Debian; Building shared library version; gsl-histogram bug and patch

[Dirk Eddelbuettel <edd at debian dot org>]

I am looking into taking over maintenance for the gsl and gsl-dev packages
for the Debian distribution.  The two packages have effectively been orphaned
since Michael Booth made the one initial release. [BTW: Does anyone here know
where Michael is?  We don't, it seems.]

I would prefer to build a shared library version. I tried a patch I saw in
the mailing list archive, but no luck. I also saw libtoolswitch.pl was
mentioned, and tried it. On both the 0.41.1 tarball sources, and 0.41.1+ from
the CVS archive, it fails with the following error message:
	updating cache ./config.cache
	ltconfig: unrecognized option `--no-reexec'
	Try `ltconfig --help' for more information.
	configure: error: libtool configure failed
	make[1]: *** [config.status] Error 1
	make[1]: Leaving directory `/home/edd/debian/gsl-0.4.1'
	make: *** [build-stamp] Error 2

I have an up-to-date Linux box (running Debian potato, the 2.2 release
targetted for January).  I am at a loss here as tried removing the
--no-reexec call from configure and configure.in, but no luck. Could anyone
help?


On another note, Debian received a bug report regarding gsl-histogram. I am
not sure if that was ever forwarded to you guys. The report is at 
	http://cgi.debian.org/cgi-bin/bugreport.cgi?bug=42116

The problem is that the last input element is counted twice:
edd@sonny:~> echo 11 12 13 | /usr/bin/gsl-histogram 11 14 3
11 12 1
12 13 1
13 14 2      <-- wrong, the last datum was counted twice


This problem is not caught by `cd histogram; make check'.  However, a simple
patch ...

--- gsl-histogram.c	Thu Nov 18 22:15:03 1999
+++ gsl-histogram.c.orig	Wed Sep 16 17:06:42 1998
@@ -39,8 +39,7 @@
 	double x;
 	status = fscanf (stdin, "%lg", &x);
 
-	if (!feof(stdin))
-	  gsl_histogram_increment (h, x);
+	gsl_histogram_increment (h, x);
 
       }
     while (status == 1);



... seems to fix the problem:

edd@sonny:~> echo 11 12 13 | gsl-histogram.new 11 14 3
11 12 1
12 13 1
13 14 1 


Best regards,  Dirk

                                                  



-- 
According to the latest figures, 43% of all statistics are totally worthless.