Re: build trouble on MacOS 10.5 [patch]

[John W. Eaton]

On 22-Dec-2009, Rob Mahurin wrote:

| On Tue, Dec 22, 2009 at 09:16:30AM -0500, Ben Abbott wrote:
| > Good news, I'm able to build with all libs except for Apple's Carbon and 
OpenGL frameworks.
| > 
| >   $ ./configure --prefix=/sw --infodir=${prefix}/share/info 
--mandir=${prefix}/share/man --libexecdir=${prefix}/lib 
--with-blas=-Wl,-framework -Wl,vecLib --without-framework-carbon 
--without-framework-opengl
| > 
| > Make check only has two failures.
| > 
| > Summary:
| > 
| >   PASS   5659
| >   FAIL      2
| > 
| > Once, I can build everything I need to be fully functional, I'll
| > look into the test failures.
| 
| I see the same, with a fresh checkout of revision 10014:cb98128649e9.
| 
|       $ ./configure 
|       CC=gcc-mp-4.3 CXX=g++-mp-4.3 F77=gfortran-mp-4.3 
LDFLAGS=-L/opt/local/lib
|       --without-framework-opengl --without-framework-carbon
| 
|         PASS   5507
|         FAIL      2
| 
| I needed the patch attached to include <config.h> before any
| <string.h>.  Is this harmless on other systems?  Is it safe to assume
| that anyone doing "make -C src lex.cc" will have a running perl, or is
| some other magic necessary?  jwe had a less ugly suggestion.

| diff -r 6918d86add73 liboctave/md5.c
| --- a/liboctave/md5.c Fri Dec 18 13:26:05 2009 -0500
| +++ b/liboctave/md5.c Sat Dec 19 12:56:16 2009 -0500
| @@ -51,6 +51,9 @@
|    1999-05-03 lpd Original version.
|   */
|  
| +#ifdef HAVE_CONFIG_H                                                         
 
| +#include <config.h>                                                          
 
| +#endif                                                                       
 
|  #include "md5.h"
|  #include <string.h>

This change is OK.

| diff -r 6918d86add73 src/Makefile.am
| --- a/src/Makefile.am Fri Dec 18 13:26:05 2009 -0500
| +++ b/src/Makefile.am Sat Dec 19 12:56:16 2009 -0500
| @@ -707,10 +707,9 @@
|  #    @echo "expect 14 shift/reduce conflicts"
|  #    $(YACC) $(YFLAGS) --output=$@ --defines=y.tab.h $<
|  
| -#lex.cc : lex.l
| -#    $(LEX) $(LFLAGS) $< > $(@F)-t
| -#    @mv $(@F)-t $@
| -
| +lex.cc : lex.ll
| +     $(am__skiplex) $(SHELL) $(YLWRAP) $< $(LEX_OUTPUT_ROOT).c $@ -- 
$(LEXCOMPILE)
| +     $(PERL) -i -pe 'print qq[#ifdef HAVE_CONFIG_H\n#include 
<config.h>\n#endif\n] if !$$i++' $@
|  
|  $(OPT_INC) : %.h : %.in
|       $(MAKE) -C $(@D) $@

Instead of doing this, I think I would prefer to add a source file
that looks like this:

  #ifdef HAVE_CONFIG_H
  #include <config.h>
  #endif

  #include <lex.cc>

and then compile that instead of compiling lex.cc directly.

| diff -r 6918d86add73 src/gl2ps.c
| --- a/src/gl2ps.c     Fri Dec 18 13:26:05 2009 -0500
| +++ b/src/gl2ps.c     Sat Dec 19 12:56:16 2009 -0500
| @@ -33,6 +33,9 @@
|   * Please report all bugs and problems to <address@hidden>.
|   */
|  
| +#ifdef HAVE_CONFIG_H                                                         
 
| +#include <config.h>                                                          
 
| +#endif                                                                       
 
|  #include "gl2ps.h"
|  
|  #include <math.h>

We will probably periodically update gl2ps.c from the upstream source
from time to time.  I'd rather not have to remember to make this
change each time we update.  I think it would be better to use a
wrapper file similar to what I proposed for lex.cc. 

Would you please make those changes and submit another patch?

Thanks,

jwe