Subject: Re: Source changes on the horizon
To: Neil A. Carson <neil@causality.com>
From: Todd Vierling <tv@pobox.com>
List: port-arm32
Date: 09/04/1998 11:29:14
On Fri, 4 Sep 1998, Neil A. Carson wrote:

: > Actually, I have some stuff to add to the arm support for egcs (the stuff
: > for softfloat.c), and at that point arm32 should be switched to egcs and a
: > snapshot drawn up.  I could probably do the dirty work if necessary; my home
: > Shark ain't doing much right now....
: 
: You don't mean switching permanently to egcs, do you? For all stuff?
: Although it works on kernels, there are apparently some very subtle user
: bugs, relating to PIC etc. OR there certainly were recently, Phil and
: Richard have been fixing them all at a rate of knots recently I believe.

Actually, I did.  The codegen for egcs, except for softfloat.c (returning a
64-bit struct in a register pair)--for which there is a patch I have from
Richard--is, in my own opinion, much more solid and stable than gcc 2.7's.  

I don't have final say as to when arm32 switches to egcs--just psychic
suggestive power ;)--but I have been running a system built with egcs over
three iterations for a while with no problems.  I am also using some rather
complex C++ applications compiled by it.  No optimizer bombs about
"unrecognizable insn"s, no spilled register errors....

Whether or not to finally switch will be left up to you and Mark, of course,
with proper input from Richard and Phil.

-- 
-- Todd Vierling (Personal tv@pobox.com; Bus. todd_vierling@xn.xerox.com)