On Sunday 26 October 2008 19:12:17 David McCullough wrote:
> Jivin Rob Landley lays it down ...
> > Does anybody see a reason why the build wrapper should _not_ be added
> > back to uClibc, with this change, so people don't have to create a
> > separate toolchain just to try out uClibc?
>
> The uClinux-dist uses a wrapper, ucfront,  based loosely on ccache.

My firmware linux project uses a wrapper drived from the one uClibc had years 
ago.

> We use it to build all combinations of stuff like little endian/big-endian,
> uClibc/uc-libc and glibc.  We use a single multilib'd compiler to build
> all combinations.

I avoid multilib, it's more complication than I generally want to deal with.

> All the toolchains we use in house (which you can download @snapgear.org)
> are built against glibc,  but they can still build fully functional uClibc
> systems without a problem.  I am fairly sure the toolchains we use are
> configured with the static libgcc.a option to make it easier.

That one's not an "easier", that's a "possible" unless you want to build a new 
libgcc_s.so from source.

> As far as I go, the only way you can be sure your dev systems headers and
> libs are not polluting your embedded system build is to use a wrapper.

The gcc path logic is just horrible, you have to --nostdinc and --nostdlib it 
to get sane behavior out of the thing.

> When we introduced it years ago we found numerous dev host header
> dependancies :-(
>
> I don't have a problem with uClibc using a wrapper,  as long as you don't
> stop us from being able to use our wrapper ;-) ;-)

Nah, this one just be supplying an option.  You could still build a native 
wrapper-less compiler ala buildroot, or use a different wrapper...

> Cheers,
> Davidm

Rob
_______________________________________________
uClibc mailing list
[email protected]
http://busybox.net/cgi-bin/mailman/listinfo/uclibc

Reply via email to