Re: Support installing headers for bootstrapping libgcc

[Roland McGrath <roland at hack dot frob dot com>]

> Yes, installing them from install-headers does seem better if there's no 
> definite reason why install-headers should not install them 
> unconditionally.  I've tested (on x86_64) the following patch to do this.

Please make it two separate changes.  bits/stdio_lim.h can go in now.
But we still need to let air the question of why we ever used the
move-if-change logic for bits/syscall.h, IMHO.

> As noted I think glibc is the right place to fix this - the point is to 
> provide a well-defined interface to install a working set of headers, 
> sufficient for bootstrapping (but hopefully as complete as possible), 
> without building the library itself.

I'll reply to your other message about this, though I might not get
to it before next week.

> I've confirmed that it is possible to build these headers, without 
> building the whole library, using fairly small changes to build rpcgen for 
> the build system separately from the rpcgen built with the newly build 
> library - so they don't need disabling after all.  I'll send cross-rpcgen 
> patches separately.

That's cool if the changes are not large.  But I'm also fine with letting
that question lie until we have more of a thorough plan for the whole
troublesome sunrpc situation.


Thanks,
Roland