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

["Dave Korn" <dave dot korn at artimi dot com>]

Robert P. J. Day wrote:
> On Wed, 2 Nov 2005, Dave Korn wrote:
> 
>> Robert P. J. Day wrote:
>>> On Wed, 2 Nov 2005, Bill Gatliff wrote:
>>> 
>>>> The fact that it doesn't seem necessary anymore suggests that we
>>>> should at least provide a developer-level option to skip it.  The
>>>> easier we can provide evidence to the gcc guys that their release
>>>> are/aren't bootstrap-able, the more likely they'll stay
>>>> bootstrap-able.
>>> 
>>> this is almost certainly a dumb question but, to be able to skip the
>>> "install glibc headers" step, this suggests that the bootstrap gcc
>>> should be able to be created using solely the installed kernel
>>> headers, is that correct?
>> 
>> Yes, building the bootstrap compiler needs nothing but the headers
>> (and of course libraries too) for the *host* system that it will run on.
>> 
>> Whether those are "installed kernel" headers, because you're
>> cross-compiling from a linux box, or win32 headers when you're
>> building your cross-compiler on a win32 box, makes no odds; the
>> point is, it's the local system's headers that are needed, and
>> nothing for/from the *target* is required at all.
> 
> that's not quite the question i asked.  i realize that all of the
> necessary headers for building the bootstrap gcc will reside on the
> host.  the question was, how many different locations should be
> checked for those headers?

  Umm, then I'm still not quite clear - when you talk about the "install glibc
headers" step, you _are_ talking about installing headers for the target on
your host machine....

> someone from the cross-LFS list just explained to me that the
> target architectures can be grouped into two categories:
> 
> need glibc headers installed:	x86, x86_64, alpha (and apparently SH)
> don't need headers installed:	MIPS, sparc, PPC
> 
> so it's very much an arch-dependent thing, no hard and fast rule here.

  Mmmm, well, it may be that those are the archs that try to pull in the dwarf
unwinder into libgcc with default configure settings - let's see if that
disable-__cxa_atexit thing helps any.


    cheers,
      DaveK
-- 
Can't think of a witty .sigline today....


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