From: pavenis AT lanet DOT lv
To: djgpp-workers AT delorie DOT com
Date: Wed, 11 Jul 2001 18:19:37 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: Comments on GCC 3.0 distribution
Message-ID: <3B4C98B9.28903.12EF25@localhost>
In-reply-to: <20010711163616.A291@lauras.lt>
References: <1438-Tue10Jul2001222330+0300-eliz AT is DOT elta DOT co DOT il>
X-mailer: Pegasus Mail for Win32 (v3.12c)
Reply-To: djgpp-workers AT delorie DOT com
Errors-To: nobody AT delorie DOT com
X-Mailing-List: djgpp-workers AT delorie DOT com
X-Unsubscribes-To: listserv AT delorie DOT com
Precedence: bulk

On 11 Jul 2001, at 16:36, Laurynas Biveinis wrote:

> 
> > Yes, we have lots of Posix _POSIX_* constants, and quite a few others.
> > I'm afraid that without out limits.h being included by the one which
> > comes with GCC, some programs which need those constants might not
> > compile.
> 
> Another poor answer, but in this case it would be ``too bad to be true'' -
> too many broken platforms.
> 

Why not to test what happens with DJGPP limits.h? It's not so difficult 
and would have been taken much less time than this discussion: 

C:\>set DJGPP=C:/DJGPP/DJGPP.ENV

C:\>sh
sh-2.04$ echo '#include <limits.h>' | gcc -M -E -
cpp0.exe: warning:  is shorter than expected
-:  c:/djgpp/lib/gcc-lib/djgpp/3.0/djgpp.ver \
  c:/djgpp/include/sys/version.h \
  c:/djgpp/lib/gcc-lib/djgpp/3.0/include/limits.h \
  c:/djgpp/lib/gcc-lib/djgpp/3.0/include/syslimits.h \
  c:/djgpp/include/limits.h
sh-2.04$

(I explicitly set DJGPP before test because of I'm normally using 
modified file with a different name. c:/djgpp/djgpp.env is the original 
one)
> 
> > The problem is, again, with any program which compiled with our
> > stddef.h because it used some data type defined by sys/djtypes.h.  It
> > could fail to compile with GCC's stddef.h.
> 
> You mean that code like
> #include <stddef.h>
> __dj_blah_t
> 
> will stop working?

Perhaps it would be nice to try to build various DJGPP packages and
report results here. Then we'll see what problems we have (maybe not 
all though). 

Something I have tried myself:
	gdb (CVS snapshots about one month ago) - no evident problems
	CVS version of gnuplot - haven't noticed additional problems
	GNU make 3.79.1 - haven't noticed problems
	GRX-2.4.2 - all works, but I had previously submitted related 
					patches 
	Allegro-3.9.36 - had to remove -fomit-frame-pointer from options
					applications which uses allegro crashes

Perhaps also others could do similar tests with different packages and
report about results.

Andris