Re: safer way to use gnulib (for other MacOS X users)

[Thomas Treichl]

Am 17.03.10 22:30, schrieb Ben Abbott:
> On Wednesday, March 17, 2010, at 04:52PM, "Thomas Treichl"<address@hidden>  
> wrote:
> > Am 17.03.10 21:31, schrieb Ben Abbott:
> > > On Wednesday, March 17, 2010, at 03:28PM, "Thomas Treichl"<address@hidden>   
> > > wrote:
> > > > Am 16.03.10 23:38, schrieb Ben Abbott:
> > > > > On Mar 16, 2010, at 6:03 PM, John W. Eaton wrote:
> > > > >
> > > > > > On 16-Mar-2010, Ben Abbott wrote:
> > > > > >
> > > > > > | The dependencies of lex.cc aren't too many, so it finished quickly. Except 
> > > > > > for the one failure I always get with data.cc, all tests pass.
> > > > > > |
> > > > > > | Thanks again!
> > > > > >
> > > > > > OK, so now we finally have Octave+gnulib working on OS X, except that
> > > > > > we need to do something about the possibility of sigemptyset and
> > > > > > sigaddset being defined as preprocessor macros.
> > > > >
> > > > > I've attached a diff for others who may be building on MacOS X.
> > > > >
> > > > > Ben
> > > >
> > > > Hi Ben,
> > > >
> > > > I wasn't able to build a development snapshot for a very long time
> > > > because of the gnulib problems. With your patch and a fresh gnulib
> > > > checkout I get the following result on my 10.6 machine
> > > >
> > > >     Summary:
> > > >       PASS   6374
> > > >       FAIL      0
> > > >
> > > > So wow, thanks to all developers!
> > > >
> > > >     Thomas
> > >
> > > Thomas,
> > >
> > > I continue to encounter a failure in data.cc.
> > >
> > >    121>>>>>   processing 
> > > /Users/bpabbott/Development/mercurial/local_clone/src/data.cc
> > >    122   ***** assert(log2(complex(0,Inf)), Inf + log2(i));
> > >    123 !!!!! test failed
> > >    124 assert (log2 (complex (0, Inf)),Inf + log2 (i)) expected
> > >    125 Inf + 2.266i
> > >    126 but got
> > >    127 Inf - NaNi
> > >    128 NaNs don't match
> > >
> > > Using Apple's gcc-4.2.1, I've not been able to resolve this by changing the 
> > > optimization setting. Which gcc are you using and what level of optimization?
> > >
> > > Ben
> >
> > Oh yes, I remember I had similiar problems. I used these flags for CXX
> > and for my g95 with low optimization -O instead of at least -O2 or even
> > -O3 that I used before the change to gnulib. But I can check once again
> > with higher optimization tomorrow
>
> Are you really using g95 and not gfortran?
>
> If so what version? and did you use it to build all the fortran dependencies?
>
> Ben

Hi Ben,

yes I'm really really using g95 and not gfortran because of two reasons
 a) I never installed gfortran on my 10.6 machine
 b) I'm using g95 that I've packed into Octave.app 3.2.3 which is

  which i386-apple-darwin8.11.1-g95; i386-apple-darwin8.11.1-g95
  --version
  /Applications/Octave.app/Contents/Resources/bin/i386-apple-
  darwin8.11.1-g95
  G95 (GCC 4.0.3 (g95 0.91!) Feb 27 2008)
  Copyright (C) 2002-2005 Free Software Foundation, Inc.

I used the libs that come with Octave.app 3.2.3 that I have built with 
the 10.4 system months ago. I'm compiling Octave against these libs and 
overwrite Octave 3.2.3 with 3.3.50+. This means that the rest of the 
system (you're right) is not really be built for 10.6 but with 10.4 SDK 
for a 10.4 system, only Octave is built against 10.6 SDK.

Today I used -O2 to compile the sources once again. The test script 
doesn't work anymore because of the following error

  octave-3.3.50+:1> fntests
  Integrated test scripts:
  invalid assignment to cs-list outside multiple assignment.
  octave-3.3.50+:2>

Regards

  Thomas