et_dyn fairly 
stabilized, we just need to make it more generic and tested on lfs. The perl 
hcc script could also detect if crt1S.o is installed, and it if isn't, don't 
offer any et_dyn options, so that all of this is very optional. It's also 
feasible the perl hcc script could detect if gcc accepts -fbounds-checking, 
or any other added gcc features; hcc isn't used very often so we wouldn't 
need to really worry if it takes an extra half second to run while it checks 
gcc extensions. The propolice maintainer hasn't answered me about the new gcc 
errors, sometimes he takes a week or three to reply.




More information about the hlfs-dev mailing list