[R-SIG-Mac] Building from source problems.

[Rolf Turner]

On 2/11/2009, at 11:26 AM, Steve Lianoglou wrote:

	<snip>

> In the mean time, do you mind if we take one or two stabs in the dark?

	By all means!  Go for it!

> So, if I'm not mistaken, it seems your only problem now is compiling a
> package from source, and you're getting some "mtune=core2" problem,
> right?

	Ack-chewally I have now installed, from CRAN, the 2.10.0 binary  
designed
	for Tiger, which Simon Urbanek built especially for me, and which
	eschews the "mtune=core2" flag.  With this binary I can now build
	spatstat from source without hiccups.  (Thank you Simon!)

	So I guess I don't really have *any* problems as it stands.

	I'd just kind of like to get *everything* working, seamlessly.
	But I'm getting the impression that this will require upgrading
	my OS from Tiger to the latest.  I have emailed our IT people
	to investigate whether this is a possibility.  Have not yet heard
	back from them ....
>
> THIS IS PROBABLY A BAD IDEA, BUT ...
>
> If we can't get your compiler to support these optimizations (maybe
> because we can't get gcc-4.2 on there), then how about we just don't
> use those optimizations when compiling? I'm not sure if that will
> cause problems, but perhaps we can try?

	This was actually suggested to me by someone else, offlist,
	with less detail.  I didn't have the nerve to try.
>
> Can you find where this is defined in your install? Doing:
>
> $ cd /Library/Frameworks/R.framework/Versions/2.10
> $ grep -r mtune=core2 .
>
> for me brings up one file:
>
> ./Resources/bin/libtool:LTCFLAGS="-mtune=core2 -g -O2"
>
> Does it for you?

	Just grepping in

		/Library/Frameworks/R.framework/Versions/2.10/Resources/bin/libtool

	gives the line that you found.

> I wonder what happens if you just comment that line out and insert  
> below it:
>
> LTCFLAGS="-g -O2"

	I guess that Simon has set things somehow so that this flag doesn't
	get used (???) under the Tiger OS, since, as I said above, the
	build from source now works for me.  So the comment-out and re-define
	trick is not necessary.  I could re-install R using the ``non-special''
	binary (with which I was getting the problem) and see if your trick  
works
	with that, but that would seem to be only of academic interest at this
	stage.  It might be an interesting learning experience for me, but I
	really should be getting on with proper work --- as should the rest of
	you whom I've been pestering about this stuff, I guess! :-)

> Would all of your woes go away?
>
> As I said, it's probably a bad idea, but if you're brave, it might be
> worth a try.
>
> (I've been sitting on this email for a while now, because I don't want
> to make boneheaded suggestions in public, but my xcode d/l is @ 75%
> and I didn't want to leave you hanging for the rest of the day).


	Any suggestion from you is almost certain to be far less
	bone-headed than anything I might do on my own.

	Sorry for muddling about on this.  I have trouble keeping my
	head clear as to what I need to do (as opposed to should or
	could do, or want to do).

	Bottom line:  I seem to be OK now, for the time being at least.
	To keep on being OK, I probably need to upgrade my OS.

	Thanks to everyone for their help and for putting up with my
	often confused and sometimes grumpy questions.

		cheers,

			Rolf

######################################################################
Attention:\ This e-mail message is privileged and confid...{{dropped:9}}