Date: Tue, 27 Jan 1998 14:56:36 +0200 (IST)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: andrewc AT rosemail DOT rose DOT hp DOT com
cc: djgpp-workers AT delorie DOT com
Subject: Re: iostream concern
In-Reply-To: <199801270120.AA052814049@typhoon.rose.hp.com>
Message-ID: <Pine.SUN.3.91.980127145610.843F-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Mon, 26 Jan 1998, Andrew Crabtree wrote:

> Apparantly, like stddef.h, gcc/egcs (whatever), wants to use
> its own generated copy of _G_config.h instead of whatever used 
> to be there.  The gererated _G_config.h define _G_fpos_t as follows. 
> 
> typedef unsigned long _G_fpos_t;
> 
> In DJGPP header files this is defined as a signed long.

By ``DJGPP header files'' you mean the files which are in lang/cxx/,
as distributed with previous versions of libg++, right?  Otherwise, I
don't understand where did you see that definition; it is certainly
not in $DJDIR/include from djdev201.zip.

> This value
> gets defined all the way down to be streampos that iostream.cc uses.
> The latest pg++ snapshot that went out had a libiostream compiled with the 
> generated header file, so people who attempt to use it with 
> the standard DJ header will get undefined references.  

I think that installing a library without the associated headers is an
*extremely* bad idea.  They should install the headers as well.