Re: About testing "self-buildability"

[Andreas Gustafsson <gson%gson.org@localhost>]

Jukka,

Earlier, I wrote:
> Both sysinst and anita recently gained support for installing the
> source sets, so that part is now easy.

Or so I thought.  See PR misc/46182.

> >  - Hardware to run it on; the important parameter is single-threaded CPU
> >    performance (qemu is single threaded, and I doubt the Xen MP support
> >    is mature enough to use for this)
> > 
> >  - A volunteer with a lot of patience to set things up.
> 
> This wouldn't necessarily need to be part of the infrastructure. It could be
> just a flag in Anita. Like
> 
>       anita self-build http://ftp.netbsd.org/pub/NetBSD/NetBSD-5.0.2/i386/

I still think a shell script wrapper around anita should be enough to
get started.  I wonder what the simplest possible build.sh command
line is...  Doing some experiments, I found that the naive "cd
/usr/src && build.sh release" does not work on a freshly installed
system, because /usr/obj does not exist.  The following worked well
enough to expose some problems with anita mistaking parts of the
build.sh output for shell prompts (fixed in anita 1.25, due out soon),
and the abovementioned misc/46182:

   anita boot --run "mkdir -p /usr/obj && cd /usr/src && ./build.sh release" 
--disk-size 8G --memory-size 256M --sets 
kern-GENERIC,modules,base,etc,comp,games,man,misc,tests,text,syssrc,src,sharesrc,gnusrc
 http://nyftp.netbsd.org/pub/NetBSD-daily/HEAD/201203091940Z/i386/

> > Given that we already know this test will fail, could you clarify what
> > the goal of the exercise is?  Is it to detect future regressions after
> > the present one has been fixed, or are you perhaps thinking of doing
> > automated bisection to help find the cause of the present regression?
> 
> Most of all, I think building the release is a very good stress test,
> which is probably also the reason why this particular bug has not
> been catched by any of the other tests. At least in terms of CPU
> time, heavy I/O, and memory pressure, this would be as good as any
> other available stress test.

True.  If it could be run on a multiprocessor (e.g., MP Xen), it would
be even better.
-- 
Andreas Gustafsson, gson%gson.org@localhost