This discussion needs to be seen by people on [EMAIL PROTECTED], because the problem has to do with a script created by libtool. I've tried to CC the mailing list a couple of times, but it has not been CCed on the whole discussion. The whole thread can be seen from these messages and their followups: http://sourceware.cygnus.com/ml/binutils/2000-03/msg00116.html http://sourceware.cygnus.com/ml/binutils/2000-03/msg00133.html Ian ------- Start of forwarded message ------- Mailing-List: contact [EMAIL PROTECTED]; run by ezmlm Precedence: bulk List-Unsubscribe: <mailto:[EMAIL PROTECTED]> List-Subscribe: <mailto:[EMAIL PROTECTED]> List-Archive: <http://sourceware.cygnus.com/ml/binutils/> List-Post: <mailto:[EMAIL PROTECTED]> List-Help: <mailto:[EMAIL PROTECTED]>, <http://sourceware.cygnus.com/ml/#faqs> Sender: [EMAIL PROTECTED] Mail-Followup-To: [EMAIL PROTECTED], [EMAIL PROTECTED], [EMAIL PROTECTED] Date: 10 Mar 2000 13:22:19 -0800 From: Ian Lance Taylor <[EMAIL PROTECTED]> To: [EMAIL PROTECTED] CC: [EMAIL PROTECTED], [EMAIL PROTECTED] In-reply-to: <[EMAIL PROTECTED]> ([EMAIL PROTECTED]) Subject: Re: [PATCH]: ld/Makefile.am Date: Fri, 10 Mar 2000 13:19:24 -0800 From: "H . J . Lu" <[EMAIL PROTECTED]> Please remember that ld/ld-new is called by collect2 which is called by the new gcc. When the new gcc runs, it sets up GCC_EXEC_PREFIX, INIT_ENVIRONMENT, COMPILER_PATH and LIBRARY_PATH which point to itself which is the new gcc. When collect2 calls ld/ld-new, it passes all those to ld/ld-new. When ld/ld-new calls "gcc", which is the correct compiler to build the whole chain, that gcc driver will see GCC_EXEC_PREFIX, INIT_ENVIRONMENT, COMPILER_PATH and LIBRARY_PATH set up by the new gcc. Of course, it won't work. ld/ld-new tries 4 times to relink the new ld. Every time it will append at least COMPILER_PATH and maybe other environment variables. It won't help anything. When ld-new is a shell script, it is doing a special case of relinking. The shell script could look down COMPILER_PATH and if it is going to reinvoke itself. If so, it could remove the directory from COMPILER_PATH. Perhaps we could make this a special patch in the ld libtool support. Ian ------- End of forwarded message -------
