Re: mailutils/mips{,el} build failure
On Thu, 2009-07-09 at 07:55 +0200, Luk Claes wrote:
> Adam D. Barratt wrote:
> > [Please Cc me on replies; I'm not subscribed to the list]
> > I'm looking at the failure of the mailutils package to build on mips and
> > mipsel, as it's stalling the transition of krb5 to testing.
[...]
> > /bin/sh ../libtool --tag=CC --mode=link cc -g -O2 -g -Wall -O2 -o muauth muauth.o ../libargp/libmuargp.a ../libcfg/libmucfg.la ../lib/libmuaux.la ../auth/libmuauth.la -lgsasl -lgnutls -lgcrypt -lldap -lpam -ldl ../mailbox/libmailutils.la -lcrypt -lresolv -lpthread -lgdbm
> > libtool: link: cc -g -O2 -g -Wall -O2 -o .libs/muauth muauth.o ../libargp/libmuargp.a ../libcfg/.libs/libmucfg.so ../lib/.libs/libmuaux.a ../auth/.libs/libmuauth.so -L/usr/lib -L/usr/local/lib/mysql /build/buildd/mailutils-2.0+dfsg1/mailbox/.libs/libmailutils.so /usr/lib/libmysqlclient.so -lnsl -lm /usr/lib/libgsasl.so /usr/lib/libidn.so /usr/lib/libntlm.so -lgssapi_krb5 -lkrb5 -lk5crypto -lcom_err /usr/lib/libgnutls.so /usr/lib/libtasn1.so -lz /usr/lib/libgcrypt.so /usr/lib/libgpg-error.so -lldap -lpam -ldl ../mailbox/.libs/libmailutils.so -lcrypt -lresolv -lpthread /usr/lib/libgdbm.so
> > /usr/bin/ld: non-dynamic relocations refer to dynamic symbol mu_gsasl_module_data
> > /usr/bin/ld: failed to set dynamic section sizes: Bad value
> > collect2: ld returned 1 exit status
>
> so this probably means there are 2 different mu_gsasl_module_data
> symbols in the code...
As I mentioned on IRC, I've managed to build the package on mips, using
the following trivial patch:
===============================================================
--- mailutils-2.0+dfsg1.orig/include/mailutils/gsasl.h
+++ mailutils-2.0+dfsg1/include/mailutils/gsasl.h
@@ -30,7 +30,7 @@
int mu_gsasl_module_init (enum mu_gocs_op, void *);
-struct mu_gsasl_module_data mu_gsasl_module_data;
+extern struct mu_gsasl_module_data mu_gsasl_module_data;
#ifdef USE_GSASL
#include <gsasl.h>
===============================================================
This also removes mu_gsasl_module_data from libmucfg's BSS segment, but
given that it was the only symbol contained in the segment anyway, I'm
assuming that's a side-effect of fixing the bug rather than an
indication of a problem.
Regards,
Adam
Reply to: