This is the mail archive of the binutils@sourceware.cygnus.com mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: A patch for configure


   Date: Thu, 18 May 2000 19:31:54 -0700
   From: "H . J . Lu" <hjl@lucon.org>

   On Thu, May 18, 2000 at 06:57:06PM -0700, Ian Lance Taylor wrote:
   >    It won't work if you don't use newlib and have those .o, so, .a and .h
   >    files installed under ${prefix}/${target_alias}. How about I also
   >    check if newlib doesn't exist?
   > 
   > Whether you use newlib or not just means additional -B options.  This
   > is a red herring.

   That is

   # ia64-cygnus-linux-gcc -v -S x.c
   #include "..." search starts here:
   #include <...> search starts here:
    /usr/ia64-cygnus-linux/lib/gcc-lib/ia64-cygnus-linux/2.9-ia64-000216-final/include
     /usr/ia64-cygnus-linux/lib/gcc-lib/ia64-cygnus-linux/2.9-ia64-000216-final/../../../../ia64-cygnus-linux/include

   It is not a problem if you use newlibc. All the C library files are
   within your source tree. But I don't use newlib. My C library stuff
   is under /usr/ia64-cygnus-linux/lib/gcc-lib/ia64-cygnus-linux. If
   I don't create

   /usr/ia64-cygnus-linux/lib/gcc-lib/ia64-cygnus-linux/2.9-ia64-000216-final

   gcc won't find my C library header files via

   /usr/ia64-cygnus-linux/lib/gcc-lib/ia64-cygnus-linux/2.9-ia64-000216-final/../../../../ia64-cygnus-linux/include


   Yes, I have another patch to deal with .o, .a and .so files.

   BTW, it was not a problem until gcc switched from

   /usr/ia64-cygnus-linux/ia64-cygnus-linux/include

   to

   /usr/ia64-cygnus-linux/lib/gcc-lib/ia64-cygnus-linux/2.9-ia64-000216-final/../../../../ia64-cygnus-linux/include


I understand the problem.  Heck, I put the ../.. stuff into gcc in the
first place, in order to make it easier to move an installed tree.

As I said before: you just need some extra -B options.  You have so
far failed to explain why that is not the case.

I want you to explain why this problem justifies your putting a hard
error into the binutils configury that prevents me from building a
cross-toolchain without installing it.  Putting in a difficult to
remember configure option makes it a tiny bit better, but only a very
tiny bit.

This gcc issue has nothing whatsoever to do with the binutils.  This
patch should not have been checked in.

Ian

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]