[ale] Moving Right Along..

Michael H. Warfield mhw at alcove.wittsend.com
Mon Jul 20 12:41:24 EDT 1998


Jeremy T. Bouse enscribed thusly:
> On Mon, 20 Jul 1998, Dave Brooks wrote:

> > Date: Mon, 20 Jul 1998 11:18:43 -0400
> > From: Dave Brooks <db at trusted.net>
> > To: "Jeremy T. Bouse" <undrgrid at undergrid.net>
> > Cc: ale at cc.gatech.edu
> > Subject: Re: [ale] Moving Right Along..

> > The egcs people can't seriously blame the libraries, can they?  I mean come
> > on -- those kernels have been compiling fine under gcc with libc5.x.x OR
> > glibc2 for months now...

> 	I don't know... at one point there was a heat'd thread between
> Linus and the developers of egcs on this issue... You could subscribe to
> the egcs mailing list (majordomo at cygnus.com) and check with them... I just
> unsubscribed the other day because of too much traffic and I need'd to get
> on some other lists that I held as higher priority... 
> 	Jeremy

	It has been stated on the linux-kernel list that egcs as well as
gcc 2.8.1 will NOT correctly compile the 2.0.x kernels but WILL correctly
compile the 2.1.x kernels.  That has been stated by several authoritative
developers including, but not limited to, Linus, Alan Cox, and (I think)
Dave Miller.  I would take that as a pretty firm statement.

	It seems really REALLY rediculous to blame the libraries for the
kernel not correctly compiling, considering that the libraries are not
used by the kernel (think about it - the libraries use the kernel - not
the other way around).  Now if it was something in the libraries which
was causing egcs to screw up...  Nah...  Too far fetched...

	Best explaination I heard was a discussion over "grey areas" of
compiler interpretation.  Things like calling functions with other functions
in parameter lists and all of them dependent on global variables.  One
compiler may order things one way, while another may order it differently.
Neither are "wrong" per se, since they are working in an "implimenation
dependent" area.  I think one person even submitted a snippet of code that
compiled with one result under gcc 2.7.x and another under egcs.  The code
"looked" ok, but it innocently hit an area where the compiler behavior is
not defined by any standard and it differed...  It was an artificial
condition, but it did illustrated that compiler dependencies can occur in
legitimate looking code.

	If even one thing in the 2.0.x kernels had something that was
in the "implimentation dependent" code interpretation arena, then
switching to egcs could very easily have broken something.  Under that
circumstance, you can do all the finger pointing you like and both sides
will be fully able to justify and support their positions.  Neither are
"wrong" but the combination is still broken.  So, as the old country doctor
says, "If it hurts when you do that - stop doing that".

>     Jeremy T. Bouse - SouthNet TeleComm Services, Inc - www.STSI.net
>   PGP ID/Fingerprint: 1024/E83D9AE5/4ACC03F098D78198  19D0593E50E597E9
>  Public PGP key available by sending email with 'send pgpkey' in subject
>      undrgrid at UnderGrid.net - NIC Whois: JB5713 - undrgrid at STSI.net
>           /earth is 98% full ... please delete anyone you can.
> 

	Mike
-- 
 Michael H. Warfield    |  (770) 985-6132   |  mhw at WittsEnd.com
  (The Mad Wizard)      |  (770) 925-8248   |  http://www.wittsend.com/mhw/
  NIC whois:  MHW9      |  An optimist believes we live in the best of all
 PGP Key: 0xDF1DD471    |  possible worlds.  A pessimist is sure of it!






More information about the Ale mailing list