[CFT] WITH_META_MODE: Working incremental build [only on i386 and amd64]

[Bryan Drewery]

On 5/29/16 3:53 PM, Mark Millard wrote:
> Quoting the original note about WITH_META_MODE ( https://lists.freebsd.org/pipermail/freebsd-current/2016-May/061481.html ):
> 
>> You will also need to load the filemon(4) module with 'kldload filemon'.
> 
> But head's sys/modules/Makefile says:
> 
>> .if defined(MODULES_OVERRIDE) && !defined(ALL_MODULES)
>> SUBDIR=${MODULES_OVERRIDE}
>> .else
>> SUBDIR= \
> 
> . . .
>>        ${_filemon} \
> 
> . . .
>> .if ${MACHINE_CPUARCH} == "i386" || ${MACHINE_CPUARCH} == "amd64"
> . . .
>> _filemon=       filemon
> . . .
> 
> as the only contexts that provide a filemon.ko to use with kldload.
> 
> Thus, for example, arm variants (32 bit and 64 bit) and powerpc variants (32bit and 64 bit) do not have WITH_META_MODE as an option as things are set up.
> 
> I had been hoping to cut down on the time for clang-related rebuilds during native buildworld runs on my slower buildworld contexts (armv7a/cortex-a7, powerpc, powerpc64). But it was not to be.
> 
> It appears that, once some arm variants are officially tier 1, WITH_META_MODE will not span all tier 1 platforms.
> 
> [Since I tend to use non-tier-1 platforms I tend to notice some of the statements about FreeBSD that are true of only tier 1 without being explicit about it. But initially it takes some research to discover that status for each such point. WITH_META_MODE is an example.]
> 

Ah, I wasn't aware of the restriction. I am testing building it for
other archs now.  Most of the arch-dependent code has been removed since
the restriction was added.


-- 
Regards,
Bryan Drewery