is that "install glibc headers" step in crosstool necessary?

[Dan Kegel]

Mike Frysinger wrote:
>>In summary, it *is* possible to build cross-toolchains on some arch's
>>without the Glibc headers step...
> 
> 
> and it's quite trivial to patch all the other arches (except for ia64) to not 
> need libc or kernel headers

Problem is, we'd have to carry those patches in crosstool
for the old versions.  Is it worth the bother?
For new versions, the patches should be submitted upstream
if they're not already in.

What's the deal with ia64?

>>trust me.. I've done the hard
>>yards with Google on this topic and the evidence confirms that building
>>Glibc based cross-toolchains without the Glibc headers step is just plain
>>wrong.
> 
> 
> the stuff ive found have always talked about mismatch exception handling 
> between libgcc_s and the libc ...
> 
> so what if we do:
> - binutils
> - bootstrap C (no libc headers)
> - kernel headers
> - libc
> - 'normal' C (and optional C++/whatever addons)
> 
> will the final toolchain be OK ?

I'm pretty sure that's not enough for NPTL.  It'd be nice
if it were enough for Linuxthreads, but I think the problem
is libgcc_s etc. needs to be built with glibc headers.
The fix for that would be to break out libgcc_s etc. so it can
be built separately from gcc proper.
- Dan

-- 
What does Linux need to succeed on the desktop?  See http://kegel.com/linux/comfort

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