ports/73685 - anyone interested?

[Kevin Oberman]

> From: Joe Kelsey <joe at zircon.seattle.wa.us>
> Date: Tue, 09 Nov 2004 19:36:08 -0800
> Sender: owner-freebsd-gnome at freebsd.org
> 
> On Wed, 2004-11-10 at 11:03 +0900, Alexander Nedotsukov wrote:
> > Joe Kelsey wrote:
> > 
> > >Yesterday, I submitted ports/73685 about problems with all *mm ports
> > >(gtkmm, vfsmm, etc.)  I cannot finish my 2.8 upgrade due to inability to
> > >compile these ports.
> > >
> > >This is a critical and high priority problem.
> > >
> > >The problem appears to relate to attempting to use g++34 to compile the
> > >ports.  Should it really use g++34 on a 4.10 system?
> > >
> > >Anyone who has any ideas, please let me know.  I cannot think of where
> > >to go from here.
> > >  
> > >
> > Forgot to CC here. For anyone interested please see PR folowup.
> 
> The answer posted to the PR did not help in the slightest.
> 
> While waiting in vain for someone to help with this problem which has
> bothered me for at least two weeks, I finally stumbled around with more
> or less "random" attempts to solve the problem by working backward from
> gtkmm until I finally did
> 
> portupgrade -f libsigc++
> 
> which actually ended up replacing three different versions of libsigc++,
> including the miscompiled 2.0 version.
> 
> Of course the snide remark made in the PR about "use portupgrade" really
> makes me mad because it simply indicates that the responder did not even
> attempt to read the actual PR which clearly indicates use of
> portupgrade.
> 
> Again, thanks for nothing.

I'm afraid the "use portupgrade" response is all too commonly given
without adequate analysis because it USUALLY works. This is a weird and
ugly case. It involves the use of ports which build header files based
on the system header files and, when the compiler or the system headers
files change, you need to re-install the same version again to get it to
work again.

I was bitten twice by libsigc++ and once by the gcc3.2 port on V5 and
OpenOffice-1.1.3. These can be REALLY tricky to track down, too, as you
learned. 
-- 
R. Kevin Oberman, Network Engineer
Energy Sciences Network (ESnet)
Ernest O. Lawrence Berkeley National Laboratory (Berkeley Lab)
E-mail: oberman at es.net			Phone: +1 510 486-8634