[lfs-support] can't compile gcc-4.7.1 pass 1
bruce.dubbs at gmail.com
Tue Jan 15 10:40:21 PST 2013
Dr. George E. Moore wrote:
>> Message: 3
>> Date: Sat, 05 Jan 2013 15:36:34 -0600
>> From: Bruce Dubbs <bruce.dubbs at gmail.com>
>> Subject: Re: [lfs-support] lfs-support Digest, Vol 2764, Issue 1
>> To: LFS Support List <lfs-support at linuxfromscratch.org>
>> Message-ID: <50E89CE2.1060705 at gmail.com>
>> Content-Type: text/plain; charset=ISO-8859-1; format=flowed
>> Dr. George E. Moore wrote:
>> Please don't reply to digest email without trimming.
>>> The output of the "version-check-sh" clearly indicate I have missing and
>>> perhaps in appropriate versions. I read this and still failed to run
>>> this script before now. My apology for not following directions. I'll
>>> make the correction and re-try the GCC-4.7.1 make.
>>> The content of the script follows:
>>> gem at precision690:~$ ./version-check.sh
>>> bash, version 4.1.5(1)-release
>>> /bin/sh -> /bin/dash
>> Needs to be bash
>>> Binutils: (GNU Binutils for Debian) 2.20.1-system.20100303
>>> ./version-check.sh: line 8: bison: command not found
>>> yacc not found
>> Install bison. yacc should then be set up automatically, but check that.
>>> bzip2, Version 1.0.5, 10-Dec-2007.
>>> Coreutils: 8.5
>>> diff (GNU diffutils) 3.0
>>> find (GNU findutils) 4.4.2
>>> ./version-check.sh: line 17: gawk: command not found
>>> /usr/bin/awk -> /usr/bin/mawk
>> Install gawk and check the symlink
>>> gcc (Debian 4.4.5-8) 4.4.5
>>> (Debian EGLIBC 2.11.3-4) 2.11.3
>>> GNU grep 2.6.3
>>> gzip 1.3.12
>>> Linux version 2.6.32-5-686 (Debian 2.6.32-46) (dannf at debian.org) (gcc
>>> version 4.3.5 (Debian 4.3.5-4) ) #1 SMP Sun Sep 23 09:49:36 UTC 2012
>>> m4 (GNU M4) 1.4.14
>>> GNU Make 3.81
>>> patch 2.6
>>> Perl version='5.10.1';
>>> GNU sed version 4.2.1
>>> tar (GNU tar) 1.23
>>> Texinfo: makeinfo (GNU texinfo) 4.13
>>> xz (XZ Utils) 5.0.0
>>> gcc compilation OK
>>> gem at precision690:~$
>> The rest is OK.
>> -- Bruce
> Thanks for getting back with me.
> I have been able to get the gcc-4.7.1 pass 1 to run to completion;
> however, I have some concerns as to its validity, given what I had to
> do. Let me explain:
> After linking /bin/sh -> /bin/bash; installing bison and gawk;
> validating that /usr/bin/yacc -> /usr/bin/bison.yacc, and
> that /usr/bin/awk -> /usr/bin/gawk, I re-configured and ran make again
> in the gcc-build directory. The compilation stilled failed while
> building gcc/c-family/c-cppbuiltin.c; with the same error -
> GNU_USER_TARGET_OS_CPP_BUILTINS() undefined.
We recommend you remove the source directory, and in this case
gcc-build, and start the section fresh. It appears you started from the
failed position and that results in an undetermined state.
> I rechecked the instructions and the host requirements; however, it
> appears that the guidelines are being following.
> I located the macro definition for GNU_USER_TARGET_OS_CPP_BUILTINS() in
> gcc-4.7.1/gcc/config/gnu.h header and added it to c-cppbuiltins.c.
> The compilation then error when compiling gcc/gcc.c. with
> GNU_USER_DYNAMIC_LINKER and GNU_USER_LINKER_EMULATION as undefined.
> I located these #define in gcc-4.7.1/gcc/config/i386/gnu.h and added
> this header to gcc.c.
This should not be necessary. All the instructions needed are in the book.
> My question is whether this pass 1 build is valid? It seems that my
> having to add headers to .c file is some what unorthodoxy, and could
> result in a bad build. I'm a little reluctant to go forward without some
> feedback. I would have assumed that these defines would have been
> included in some "common" directory path search or not needed for
> compiling gcc-4.7.1 pass 1.
More information about the lfs-support