Path: utzoo!attcan!uunet!mcvax!hp4nl!philmds!nlgvax!johan
From: johan@nlgvax.UUCP (Johan Stevenson)
Newsgroups: comp.os.minix
Subject: Re: MINIX-ST 1.1 patch #2: fixed binary /usr/lib/cv
Message-ID: <162@nlgvax.UUCP>
Date: 24 Nov 88 12:10:43 GMT
References: <159@nlgvax.UUCP> <277@lzaz.ATT.COM>
Reply-To: johan@nlgvax.UUCP (Johan Stevenson)
Organization: Philips Research Geldrop
Lines: 29

In article <277@lzaz.ATT.COM> hcj@lzaz.ATT.COM (HC Johnson) writes:
> A word of caution. I undecoded cv as instructed, and the checksum
>is ok. BUT, Its broken.  When the minix sources .../kernel/... are
>made, cv fails, complaining about 'only longs can be relocated'.
>The resulting kernel, minix.img is then brain damaged.
>
>Note, this even fails on the standard distribution copy of minix/kernel.

Are you sure?

Indeed, the new cv complains about 'only longs can be relocated'.
But this is a warning. The same limitations was there in the old cv,
but no message was generated.
The trick is that kernel.mix does not need the relocation information
at all.
I just compared the output of the old and new cv on the kernel.out
as present in MINIX-ST 1.1, and they result in identical kernel.mix.
To avoid the warning, you can add the -R option to the new cv,
so you can replace in kernel/Makefile the line
	$(CV) kernel.out $@
by
	$(CV) -R kernel.out $@
This suppresses generation of relocation info altogether, resulting
in a slightly smaller kernel.mix.

I am willing to investigate further, but please give me better
evidence that something is wrong at all.
--
Johan W. Stevenson	johan@pcg.philips.nl		Philips Research