Re: [Bug-gnubg] Compile latest version under Windows

[Jim Segrave]

On Wed 27 Nov 2002 (16:05 +0100), Holger wrote:
> Hi Jim,
> 
> At 23:31 26.11.02 +0100, Jim Segrave wrote:
> >I'm more than willing to look at the errors - I don't know how much I
> >can help with the Windows environment errors, but if you send the
> >output of make > make.out 2>&1 (or however you get stdout and stderr
> >to a file in that system, I am more than willing to try to identify
> >the problems). I may be able to do more under Solaris, as we do run it
> >on a few machines, but I can't install gtk on production
> >machines. Once again, if there are portability issues, I am happy to
> >try to help resolve them.
> 
> A new day, another try. I got further than yesterday, but still not until
> the end.
> 
> I'll attach both my Makefile and the output of make.
> 
> It gives some circular dependencies and undefined references.
> 
> Hope you can help me.

The circular dependencies aren't really a problem. They are a result
of a vastly different Makefile style. I think it's assuming every .o
file in OBJS is dependant on every C and h file in SOURCE, which will
include br1.c. But br1.c is dependant on makebearoff, which is built
with files like eval.o and osr.o. Since make assumes that eval.o is
dependant on br1.c and br1.c is dependant on a file built from eval.o,
it issues a warning. But it looks like it's doing the right thing.

Other issues - there are some more files to add to SOURCE and OBJS:

address@hidden:/tmp$ diff -u Makefile.orig Makefile
--- Makefile.orig       Wed Nov 27 16:35:25 2002
+++ Makefile    Wed Nov 27 16:35:38 2002
@@ -21,7 +21,9 @@
       gtkchequer.c gtkexport.c gtkgame.c gtkprefs.c gtktheory.c  guile.c \
       html.c i18n.c import.c kleinman.c latex.c matchequity.c  matchid.c \
       play.c positionid.c postscript.c pub_eval.c record.c rollout.c \
-      set.c sgf.c sgfl.c sgfp.c show.c sound.c strdup.c text.c osr.c br1.c
+      set.c sgf.c sgfl.c sgfp.c show.c sound.c strdup.c text.c osr.c br1.c \
+      bearoffgammon.c bearoffgammon.h bearoff.c bearoff.h \
+      onechequer.c onechequer.h
 
 OBJS   = analysis.o bearoffgammon.o copying.o credits.o database.o dice.o \
       drawboard.o erf.o eval.o export.o external.o getopt.o getopt1.o \
@@ -29,7 +31,8 @@
       gtkchequer.o gtkexport.o gtkgame.o gtkprefs.o gtktheory.o  guile.o \
       html.o i18n.o import.o kleinman.o latex.o matchequity.o matchid.o \
       play.o positionid.o postscript.o pub_eval.o record.o rollout.o \
-      set.o sgf.o sgfl.o sgfp.o show.o sound.o strdup.o text.o osr.o br1.o
+      set.o sgf.o sgfl.o sgfp.o show.o sound.o strdup.o text.o osr.o br1.o \
+         bearoffgammon.o bearoff.o onechequer.o

That should get your make to complete.

The various warnings I'm looking at - none are fatal, all could do
with cleaning up. The puzzling one is:

getopt.c:647: warning: implicit declaration of function `strncmp'

  Do you know what .h file you have under Windows which defines strncmp?
  Under unix, it's string.h

I'll check in some updates to clear up most of the other warnings,
although they aren't critical (even the missing prototype for strncmp
won't hurt).

-- 
Jim Segrave           address@hidden