Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!tut.cis.ohio-state.edu!JUNE.CS.WASHINGTON.EDU!pardo
From: pardo@JUNE.CS.WASHINGTON.EDU
Newsgroups: gnu.g++.bug
Subject: Bug in installation instructions.
Message-ID: <8905110006.AA18733@june.cs.washington.edu>
Date: 11 May 89 00:06:13 GMT
Sender: daemon@tut.cis.ohio-state.edu
Distribution: gnu
Organization: GNUs Not Usenet
Lines: 31

The file .../g++-include/stdio.h must be updated on a
machine-by-machine basis.  On *some* machines, the directive at line
55 of 1.35.2's stdio.h is simply wrong.

    #define SPRINTF_RETURNS_INT

It should be commented out.

I did not see particular warnings about this in the installation
guide.  In general, the preferred practice is that when no machine is
provided for machine-dependent code, the code will fail to compile.
IMHO, stdio's willingness to compile incorrectly is a bug.

The installation scripts could take care of patching up the .h files
to match the machine type, or the .h files should barf when they are
used the first time, or there should be a big red WARNING label in the
instructions (there might be one and I just missed it).

I can immagine a machine-dependent `profile':

    stdio.h:#define HAVE_BUFSIZE
    stdio.h:/* #define SPRINTF_RETURNS_INT */

and a magic comment in the .h file

    /*! patch */

and finally a `patchincludes' script that applies the correct
(machine-dependent) profile to each .h file at installation.

	;-D on  ( Soon: instructions for bug installation )  Pardo