On Wed, 2008-03-26 at 11:21 +1100, Max Matveev wrote:
> On Wed, 26 Mar 2008 11:06:27 +1100, Nathan Scott wrote:
>
> nscott> On Wed, 2008-03-26 at 10:59 +1100, Max Matveev wrote:
> >>
> nscott> Most users expect to be able to do what James did - type:
> nscott> ./configure --prefix=xxx && make
> >>
> nscott> That basic function does not work in the PCP build atm,
> nscott> and the only sensible way I see to fix it, while keeping
> nscott> the automatic ./configure via make, is the above change.
> >> But it's the wrong change because it creates extra crap which is not
> >> needed except as a crutch to crippled make.
>
> nscott> By "wrong" you mean "suboptimal" or "not ideal"?
> I mean it goes against the original intention of a person who added
> config.done (and it was makc if you're worndering).
Heh, I'd figured that much out.
> Kind of morally wrong, not legally or technically wrong.
OK - so, now you're questioning my morality? :)
> nscott> Why is it dangerous? Is it less dangerous than installing
> nscott> to /usr when user requests /usr/local?
> It's dangerous because it could mean that a stale configure data can
> be picked up in the production build thus making builds less
> reproducible.
> >> Makefile rules to depend on this file instead of config.done.
>
> nscott> Yeah, thats another option. I guess any of the AC_OUTPUT
> nscott> files (or all?) would do.
> Yeap.
Fair enough, I'll update it to do this & nuke config.done.
cheers.
--
Nathan
|