Re: c++/7009: please provide command line option to turn off new include warnings

Re: c++/7009: please provide command line option to turn off new include warnings

[Mark Phillips]

The following reply was made to PR c++/7009; it has been noted by GNATS.

From: Mark Phillips <msp@nortelnetworks.com>
To: gcc-gnats@gcc.gnu.org, gcc-prs@gcc.gnu.org, Hugues.Talbot@cmis.CSIRO.AU,
   gcc-bugs@gcc.gnu.org
Cc:  
Subject: Re: c++/7009: please provide command line option to turn off new
 include warnings
Date: Wed, 10 Jul 2002 17:50:58 +0100

 http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7009
 
 The new warnings also stop gcc 3.1itself from building...
 
 In my case gcc gettext and libiconv are all installed in the same 
 directory, but the existing compiler does not pickup that include/lib dir.
 
 If I use --with-libiconv-prefix=xxx it almost works except, but libjava 
 fails because its configure script thinks loads of things are missing 
 because the tests fail when they see the unexpected warning:-
 
 cc1plus: warning: changing search order for system directory "/xxx"
 cc1plus: warning:   as it has already been specified as a non-system 
 directory
 
 A similar problem also effects several other packages.
 
 The new deprecated C++ library warnings also break the configuration of 
 some packages (though that can be fixed with a command line flag).
 
 Cheers
 Mark
 

Re: c++/7009: please provide command line option to turn off new include warnings

[nathan]

Synopsis: please provide command line option to turn off new include warnings

State-Changed-From-To: open->closed
State-Changed-By: nathan
State-Changed-When: Sat Sep 14 14:44:35 2002
State-Changed-Why:
    fixed in cvs.btw saying -isystem foo -I foo is contradictory

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=7009

c++/7009: please provide command line option to turn off new include warnings

[Hugues Talbot]

>Number:         7009
>Category:       c++
>Synopsis:       please provide command line option to turn off new include warnings
>Confidential:   no
>Severity:       non-critical
>Priority:       low
>Responsible:    unassigned
>State:          open
>Class:          change-request
>Submitter-Id:   net
>Arrival-Date:   Wed Jun 12 23:26:00 PDT 2002
>Closed-Date:
>Last-Modified:
>Originator:     Hugues Talbot
>Release:        3.1
>Organization:
CSIRO DMS
>Environment:
System: OSF1 fire.nsw.cmis.CSIRO.AU V5.1 732 alpha
Machine: alpha
	
host: alphaev67-dec-osf5.1
build: alphaev67-dec-osf5.1
target: alphaev67-dec-osf5.1
configured with: ../gcc-3.1/configure --enable-languages=c,c++,f77 --disable-nls
>Description:
	I find this warning unbearable:

c++  -D_REENTRANT -DOSF1  -I.   -isystem /usr/local/include -DHAVE_CONFIG_H -I/usr/local/include/freetype2 -I/usr/local/include/freetype2 -I/usr/local/include -O2 -Wall -Wno-return-type  -I../../.. -I.. -c arbitrary.cxx
cc1plus: warning: changing search order for system directory "/usr/local/include"
cc1plus: warning:   as it has already been specified as a non-system directory
cc1plus: warning: changing search order for system directory "/usr/local/include"
cc1plus: warning:   as it has already been specified as a non-system directory

0- It is not generally useful. 
1- It drowns the real useful warning information into a sea of useless messages 
2- It is needlessly repeated multiple times, not just across whole projects but as the result of
a single call to the compiler. This is way overkill.
3- I've already indicated -isystem /usr/local/include on the command line, why does gcc continue
to bitch?
4- My compilation stanza are derived from a complex autoconf mechanism, I have no wish to make
it even more complex to fix this non-existent problem. Think of how this must be done: autoconf
must find out where gcc has been installed, and then if it has been installed in the same directory
as some other library I also need, then suppress the excess -I that I would otherwise need. The gcc
maintainers are welcome to propose a solution to this thorny problem.

All I want and I'm sure many others is a way of turning this message off. I couldn't
find any short of hacking the gcc source. It's OK to leave it on as default if you think
it's important (obviously I don't).


AFAIK the gcc user community has done very well without this warning for years, I don't know
what prompted the gcc developers to suddently lecture everybody.


>How-To-Repeat:
Trivial to repeat.
>Fix:
-Wno-include-order-warning or something like that, Pleeeeze?


>Release-Note:
>Audit-Trail:
>Unformatted: