Re: gcc 2.95.2 on rs6000-ibm-aix4.3.3.0

["Richard Shen" <richard dot shen at phc dot aetna dot com>]

I'm afraid I was premature in suggesting the make worked.  True, the make
finished without errors nor complaints.  However, a subsequent compile of lsof
(which worked great with gcc 2.95) failed miserably early, in some cpp step (too
many params or environment too big -- so it says).  I am in the process of
recompiling gcc 2.95.2 without --enable-shared switch.  I will forward on my
results as soon as I can, if you are interested.




David Edelsohn <dje@watson.ibm.com>
Thursday January 27, 2000 11:01 AM
To:   Richard Shen/PHG/Prudential@PHC
cc:   gcc@gcc.gnu.org
      Bruce Korb <autogen@linuxbox.com>
Subject:  Re: gcc 2.95.2 on rs6000-ibm-aix4.3.3.0




>>>>> "Richard Shen" writes:

Richard> Output from config.guess:
Richard> rs6000-ibm-aix4.3.3.0

Richard> However, links:
Richard> /usr/local/lib/gcc-lib/rs6000-ibm-aix4.3.3.0/2.95.2/include/Mrm ->
Richard> Motif2.1/Mrm
Richard> /usr/local/lib/gcc-lib/rs6000-ibm-aix4.3.3.0/2.95.2/include/uil ->
Richard> Motif2.1/uil
Richard> point to non existent directories.

     This looks like a problem with the header fixing functionality of
GCC.  /usr/include/Motif2.1/... does exist.  I am sort of surprised that
directly including through <Mrm/...> or <uil/...> is okay, but I guess it
is there.  The problem is that nothing needs to be fixed in the
Motif2.1/Mrm directory yet the symbolic link remains.  When cpp searches
for a header file, it finds the include path in the gcc header file cache
but does not find the actual directory.  Somehow when the directory is
found unnecessary, the symbolic link needs to be removed.

David
===============================================================================
David Edelsohn                                      T.J. Watson Research Center
dje@watson.ibm.com                                  P.O. Box 218
+1 914 945 4364 (TL 862)                            Yorktown Heights, NY 10598