From: pjfarley AT dorsai DOT org (Peter J. Farley III)
Newsgroups: comp.os.msdos.djgpp
Subject: Rebuilding gcc: Why isn't perform.h used for go32 compile of libgcc1.c?
Date: Wed, 10 Sep 1997 02:09:48 GMT
Organization: None
Lines: 31
Message-ID: <34160074.19893136@snews.zippo.com>
NNTP-Posting-Host: news.newsdawg.com
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Precedence: bulk

[Sorry if this post is a duplicate.  I originally crossposted from
gnu.gcc.help, but since it hasn't showed up on either my djgpp
newsfeed or on DJ's archive, I felt I needed to post it separately
here.]

I have been reviewing the linux configure specifications in the
gcc-2.7.2.3 configure file, and though I don't completely understand
the dependancies and thinking behind those setups, I'm working on
doing just that.  In the meantime, and as a part of my learning
process, I have a question: Why isn't the perform.h "fix" used for
go32 compiles?  Or is this something that could not be used until a
go32 compiler existed?  Or is the perform.h definition incompatible in
some way with the go32 implementation?

In DJGPP terms, why does there need to be a "mklibgcc" executable and
"mklibnow.bat" generated by it to compile libgcc1.a?  Why not use an
include of the "perform.h" file in the config/i386 directory to get
around the problem, using the DJGPP gcc compiler as the "native" cc?

It seems to me that a full DJGPP/go32 configuration is a reasonable
imitation of a linux environment (up to a point, of course).
Therefore it occured to me that an i386-go32-msdos configuration could
conceivably be a (partial) imitation of one of the vanilla linux
implementations, at least as far as the configure script is concerned,
and could use some of the same tricks that a linux configuration uses
to compile libgcc1 using gcc as the "native" compiler.

IOW, what am I missing here?  Or was this path just not yet explored?

----------------------------------------------------
Peter J. Farley III (pjfarley AT dorsai DOT org)