Subject: MI changes, was Re: ktrace records now get lwp id
To: Jaromir Dolecek <>
From: Bill Studenmund <>
List: tech-kern
Date: 07/02/2003 16:48:29
On Sat, 28 Jun 2003, Jaromir Dolecek wrote:

> Indeed. Making a cross-architecture change without bothering to even
> compile the kernel is not acceptable.

Ok, so I've been looking at this. I agree Darren should have tested more
than he did. But where do we draw the line? I've started cross-compiling
kernels to see what else needed work, and checked in a few changes. I'm
building up more.

The thing though is that we define around 265 kernel configurations as
best I can tell ("ls -d */conf/[A-Z]* | grep -v CVS | grep -v Make | wc")
and compiling all of them would be silly. Many of these files are
duplicates/include parts. But we still have a mess of kernel configs. How
many is enough?

One per machine would be good, and "ALL" (if present) or "GENERIC" would
be a good candidate. But what about the ev's? Do you have to do each one?

Also, since we're getting into super-scripting, once we get a list, I'd
like to suggest someone write a script that will build all the tools, then
build each of these kernels. I figure if we're going to make this a
requirement, we should instrument it as much as possible.


Take care,