Date: Wed, 23 Sep 1998 20:20:23 -0400 (EDT)
Message-Id: <199809240020.UAA24396@indy.delorie.com>
From: DJ Delorie <dj AT delorie DOT com>
To: snowball3 AT usa DOT net
CC: djgpp-workers AT delorie DOT com
In-reply-to: <199809240014.AAA143658@out4.ibm.net> (snowball3@usa.net)
Subject: Re: Has PE-COFF switch been considered?
Precedence: bulk


> 1) After updating the headers to add the additional fields used by 
> PE-COFF, you may only need to tweak the appropriate code and 
> not have to wholesale rewrite it. Of course binutils and everything 
> else would have to reconfigured to generate PE-COFF instead of 
> COFF.

Hmmm, no work for no gain.  No, not a bargain.  The advantages of ELF
are weak symbols and better debugging, neither of which PE-COFF has.

> 2) Would allow for the adding of the occasionally requested .DLL 
> support.

No, it wouldn't.  You'd need to add DLL support in MS-DOS first.

> 3) Would allow for an easier package for writing Win32 programs 
> with DJGPP (another frequent request) since DJGPP and Win32 
> would share the same file format.

No, there's a lot more than that involved.  The main advantage of
DJGPP is the runtime (libc), which you can't use at all when you build
a Windows program.  If this is what you want, use mingw or cygwin.

> 4) From what Mumit Khan told me, PE-COFF is just as good with 
> template instantiation as ELF.

Which means that DJGPP's COFF will do just as well, since PE-COFF is
just regular COFF with well-understood extra sections for DLLs, a DOS
stub, and a few bugs.