Solution to MacOSX on Intel?

[Ben Abbott]

On Dec 6, 2007, at 10:58 PM, Ben Abbott wrote:

> *sigh* ... I must admit to an embarrassing error. I was under the  
> impression that config.h was used by configure, rather than being  
> generated by it. Thus I'd made changes to config.h and then run ./ 
> configure ... which reversed the changes :-(
>
> So I'm going through the exercise in *explicit* fashion. Please  
> review my steps will *low* expectations of my knowledge and  
> expertise in such matters ;-)
>
> In any event, I've successfully built octave using g95 on MacOSX  
> 10.5 on Intel. The details are below.
>
> (1) I've applied the patch from John.
>
> Index: configure.in
> ===================================================================
> RCS file: /cvs/octave/configure.in,v
> retrieving revision 1.600
> diff -u -u -r1.600 configure.in
> --- configure.in	4 Dec 2007 20:51:25 -0000	1.600
> +++ configure.in	6 Dec 2007 20:00:42 -0000
> @@ -989,9 +989,13 @@
>      DL_LDFLAGS='-bundle -bundle_loader $(TOPDIR)/src/octave $ 
> (LDFLAGS)'
>      MKOCTFILE_DL_LDFLAGS='-bundle -bundle_loader $$BINDIR/octave-$ 
> $OCTAVE_VERSION$$EXEEXT'
>      SH_LDFLAGS='-dynamiclib -single_module $(LDFLAGS)'
> -    CXXPICFLAG=
> -    CPICFLAG=
> -    FPICFLAG=
> +    case "$canonical_host_type" in
> +      powerpc-*)
> +        CXXPICFLAG=
> +        CPICFLAG=
> +        FPICFLAG=
> +      ;;
> +    esac
>      SHLEXT=dylib
>      SHLLIB='$(SHLEXT)'
>      SHLEXT_VER='$(version).$(SHLEXT)'
>
> (2) I ran autogen
> ./autogen.sh
>
> (3) and then configure
> ./configure '--prefix=/sw'  '--infodir=${prefix}/share/info' '-- 
> mandir=${prefix}/share/man' '--libexecdir=${prefix}/lib' '-- 
> build=i386-apple-darwin' 'build_alias=i386-apple-darwin' 'CFLAGS=-O0  
> -g -ggdb3' 'LDFLAGS=-L/sw/lib' 'CPPFLAGS=-I/sw/include' 'CXXFLAGS=- 
> O0 -g -ggdb3' 'F77=/sw/bin/g95' 'FFLAGS=-O0'
>
> (4) *now* I've edited config.h (thanks again to John's direction).
> edit the generated config.h file and change the line
> - /* #undef OCTAVE_HAVE_BROKEN_STRPTIME */
> + #define OCTAVE_HAVE_BROKEN_STRPTIME 1
> - #define HAVE_STRUCT_TM_TM_ZONE 0
> + #define HAVE_STRUCT_TM_TM_ZONE 1
>
> (5) build
> make
>
> (6) test
> make check
>
> -------------------------
> [...]
>   scripts/time/datevec.m ................................. PASS    
> 11/11
> [...]
> Summary:
>
>   PASS   3770
>   FAIL      0
>
> There were 1 expected failures (see fntests.log for details).
> -------------------------
>
> I'll spend more time determining if both changes to config.h are  
> needed and if the patch to configure.in is needed.
>
> In the meantime, perhaps someone else can duplicate/verify this  
> result with g95, or better yet, gfortran.

I've tried both changes to config.h in isolation on MacOSX-10.5/Intel.  
The TM_TM_ZONE part is not needed. Thus the change to config.h should be

- /* #undef OCTAVE_HAVE_BROKEN_STRPTIME */
+ #define OCTAVE_HAVE_BROKEN_STRPTIME 1

Ben

-------------- next part --------------
An HTML attachment was scrubbed...
URL: https://www.cae.wisc.edu/pipermail/bug-octave/attachments/20071207/b5f33e3b/attachment.html