cvs commit: src Makefile

[Robert Watson]

On Sun, 25 May 2008, John Birrell wrote:

> On Sun, May 25, 2008 at 11:00:40PM +0100, Robert Watson wrote:
>> I think you should consider backing out the change -- the point of "make 
>> universe" is to build as much as we possibly can of the tree, including all 
>> variations on architectures, as many obscure kernel configurations as we 
>> can find, etc.  People run make universe to confirm that their changes 
>> haven't broken the build for as many cases as possible, and sun4v remains a 
>> valid case to detect.
>
> I will back out the change, but I think you are making the 'universe' target 
> out to be more than was intended. We used to talk about tiers. We seem to 
> have lost sight of that.

According to our tier documentation, sun4v is a tier 2 platform -- either on 
the way up, or on the way down, depending on how you look at it.  All of the 
other tier 2 platforms, including ARM, PowerPC, ia64 and sparc64, appear in 
make universe, and I would expect objections if any of them were removed from 
make universe also.

>> If you think we should write off sun4v at this point, which may well be a 
>> valid proposal, we should have the larger discussion about that before we 
>> remove it from make universe unless we have a pretty practical reason 
>> otherwise (i.e., dtrace and sun4v are fundamentally incompatible).
>
> This isn't related to DTrace. It's more about what builds developers are 
> expected to do before committing stuff. Much of the buildworld breakage over 
> the last month would have been detected if a universe build had been done. 
> Even the current breakage.

Yes, sounds like people need to start running it before they commit major 
changes.  And it's a cascading problem -- if people don't keep things 
building, it becomes much harder to build test further changes.

Robert N M Watson
Computer Laboratory
University of Cambridge