It looks like putting -L%b in LDFLAGS solves the problem, since that occurs in the linker lines before -lruby and -lruby-static, and so %b gets checked out first.
I've uploaded updated .info and .patch files to my experimental directory: http://www.finkdeveloper.net/svn/users/akh/experimental/ruby-update/ The -10.5 version uses the PostInstallScript that Martin sent out earlier--I decided to leave the 10.4 script as it was. On Apr 17, 2008, at 2:02 PM, Alexander K. Hansen wrote: > My assumption was technically right, but... > > On checking through my build directories, I found that in the > "dirty" case > (i.e. where ruby built) there were a bunch of "-lruby-static", and, of > course, libruby-static.a is generated by ruby itself. The linker, > as is its > wont, browses around the system rather than looking in the package's > build > directory, where there's a perfectly good libruby-static.a sitting > around > atI'm not sure, however, whether we want to change this globally via > a patch > to configure.in the top level. It looks like the LIBRUBYARG_STATIC > variable > carries this information. > > $ grep -r LIBRUBYARG_STATIC * > #Makefile.in#:LIBRUBYARG_STATIC = @LIBRUBYARG_STATIC@ > Makefile.in:LIBRUBYARG_STATIC = @LIBRUBYARG_STATIC@ > bcc32/Makefile.sub:s,@LIBRUBYARG_STATIC@,$$(LIBRUBY_A),;t t > configure:LIBRUBYARG_STATIC > configure:LIBRUBYARG_STATIC='-l$(RUBY_SO_NAME)-static'I guess I'd > say to try > configure:LIBRUBYARG='$(LIBRUBYARG_STATIC)' > configure:LIBRUBYARG_STATIC!$LIBRUBYARG_STATIC$ac_delim > configure.in:LIBRUBYARG_STATIC='-l$(RUBY_SO_NAME)-static' > configure.in:LIBRUBYARG='$(LIBRUBYARG_STATIC)' > configure.in:AC_SUBST(LIBRUBYARG_STATIC) > doc/ChangeLog-1.8.0: config["LIBRUBY_A"]. don't link > $LIBRUBYARG_STATIC > if > doc/ChangeLog-1.8.0: * configure.in (LIBRUBYARG_STATIC, > LIBRUBYARG_SHARED): > linker > doc/ChangeLog-1.8.0: * Makefile.in (LIBRUBYARG_STATIC, > LIBRUBYARG_SHARED): > added. > lib/mkmf.rb: 'LIBS' => "# > $LIBRUBYARG_STATIC > #{opt} #$LIBS") > lib/mkmf.rb:LIBRUBYARG_STATIC = #$LIBRUBYARG_STATIC > lib/mkmf.rb: $LIBRUBYARG_STATIC = config['LIBRUBYARG_STATIC'] > win32/Makefile.sub:s,@LIBRUBYARG_STATIC@,$$(LIBRUBY_A),;t t > wince/Makefile.sub:s,@LIBRUBYARG_STATIC@,$$(LIBRUBY_A),;t t > > > On Tuesday 15 April 2008 10:58:50 pm Alexander Hansen wrote: >> Forwarding on to fink-devel and the maintainer: >> >> In summary: ruby18 doesn't build on clean systems on 10.5.2--we're >> not sure about 10.4 , in the absence of reports from that OS version. >> >> In addition to the build logs, I've posted build directories at that >> location, too, for successful (dirty) and unsuccessful (clean) >> builds. Because it built on a system with a bunch of stuff >> installed, >> I'm assuming that there's a missing BuildDepend. >> >> Begin forwarded message: >>> From: Alexander Hansen <[EMAIL PROTECTED]> >>> Date: April 13, 2008 8:55:34 PM EDT >>> To: Michael Hund <[EMAIL PROTECTED]> >>> Cc: [EMAIL PROTECTED] >>> Subject: Re: [Fink-beginners] Ruby18 build problem Re: Problems >>> installing kexi >>> >>> (ccing John since he had the same problem) >>> >>> I've reproduced the failure by building in a clean environment. It >>> appears that there is a missing build dependency, but I'm not sure >>> what it is. I've posted build logs here: >>> >>> http://akhmac.blogdns.net/~hansen/finklogs/ruby_test/ >>> >>> <snip> ------------------------------------------------------------------------- This SF.net email is sponsored by the 2008 JavaOne(SM) Conference Don't miss this year's exciting event. There's still time to save $100. Use priority code J8TL2D2. http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone _______________________________________________ Fink-devel mailing list Fink-devel@lists.sourceforge.net http://news.gmane.org/gmane.os.apple.fink.devel
