Re: aic7xxx/aicasm manpage

Scott Long wrote:
> Chuck Robey wrote:
> 
>> My kernel build is busted, it can't find:
>>
>> cd ../../../modules; 
>> MAKEOBJDIRPREFIX=/usr/src/sys/amd64/compile/JULY/modules 
>> KMODDIR=/boot/kernel DEBUG_FLAGS="-g" MACHINE=amd64 
>> KERNBUILDDIR="/usr/src/sys/amd64/compile/JULY" make  all
>> ===> aac (all)
>> ===> accf_data (all)
>> ===> accf_http (all)
>> ===> agp (all)
>> ===> aha (all)
>> ===> aic7xxx (all)
>> ===> aic7xxx/aicasm (all)
>> make: don't know how to make aicasm.1. Stop
>> *** Error code 2
>>
>> Stop in /usr/src/sys/modules/aic7xxx.
>> *** Error code 1
>>
>> Stop in /usr/src/sys/modules.
>> *** Error code 1
>>
>> I know that to fix this, the simplest way is to merely stick a
>> NOMAN= <empty>
>>
>> into the Makefile at /sys/modules/aic7xxx/aicasm.  I don't know if 
>> that's the way you'd want to fix it, but it seems to break the build 
>> for me, so a prompt fix is in order, I think.
> 
> 
> It's lingering breakage from the NO_MAN/NOMAN thing.  I thought that 
> some shims were put in to deal with this, but maybe not?  How are you
> compiling the kernel?  Do you have a current world built and installed?

Doing it the hard way, cd to /sys/amd64/conf, do a config witha 
brand-new config (just installed), cd to ../compile/(myname), then do a 
make depend, then do a make (not combine them).

I figured, this was so obvious, there was probably some sort of shim 
that I couldn't spot.

You know what's always bummed me out about our make?  What we oughta 
have is a multi-phase include ... so that, particular include statements 
would not activate until a particular target was completed, then they 
would read and modify things.  Having depends be forced to be either 
done wrong (and most folks misunderstand how it works, and you know it) 
or to be done in two steps.  It ought to be done in one step.

I haven't figued out the syntax ... but I might suggest, something like:

include (filename) (target)

where if the target was left out, it would work as it is today, but if a 
second parameter was supplied, then the include statement would not work 
until/unless that target was satisfied.  Then, additionally, whenever 
such a target was completed, all such includes would be automatically 
make current.

Does that sound good?  I'm a makefile crazy, so if it does sound good, 
leave me with it.  I know how our software dev works, and I want neither 
a guarantee nor anything else like gets often asked for, but I would 
appreciate a word if you have a better notion.
> 
> Scott