Problem compiling v2.5.2 under Solaris 9 on Netra t105 system

Julien ÉLIE julien at
Fri Aug 26 16:09:43 UTC 2011

Hi Craig,

> Attempting to upgrade from one of the v2.4 releases to v2.5.2.
> I used the exact same configure script options as I did with the 2.4 
> install I'm running currently, namely:
> sh ./configure --prefix=/news --enable-uucp-rnews --enable-libtool \
>          --with-perl --with-python --with-sendmail=/usr/lib/sendmail \
>          --with-log-compress=gzip

I see that you are using libtool.
Thanks for the hint.

> and then attempted to do 'make update' to build the new version with the 
> intention of replacing the current one.

I think "make" and then "make update" is safer.  (It also allows you to run
"make check" before "make update".)  Anyway, it is not related to the issue
you are reporting.

> The build failed when linking libinn with the following errors:
> ld: fatal: symbol `skip_fws' is multiply-defined:
>          (file .libs/date.o type=FUNC; file .libs/headers.o type=FUNC);
> ld: fatal: symbol `IsValidHeaderName' is multiply-defined:
>          (file .libs/date.o type=FUNC; file .libs/headers.o type=FUNC);
> ld: fatal: symbol `skip_cfws' is multiply-defined:
>          (file .libs/date.o type=FUNC; file .libs/headers.o type=FUNC);
> ld: fatal: File processing errors. No output written to 
> .libs/

Yep, I can reproduce it with the INN 2.5.2 release.
I feel terribly sorry not to have tested a libtool build before
its release.
(As a side note, Autobuild will appear to be useful for preventing such
issues from happening!)

The good point is that I cannot reproduce the bug with the latest STABLE
2.5 source code.

Here is the patch you are looking at:

July, 11th 2010, 02:27:43 (14 months ago)
Remove the mention of an unnecessary (and weird) include file.
(See commit 8871.)

Index: /trunk/lib/Makefile
--- /trunk/lib/Makefile	(revision 9081)
+++ /trunk/lib/Makefile	(revision 9082)
@@ -127,6 +127,5 @@
 date.o: date.c ../include/config.h ../include/inn/defines.h \
   ../include/inn/system.h ../include/inn/options.h ../include/clibrary.h \
-  ../include/config.h headers.c ../include/inn/libinn.h \
-  ../include/inn/defines.h
+  ../include/config.h ../include/inn/libinn.h ../include/inn/defines.h
 dbz.o: dbz.c ../include/config.h ../include/inn/defines.h \
   ../include/inn/system.h ../include/inn/options.h ../include/clibrary.h \
Index: /trunk/lib/date.c
--- /trunk/lib/date.c	(revision 9019)
+++ /trunk/lib/date.c	(revision 9082)
@@ -9,5 +9,4 @@
 #include "config.h"
 #include "clibrary.h"
-#include "headers.c"
 #include <ctype.h>
 #include <time.h>

My fault, sorry.

> Given that v2.5.2 has been released for quite a while I'm hoping someone 
> may have come across this issue or worked out what's causing it.

Which means that building INN with libtool is probably not that frequent...

I hope this patch will do the trick for you.  (At least, the build works
for me.)

I have not tested to run INN with libtool support, so in case you see an
unexpected behaviour, I would recommend you to run INN without libtool.
(And to make us aware of the unexpected behaviour.)

Also, I wish to inform you that the release of INN 2.5.3 will come pretty
soon (September) so you may want to wait for it.  Another possibility
is of course to update to INN 2.5.2 with the patch above right now, and
to INN 2.5.3 when you want.

Changes in 2.5.3 are documented here:

P.-S. to Florian:  I will have a look at your work on procbatch
documentation next week.  It will obviously be present in INN 2.5.3.

Julien ÉLIE

« Petite annonce : Sourd rencontrerait sourde pour terrain
  d'entente. »

More information about the inn-bugs mailing list