Installworld and /usr/include/*.h modification times

[Mark Andrews]

In message <CA+7WWSewNfRE8XZ3H5HuHWw78yaXv7DKMyaiVzaMoy4kUz1oiQ at mail.gmail.com>
, Kimmo Paasiala writes:
> On Fri, Jun 1, 2012 at 8:45 PM, Lowell Gilbert
> <freebsd-stable-local at be-well.ilk.org> wrote:
> > Kimmo Paasiala <kpaasial at gmail.com> writes:
> >
> >> Why are /usr/include files installed with "install -C" during "make
> >> installworld" =C2=A0when almost everything else is installed without the=
>  -C
> >> flag? This makes it harder to track which files were actually
> >> installed during the last "make installworld". One can easily find
> >> obsolete files =C2=A0(that are not covered with make delete-old(-libs))
> >> with "find -x / -type f -mtime +suitable_time" but this doesn't work
> >> for /usr/include files because the modification times are not bumped
> >> on "make installworld".
> >
> > "make" uses timestamps to determine whether to trigger a rule. Changing
> > timestamps on source files without changing the contents is a bad idea.
> 
> Yes, I'm aware of how make uses timestamps for figuring out out of
> date targets. However I would argue that after updating world with
> "make installworld" (which is done in single user mode there for
> requiring at least one reboot) you should start any compilations from
> scratch. The ports system does this by default and cleans up any
> previous work files before new compilation. I just don't see where
> bumping of mtimes for those files would have that great impact, does
> anyone?

You obviously havn't had to deal with multi-day builds and also having
to repair the OS.  Preserving timestamps preserves re-startability.

Mark
-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: marka at isc.org