5.3 STABLE kernel compilation oddity

[Martin Nilsson]

Robert Watson wrote:
> I've recently seen an odd nit involving -j kernel compiles, in which
> occasionally the build breaks because of a parse problem and vnode_if.awk. 
> I've been meaning to sit down and see if I can reproduce it, I suspect a
> dependency problem.  -j should always work for the kernel, but I wonder if
> some of the tool upheaval, especially relating to the vnode include stuff
> in -CURRENT, has broken it.

I've seen something similar to this on 5.3 and 5.3-STABLE buildworld it 
happens on fast machines (both SMP & UP) when running 'make -j n 
buildworld' with n larger than about 8. It does not happen everytime, 
just often enough to make you wonder if the hardware is flaky or not.

I have seen this on P4E (Prescott), Xeon (533FSB & 800FSB) as well as 
SMP Opterons. I can't remember getting it on the Celerons. If anybody 
wants patches that fixes this tested, I'd be happy to help. It is very 
annoying when using buildworld to see if a newly built machine is 
working OK.


-- 
Martin Nilsson, CTO & Founder, Mullet Scandinavia AB, Malmö, SWEDEN
E-mail: martin at mullet.se, Phone: +46-(0)708-606170, Web: www.mullet.se

Our business is well engineered servers optimised for FreeBSD & Linux