On Thu, Feb 08, 2001 at 04:28:23PM +0700, Robert Elz wrote:
> I doubt I need to do that anyway - old userlands generally work
> fine on new kernels (which is certainly what I would try here,
> just a new install kernel, new kern.tgz, and the rest of the
> system the standard 1.5 tarballs).
> 
> (Even if I did, compile time isn't an issue)
> 
> The question is more whether there are likely to be more fixes in
> current that would be useful to have, or whether there my have been
> some instability introduced which would not be nice to keep.

Well, if you're looking for stability I would recommend the 1-5
branch instead :)

> 
> Also, since actually doing a test is going to involve a very long
> turnaround, are there any kernel options (or anything else I perhaps
> should be doing to assist debugging) when I build the install
> kernel.   I will most likely not strip it, and perhaps even make
> a -g kernel (since it will have to boot off CD - there won't be
> anyone available to change floppies when requested... it doesn't
> matter how big it gets for an alpha install).
> 
> I guess SCSIVERBOSE and PCIVERBOSE would be useful - is there
> anything else that might help?

Well, a kernel compiled with -g, and stack trace when it panics would
be a good starting point. Other debug options may be way too verbose if the
kernel works fine.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--