Got it (Re: KERNLOAD undeclared error during buildkernel)

[Simon 'corecode' Schubert]

Matthew Dillon wrote:
:    How odd.  It's got to be that somehow __i386__ wasn't defined
:    before, and is now (maybe a compiler or preprocessor change),
:    because I don't see a KERNLOAD anywhere in the codebase other
:    then in that assert.
    Figured it out!

    Simon's commit today did it:

    If compiling a module with the kernel, suck in opt_global.h.

    Before this commit, INVARIANTS was *NEVER* defined during the
    module build because, well, they were modules and until this commit
    modules did not inherit kernel options.  Since INVARIANTS was not
    being used, the KASSERT() was #define'd to nothing.
aha! I noticed that build somehow broke, but I really didn't get the 
reason *why* that happened. I knew it had to be something with the 
preprocessor, but in the short time I had at hands wasn't able to track 
it down.

Thanks,
  simon