Ignatios Souvatzis wrote:

> Hi,
> 
> in short: madplay 0.12.5b as built form pkgsrc is dumping core on
> a 1.5 DNARD system.
> 
> I forced it to statically compile in libmad.a instead of using the shared
> library, and it works. Now, whats broken... madplay, the package, the compiler,
> the linker, ld.so_aout, libtool?

madplay's own libtool doesn't know about shared libraries, so
the following at least produces something that works on arm32
(just ignore the warnings about not finding libmad.so.0.0):

Index: Makefile
===================================================================
RCS file: /cvsroot/pkgsrc/audio/madplay/Makefile,v
retrieving revision 1.11
diff -d -p -u -r1.11 Makefile
--- Makefile	2001/02/16 13:30:46	1.11
+++ Makefile	2001/03/13 03:07:17
@@ -15,7 +15,9 @@ BUILD_DEPENDS+= msgfmt:../../devel/gette
 
 GNU_CONFIGURE=	YES
 USE_LIBINTL=	YES
+.if (${MACHINE_ARCH} != "arm32")
 USE_LIBTOOL=	YES
 LTCONFIG_OVERRIDE=	${WRKSRC}/ltconfig
+.endif
 
 .include "../../mk/bsd.pkg.mk"

I did try

	CONFIGURE_ARGS=--disable-shared

but while it did get passed to ./configure it still made shared
libraries.

This obviously falls into the "gross hack" department, but since this
is really the only usable mp3 player on a shark, should I commit it
(with some appropriate XXX comments)?

Simon.
--
Simon Burge                            <simonb@wasabisystems.com>
NetBSD CDs, Support and Service:    http://www.wasabisystems.com/