RE: Cross compiling for systems with existing GLIBC questions ...

["Jamie Risk" <jamie at panacis dot com>]

I was able to answer some of these with a lot of
experimenting ...

> Thanks very much to this post:
>    http://gcc.gnu.org/ml/gcc-help/2009-02/msg00122.html 
> my thinking that I had to build (cross compile all of)
> GLIBC has (I hope) been corrected. But I still have a
> few questions before *I* can leave the topic.
> 
> 1. After compiling BinUtils for the ARM, do I need to
>    put those programs at the beginning of my path to
>    get gcc to build as a cross compiler? 

I have been compiling _something_ without having to alter
the path.

> 2. ... [Does] the gcc configure
>    script have to be told where those GLIBC
>    headers files are?

When I try to cross compile with my new binaries, the 
compiler can't find any of the standard include files;
which seqways into 3/4 below.

> 3. How can I determine which version of GLIBC is running 
>    on my target so that I can do this?

I looked at the ELF headers for the so file; I will 
use the highest version of glibc that I found in the
headers.

> 4. (Arguably offtopic) Do I extract those headers out
>    of the appropriately versioned GLIBC tarball, or
>    will I need to invoke a makefile target on the
>    glibc source to produce just the header files?
> 5. Finally, is the necessary switch on the gcc
>    configure command that locates those GLIBC headers
>    simply:  
>    --with-headers=...  
>    ?

These are still a challenge.

The best possible answer to the above questions are 
instructions on how to build a 4.2 (or later) gcc 
cross-compiler that uses the targets 2.3 GLIBC.

- Jamie

p.s. Before summarily dismissing this thread as tedious,
please assume that I'm here having spent the past week
consulting hundreds of (outdated and incorrect) web pages
dealing with the very specific errors that I've been
getting, and that alternate tools such as Kegel's
'crosstool' or it's predessor 'ct-ng' don't address the 
specific versions above.