linux-mips
[Top] [All Lists]

Re: IP32 - issues with last CVS snapshoot

To: "Ilya A. Volynets-Evenbakh" <ilya@total-knowledge.com>
Subject: Re: IP32 - issues with last CVS snapshoot
From: "Maciej W. Rozycki" <macro@linux-mips.org>
Date: Fri, 11 Feb 2005 19:27:13 +0000 (GMT)
Cc: "Stephen P. Becker" <geoman@gentoo.org>, Frederic TEMPORELLI - astek <ftemporelli@astek.fr>, linux-mips@linux-mips.org
In-reply-to: <420D006E.3000107@total-knowledge.com>
Original-recipient: rfc822;linux-mips@linux-mips.org
References: <420CEE7F.3080201@astek.fr> <420CF611.5030705@gentoo.org> <Pine.LNX.4.61L.0502111825300.30117@blysk.ds.pg.gda.pl> <420D006E.3000107@total-knowledge.com>
Sender: linux-mips-bounce@linux-mips.org
On Fri, 11 Feb 2005, Ilya A. Volynets-Evenbakh wrote:

> O64 may not be supported ABI, but it provides us with a feature that is really
> usefull:
> specifically, it generates 32 bit symbol addresses instead of 64 bit ones.
> This cuts
> down on code size considerably. If this feature was implemented in toolchain
> as separate
> switch, O64 hack could go away.

 Well, the topic has been beaten to death here, so you don't really need 
to illuminate me -- it's only due to this popular request I've implemented 
the ability to do 32-bit builds for 64-bit kernel.  I just wonder why 
people insisting on such a setup don't actually contribute some code to do 
that cleanly and keep switching between hacks as they stop working one by 
one...

> With that said, you are of course right - IP32 code and some drivers are
> broken, because
> they do rely on this feature in many places.

 Having a compiled tree in place these bugs are trivial to track down with 
"find", "objdump", "grep" and some usual shell script magic.

  Maciej

<Prev in Thread] Current Thread [Next in Thread>