Message-ID: <B0000087307@stargate.astr.lu.lv>
From: "Andris Pavenis" <pavenis AT lanet DOT lv>
To: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
Date: Tue, 18 May 1999 14:15:14 +0300
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Re: gcc 3.0
CC: djgpp-workers AT delorie DOT com
References: <Pine DOT A41 DOT 4 DOT 05 DOT 9905170950280 DOT 58860-100000 AT ieva01 DOT lanet DOT lv>
In-reply-to: <Pine.SUN.3.91.990518135600.24330b-100000@is>
X-mailer: Pegasus Mail for Win32 (v3.11)
Reply-To: djgpp-workers AT delorie DOT com

On 18 May 99, at 13:56, Eli Zaretskii wrote:

> 
> On Mon, 17 May 1999, Andris Pavenis wrote:
> 
> > The problem that must be resolved is building upcomming version for DJGPP
> > and I expect problems there:
> > 	forcing --prefix=$DJDIR through configure
> > 	scripts and makefiles, it is needed there and we cannot avoid it. 
> 
> What about building on non-LFN platforms?  Does it work?

Building egcs-1.1.2 for DJGPP requires LFN support. Removing this limitation
may require large amount of work. Even without that we'll have enough 
problems with upcoming version.

> 
> > 	Of course we can find some hack to get environment variable
> > 	resolved in other way but I don't think gcc is appropriate place
> > 	where to use such hacks.
> 
> Is this something that /dev/env cannot solve?

Please note that then we'll get /dev/env also in dependencies generated
by "gcc -M". As result also make and maybe something else must understand
/dev/env/... 

> What about -imacros in specs--does this have any relevance to the
> official distribution, or is this something that is added manually to
> the DJGPP binary zip?

There are some things that could be done with egcs-1.1.2 distribution for DJGPP:
	- adding -imacros=... to specs
	- moving  lang/cxx/headers.gcc from gppe112b.zip to gcce112b.zip and
          modifying specs to use it as told in readme.djgpp
       - cross-compiling stuff in egcs112s.zip. Currently modified collect2.c does not
          compile under Linux when building cross-compiler. Also perhaps some
          scripts for building cross-compiler would be nice
       -  cleaning source archives from names invalid without LFN support (I did it 
          partially with libraries, but not gcc. Currently I haven't touched that for
          about a month). Support of building without LFN support is next and more
           difficult step. 

There is no problems with first 2 things. 

Andris