Re: Cross compile error question: i686 to powerpc--linux-gnu

[Joe Sislow <flynn at cosmodog dot com>]

David Korn wrote:

>   The rest of the problem (D'oh!!  I've only just noticed, I was so
thrown
> by you target triple) is probably caused by the fact that newlib
doesn't
> work with linux.  Use glibc instead.
>

Ok, that's helpful, but confusing (as NONE of the CrossGCC faq's cover
using
glibc).  Maybe I can just cut to the chase here and see if you can
help...

We're creating our own hardware with a PowerPC 603e chip.  We're going
to have to
write our own libraries (we understand this).  However, we just want a
barebones
compiler that we can work from without anything weird.

Why does GCC need all of these libraries recompiled to output standard C
code?
I can understand that the linker and assembler might need to know more
about the
hardware, but why is it so hard to get GCC to just spit out the right
code to
binutils?  The binutils has (seemingly) been the most painless up to
this point.

Basically, I know that my compiler is working, as it seems to be
generating
accurate PowerPC assembly code without the last recompile of the
bootstrap
compiler.  Linking doesn't work with GCC, but we know that to link the
assembly
code, we're going to need to specify information about the memory
layout.

Anyway...long story short...if we want to provide our own library
support, do
I NEED to go any further than making the bootstrap compiler?  I'm not
going to
use Linux, I'm not going to use GNU (most likely)...so, am I going too
far in
something I don't need to do?

And on a more philosophic note, shouldn't a compiler just look for
prototype
descriptions and hand the rest off to the linker to resolve?  Why does
GCC need
to know ANYTHING about system-specific libraries?  Is there some
integration with
binutils that's going on here I don't understand?




------
Want more information?  See the CrossGCC FAQ, http://www.objsw.com/CrossGCC/
Want to unsubscribe? Send a note to crossgcc-unsubscribe@sourceware.cygnus.com