FYI: Fix PR632: patches/security/20110215/6878713.patch breaks shark zero build

[Mark Wielaard]

On Tue, 2011-02-22 at 12:05 +0000, Dr Andrew John Hughes wrote:
> On 10:39 Tue 22 Feb     , Mark Wielaard wrote:
> > On Mon, 2011-02-21 at 20:52 +0000, Dr Andrew John Hughes wrote:
> > > I can see how it would be advantageous from a speed perspective, but for me,
> > > if I'm going to bother building one of the alternative VMs, I want to know it
> > > actually works.  Now I guess you can run 'check-hotspot' with each of the alternative
> > > VMs but I don't see why that saves anything over just doing the full build,
> > > which tests the VM heavily with lots of javac invocations.
> > 
> > So we need targets for at least check-hotspot and check-langtools to be
> > run with alternative runtimes build in. This might even be a nice idea
> > for pure hotspot client/server/tiered testing. Ideally also
> > make-check-jdk, but that on itself takes ages.
> > 
> 
> Or just use the --enable options which already work with make check.

I think you misunderstand why --enable-additional-vms is very useful.
If you use the --enable options you have to rebuild everything
completely for each alternative runtime. Which by now is cacao,
shark/zero, alternate hotspot (although that doesn't have an
additional-vms variant yet) and jamvm. Having to rebuild everything five
times is just not very efficient and takes way too much time. The idea
is that we extend the tests so they can be run with each alternative
runtime installed/compiled in instead.

Cheers,

Mark