I've fixed the problem with etc/io.d and etc/sched.d causing runtime
errors on those platforms where we cannot compile ctfconvert. This
is only a partial fix, as its not acceptable to preclude ctfconvert
from being available. ctfconvert is important so we can access Linux
kernel structures from D scripts, and this impairs the ability of
dtrace to do interesting things. (On these platforms, fbt + syscall
are still available).
I've rewritten the interrupt routines for i386 and amd64 - the interrupts
are now heavily macro-ised, which will help with future support and should
fix some issues.
I note on one of my kernels, that the io::: provider will hang/crash
the kernel, so I need to validate what is causing this breakage.
I have added a better 'make test' which can be used to to some minor
stress testing, and will add more functional tests as I hit bugs.