[ On Tuesday, January 15, 2002 at 20:31:42 (+0100), Alistair Crooks wrote: ]
> Subject: Re: Simultaneous builds
>
> Funny, someone thanked me the other day for adding OBJHOSTNAME,
> and said it was on his list of things to do.

I submit that whomever said they wanted such a feature didn't think
through the problem completely.  It is entirely and totally unnecessary,
even without the change I've suggested (the change I suggested is only
necessary to support simultaneous builds on multiple hosts sharing the
same pkgsrc source tree).  OBJHOSTMACHINE simply duplicates the ability
already inherent in WRKOBJDIR (by virtue of the fact WRKOBJDIR can point
to a place that will be unique on every build host, a place that can
even be a symlink itself thus further making it totally adjustable and
customisable).

> It is hardly
> "complication".

It is indeed a complication, by the very definition that you've added
more complexity to the expresisions which implement it, albeit a minor
one in pkgsrc (though not nearly so minor in /usr/src).

On an otherwise virgin machine OBJHOSTNAME necessarily also causes at
least one invocation of 'uname -n' too....  (one for every time make
reads bsd.pkg.mk if I'm not mistaken)

>  There is even a way to do this without the
> OBJHOSTNAME definitions that I added, but I thought it best to
> codify practice.

Codifying bad practice doesn't help.....

> Please remember that we don't all share your ease
> of modification of bsd.pkg.mk, familiarity with make(1), or ability
> to use cvs (I'm not being facetious here - just pointing out that,
> often, the intended audience is people who have been handed a
> machine and told to administer it).

I understand, but that's all the more reason to avoid adding yet another
unnecessary knob (i.e. one clearly duplicating existing capabilities)
that makes understanding what's really necessary in a given scenario
even more difficult.  If inexperienced people are expected to be able to
drive this process then there needs to be one and only one way of doing
any particular thing (and of course that one knob for each feature needs
to be carefully and clearly documented too).

(you've no idea how long it took even me to figure out that OBJMACHINE
was both unnecessary and even evil -- I spent many hours staring at the
mess of intricate and inter-related expansions in bsd.obj.mk, the
confusing and all-too-often incorrect documentation in bsd.README, and
that's after doing a whole bunch of experiments to try to see what
effect carefully calculated changes would have!)

> Moreover, I could have done
> it without changing bsd.pkg.mk at all, by advising everyone to
> set WRKDIR_BASENAME for themselves, but I prefer a handy knob for
> these kind of things.

The only knob that's necessary is the existing WRKOBJDIR setting.  It
can even be "hard-coded" in a shared mk.conf file so long (and MUST be
at least consistently defined on all machines sharing a pkgsrc tree).

> > ***************
> > *** 2506,2514 ****
> >                 fi;                                                     \
> >         fi
> >   .  ifdef WRKOBJDIR
> > !       -${_PKG_SILENT}${_PKG_DEBUG}                                    \
> > !       ${RMDIR} ${BUILD_DIR} 2>/dev/null;                              \
> > !       ${RM} -f ${WRKDIR_BASENAME}
> >   .  endif
> >   .endif
> >   
> > --- 2597,2603 ----
> >                 fi;                                                     \
> >         fi
> >   .  ifdef WRKOBJDIR
> > !       -${_PKG_SILENT}${_PKG_DEBUG}${RMDIR} ${BUILD_DIR} 2>/dev/null;
> >   .  endif
> >   .endif
>
> If you'd like to wrap these changes together under a suitable definition,
> and send-pr them, I'm sure we'd look at them.

I could, but until/unless you "buy" into my assertion that OBJHOSTNAME
is a totally unnecessary knob (which in context of read-only source
partitions is equally as evil as OBJMACHINE) and agree that WRKOBJDIR
should be set to a common target on all systems sharing a pkgsrc tree,
my change above is irrelevant and useless to anyone but me (who will
never in a lifetime ever set OBJHOSTNAME or OBJMACHINE (again) :-).

Meanwhile if you do agree with me on this topic then the need for the
above change is very self obvious and necessary in conjunction with
allowing simultaneous builds on separate hosts.

Indeed my change may even be necessary with OBJHOSTNAME, but disliking
OBJHOSTNAME as much as I do I'll leave the reason unsaid....  :-)
(anyone using OBJHOSTNAME and doing enough simultaneous builds on enough
machines at the same time will inevitably discover it  :-)

> Anyway, your mail seems to object to a small part of the changes
> I'm proposing, which went in two days ago, which other people want,
> and which is a convenient knob.

Indeed that's all I'm complaining about.  I ask anyone who disagrees
with me to do so by showing a concrete example where a common shared
setting of WRKOBJDIR is not more than sufficient to allow for shared use
of pkgsrc (either with my change above, or without also allowing for any
simultaneous builds in any way).

> You have not commented on the main
> part of the changes at all - the shlock ones, and the splitting of
> the do-extract target.

I'm personally not really concerned about allowing simultaneous builds,
just so long as the attempt to do so doesn't adversely affect things
(especially the overall performance of "make" in pkgsrc).  While I agree
with David that there are many times when better use of resources can be
attained by doing simultaneous builds in pkgsrc, I'm not likely to try
doing so myself any time soon.  I don't even use 'make -j' -- 'cc -pipe'
is as far as I go to building things "simultaneously".  :-)

-- 
								Greg A. Woods

+1 416 218-0098;  <gwoods@acm.org>;  <g.a.woods@ieee.org>;  <woods@robohack.ca>
Planix, Inc. <woods@planix.com>; VE3TCP; Secrets of the Weird <woods@weird.com>